64bit linux in stable


Advanced search

Message boards : Number crunching : 64bit linux in stable

AuthorMessage
abc@home staff
Forum moderator
Project administrator
Project developer
Send message
Joined: Nov 8 06
Posts: 342
Credit: 44,383
RAC: 0
Message 2064 - Posted 14 Feb 2007 11:17:56 UTC

    Last modified: 14 Feb 2007 11:18:22 UTC

    Enjoy the triple speed of them :)
    Windows 64bit is still a bit tricky but I think will follow soon, next week or so.

    Profile [AF>HFR>RR]Celtar
    Avatar
    Send message
    Joined: Jan 7 07
    Posts: 42
    Credit: 1,751,293
    RAC: 0
    Message 2067 - Posted 14 Feb 2007 13:13:56 UTC - in response to Message 2064.

      Enjoy the triple speed of them :)
      Windows 64bit is still a bit tricky but I think will follow soon, next week or so.


      Thanks Hendrik, i am putting my quad on it tonight.

      As for credits granted which formula have you chosen ? Do you give the same credits included in each wu like in 32 bits application ?

      ____________

      abc@home staff
      Forum moderator
      Project administrator
      Project developer
      Send message
      Joined: Nov 8 06
      Posts: 342
      Credit: 44,383
      RAC: 0
      Message 2069 - Posted 14 Feb 2007 13:51:19 UTC - in response to Message 2067.

        Enjoy the triple speed of them :)
        Windows 64bit is still a bit tricky but I think will follow soon, next week or so.


        Thanks Hendrik, i am putting my quad on it tonight.

        As for credits granted which formula have you chosen ? Do you give the same credits included in each wu like in 32 bits application ?


        Yes, I discussed this some time ago on the forum and on the boinc dev list.
        Nobody thought it was bad to hand out 3x as much credits per time
        unit for 64bit clients compared to other projects, and since you cannot
        grant credit per result in boinc, it's fine by me.
        32bit clients get around the same credits as other projects hand them out,
        so 64bit can be seen as a super optimization.

        Profile [AF>HFR>RR]Celtar
        Avatar
        Send message
        Joined: Jan 7 07
        Posts: 42
        Credit: 1,751,293
        RAC: 0
        Message 2070 - Posted 14 Feb 2007 13:58:22 UTC - in response to Message 2069.

          I agree with that. Thanks Hendrik
          ____________

          Profile clownius
          Forum moderator
          Avatar
          Send message
          Joined: Jan 14 07
          Posts: 242
          Credit: 6,347,407
          RAC: 0
          Message 2071 - Posted 14 Feb 2007 14:04:01 UTC

            Makes me want to hurry up and start running 64bit lol. Maybe when i finally get my Quad core machine built. I don't want to take the C2D offline until i know 64bit works for me lol.
            Any word if and official BOINC 64bit will become available?
            ____________

            abc@home staff
            Forum moderator
            Project administrator
            Project developer
            Send message
            Joined: Nov 8 06
            Posts: 342
            Credit: 44,383
            RAC: 0
            Message 2073 - Posted 14 Feb 2007 14:12:26 UTC - in response to Message 2071.

              Makes me want to hurry up and start running 64bit lol. Maybe when i finally get my Quad core machine built. I don't want to take the C2D offline until i know 64bit works for me lol.
              Any word if and official BOINC 64bit will become available?


              All I can say is they run really smoothly on 64bit, I did dozens of workunits
              in the past few hours, the longest one was around 30 mins I think.

              Astro
              Avatar
              Send message
              Joined: Nov 21 06
              Posts: 102
              Credit: 601,226
              RAC: 0
              Message 2078 - Posted 14 Feb 2007 15:13:53 UTC - in response to Message 2071.

                Makes me want to hurry up and start running 64bit lol. Maybe when i finally get my Quad core machine built. I don't want to take the C2D offline until i know 64bit works for me lol.
                Any word if and official BOINC 64bit will become available?

                I haven't even seen an official 64b Boinc being discussed. I run other projects, and so won't run any 64b Boinc which has "altered" benchmarks, but that seems to be the only source available so far. That means I won't be doing any 64b with windows.

                Profile Webmaster Yoda
                Avatar
                Send message
                Joined: Dec 31 06
                Posts: 81
                Credit: 4,544,249
                RAC: 0
                Message 2095 - Posted 15 Feb 2007 0:54:37 UTC - in response to Message 2078.

                  Last modified: 15 Feb 2007 0:58:24 UTC

                  I haven't even seen an official 64b Boinc being discussed. I run other projects, and so won't run any 64b Boinc which has "altered" benchmarks, but that seems to be the only source available so far. That means I won't be doing any 64b with windows.


                  I can't find a Linux 64 bit BOINC client to run on my Core 2 Duo. Seen one somewhere which appeared to be for AMD64 but that's all. Perhaps it would work on a Core 2 Duo, but it's not obvious as that is not an AMD processor...

                  I saw a link to another one on a message board (may have been ABC Beta) but it was a dead link :(
                  ____________


                  Join the #1 Aussie Alliance on ABC@Home

                  Profile LiborA
                  Send message
                  Joined: Nov 22 06
                  Posts: 11
                  Credit: 168,977
                  RAC: 0
                  Message 2116 - Posted 15 Feb 2007 20:41:05 UTC - in response to Message 2095.

                    Last modified: 15 Feb 2007 20:42:29 UTC

                    I haven't even seen an official 64b Boinc being discussed. I run other projects, and so won't run any 64b Boinc which has "altered" benchmarks, but that seems to be the only source available so far. That means I won't be doing any 64b with windows.


                    I can't find a Linux 64 bit BOINC client to run on my Core 2 Duo. Seen one somewhere which appeared to be for AMD64 but that's all. Perhaps it would work on a Core 2 Duo, but it's not obvious as that is not an AMD processor...

                    I saw a link to another one on a message board (may have been ABC Beta) but it was a dead link :(


                    Link from Augustine in beta forum works correct. You can find it also there http://abcathome.com/forum_thread.php?id=47&nowrap=true#2054

                    Profile Webmaster Yoda
                    Avatar
                    Send message
                    Joined: Dec 31 06
                    Posts: 81
                    Credit: 4,544,249
                    RAC: 0
                    Message 2119 - Posted 15 Feb 2007 22:49:36 UTC - in response to Message 2116.

                      Link from Augustine in beta forum works correct. You can find it also there http://abcathome.com/forum_thread.php?id=47&nowrap=true#2054


                      Thanks for that!

                      ____________


                      Join the #1 Aussie Alliance on ABC@Home

                      Profile Crunch3r
                      Avatar
                      Send message
                      Joined: Jan 12 07
                      Posts: 146
                      Credit: 3,500,943
                      RAC: 0
                      Message 2121 - Posted 15 Feb 2007 23:17:51 UTC - in response to Message 2078.


                        I haven't even seen an official 64b Boinc being discussed. I run other projects, and so won't run any 64b Boinc which has "altered" benchmarks, but that seems to be the only source available so far. That means I won't be doing any 64b with windows.


                        Well as long as you keep your "altered" attitude to your self, everything will work fine, same as before you initialized "credit wars".

                        No one forces you to use the client (although nothing has been "altered" but of course we know why you post that cr.. don't we ;-)






                        Profile kevint
                        Avatar
                        Send message
                        Joined: Dec 28 06
                        Posts: 20
                        Credit: 15,454,097
                        RAC: 0
                        Message 2125 - Posted 16 Feb 2007 2:56:21 UTC

                          Last modified: 16 Feb 2007 2:59:00 UTC


                          So, with the 64 bit app - will we need to run a 64 bit BOINC client as well ?



                          EDIT: Never mind - I found the answer - DOH!

                          Profile kevint
                          Avatar
                          Send message
                          Joined: Dec 28 06
                          Posts: 20
                          Credit: 15,454,097
                          RAC: 0
                          Message 2126 - Posted 16 Feb 2007 2:56:50 UTC - in response to Message 2078.

                            I haven't even seen an official 64b Boinc being discussed. I run other projects, and so won't run any 64b Boinc which has "altered" benchmarks, but that seems to be the only source available so far. That means I won't be doing any 64b with windows.



                            So don't. But don't cry about other getting higher credits for running optimized apps or clients.

                            If the project devs say it is ok, then it is ok.

                            Logan5@SETI.USA
                            Send message
                            Joined: Nov 23 06
                            Posts: 6
                            Credit: 3,775
                            RAC: 0
                            Message 2128 - Posted 16 Feb 2007 6:41:33 UTC - in response to Message 2078.

                              Last modified: 16 Feb 2007 7:06:35 UTC

                              I haven't even seen an official 64b Boinc being discussed. I run other projects, and so won't run any 64b Boinc which has "altered" benchmarks, but that seems to be the only source available so far. That means I won't be doing any 64b with windows.


                              Hmmmmm is it me or is this the beginning of the same old stink, just in a different place?

                              @Tony:

                              WTF are you doing dredging up past flames? Leave well enough alone already and let the man do what he does best for the community. The Dev's here don't mind Crunch3rs work, so please get over it and move on before you start something you will not be allowed to win.

                              It's beginning to look like you have a personal grudge against Crunch3r considering that you don't seem to know when to leave well enough alone and seem to be following him from project to project.. Are you stalking him? If you are that's just wrong and more then a little creepy....

                              NO ONE is forcing you to use any optimized client so what's your beef? Would you be acting this critical of Simon & HIS apps....? I think not.

                              I have nothing personal against you, but when you flame someone using words like "altered" when describing Crunch3r's work, which implies he has done something that is somehow wrong, you lose what little credibility you might have had before the words came out of your mouth and that's what I as well as many others have a BIG problem with.


                              EDIT: Too bad my 1st post here could not have been under better circumstances.... :/

                              ____________

                              Astro
                              Avatar
                              Send message
                              Joined: Nov 21 06
                              Posts: 102
                              Credit: 601,226
                              RAC: 0
                              Message 2131 - Posted 16 Feb 2007 11:30:16 UTC

                                Last modified: 16 Feb 2007 11:58:24 UTC

                                Gee, notice the three negative responses are all from team Seti.USA(same team as Crunch3r), interesting.

                                Logan5, until Crunch3r stops destroying/trying to destroy the boinc credit system I will continue to inform possible users that his products have in the past (including 5.9.0-32 and 5.9.0.64) produce benchmark optimizations that OVERCLAIM. (note: 5.9.0.64 on my AMD64 5200 weren't nearly as high as his 32b versions, yet still 50% higher, 2621/4824official vs his 2567/6630) He lacks any sense of responsibility for his actions including modification and distribution. The use of these products were falsely justified by others when they came out at Seti, because they offset the gain in speed his optimized applications produced. I say "falsely justified", because the claims were 33% higher than they should have been even then. Now however there are NO optimized apps made by him that might justify the use of a modified "Altered" Boinc Core Client.

                                Crunch3r, What's the justification for the altered benchmarks NOW???? I've ran a few of them on my AMD64 2800.

                                5.9.0.32-32bit 30.867 credits/hour
                                2/15/2007 7:45:04 PM||BOINC 5.9.0.32 - 32 bit Windows Edition by Crunch3r
                                2/15/2007 7:45:04 PM||enabled features:
                                2/15/2007 7:45:04 PM||-cpu_affinity
                                2/15/2007 7:45:04 PM||-return_results_immediately
                                2/15/2007 7:45:04 PM||Processor: 1 x AuthenticAMD AMD Athlon(tm) 64 Processor 2800+
                                2/15/2007 7:45:04 PM||Version change (5.8.11 -> 5.9.0.32); NOT running CPU benchmarks
                                2/15/2007 7:46:19 PM|| 4624 floating point MIPS (Whetstone) per CPU
                                2/15/2007 7:46:19 PM|| 10192 integer MIPS (Dhrystone) per CPU

                                5.2.13 SSE 17.327 credits/hour
                                2/15/2007 10:28:58 PM||Starting BOINC client version 5.2.13 for windows_intelx86
                                2/15/2007 10:28:58 PM||Processor: 1 AuthenticAMD AMD Athlon(tm) 64 Processor 2800+
                                2/15/2007 10:30:00 PM|| 1988 double precision MIPS (Whetstone) per CPU
                                2/15/2007 10:30:00 PM|| 6329 integer MIPS (Dhrystone) per CPU

                                5.2.13 SSE2 20.896 credits/hour
                                2/15/2007 10:31:25 PM||Starting BOINC client version 5.2.13 for windows_intelx86
                                2/15/2007 10:31:25 PM||Processor: 1 AuthenticAMD AMD Athlon(tm) 64 Processor 2800+
                                2/15/2007 10:33:12 PM||Benchmark results:
                                2/15/2007 10:33:12 PM|| 2469 double precision MIPS (Whetstone) per CPU
                                2/15/2007 10:33:12 PM|| 7561 integer MIPS (Dhrystone) per CPU

                                5.5.0 SSE2 27.063 credits/hour
                                2/15/2007 10:34:58 PM||Starting BOINC client version 5.5.0 for windows_intelx86
                                2/15/2007 10:34:58 PM||Processor: 1 AuthenticAMD AMD Athlon(tm) 64 Processor 2800+
                                2/15/2007 10:34:58 PM||Version change (5.2.13 -> 5.5.0); running CPU benchmarks
                                2/15/2007 10:35:59 PM|| 3057 floating point MIPS (Whetstone) per CPU
                                2/15/2007 10:35:59 PM|| 9933 integer MIPS (Dhrystone) per CPU

                                5.5.0SSE 25.915 credits/hour
                                2/15/2007 10:38:15 PM||Starting BOINC client version 5.5.0 for windows_intelx86
                                2/15/2007 10:38:15 PM||Processor: 1 AuthenticAMD AMD Athlon(tm) 64 Processor 2800+
                                2/15/2007 10:40:18 PM|| 2550 floating point MIPS (Whetstone) per CPU
                                2/15/2007 10:40:18 PM|| 9889 integer MIPS (Dhrystone) per CPU

                                5.8.12 Original Boinc 10.125 credits/hour
                                2/15/2007 10:44:03 PM||Starting BOINC client version 5.8.12 for windows_intelx86
                                2/15/2007 10:44:03 PM||Processor: 1 AuthenticAMD AMD Athlon(tm) 64 Processor 2800+ [x86 Family 15 Model 4 Stepping 10] [fpu tsc pae nx sse sse2 3dnow mmx]
                                2/15/2007 10:45:18 PM|| 1704 floating point MIPS (Whetstone) per CPU
                                2/15/2007 10:45:18 PM|| 3156 integer MIPS (Dhrystone) per CPU

                                Note: these are displayed in order run, and all were run yesteday. Compare ANY of them to the very bottom one which is the latest Official Boinc Client.

                                Again, I will continue to warn people about his optimized Boinc Core Clients, because he continually refuses to be responsible and do so himself. TMR had speedy67 post a warning, he even asked him to remove his products when they were affecting seti beta. Where's Crunch3r's warning??? Has never been on on his site.

                                I don't follow crunch3r. Look at the ABCbeta site My userid is lower than his. His first post was 34 days ago, mine was 86 days ago. Might wanna check some facts before jumping in feet first. It seems you got tricked by your fellow team members to jump into the water without checking the depth first. (I.E gathering your facts)

                                There IS one version produced by him that doesn't alter the benchmarks as far as I can tell and that's 5.4.11 for AMD64. To me that makes it even worse for him, since he can't claim it's not his fault for all the others with "altered" benchmarks. (I.E this proves he knows how NOT to alter them), so why does he alter them?????

                                I didn't start any credit war, if he'd never made the altered versions, then they would never have happened on any project. Bringing these facts to light doesn't mean I started them. He was asked to include a disclaimer in December 2005, he was asked to intall a mechanism that would prevent their use on other projects, he was even questioned about his benchmark methodology by Paul D Buck (creator of the wiki), and was told by Crunch3r to "stick to your wiki".

                                Seti has a validation process which will weed out errant claims, and it works as long as the percentage of users is low. As that percentage grew there were ever increasing instances of two resultids claiming falsely and being granted. (I.E read destruction of the integrity of the credit process)

                                Let's look at what has happened since;
                                Dr. Anderson and Seti spent good time and precious money creating the Fpops credit system to stop the overclaiming and bring credit back into equalization. This removed credit calculation from Boinc and placed it in the science application. What does Crunch3r do? He starts optimizing the seti application at seti beta. He made version 5.11 which had a "load store adjustment" equal to the earlier version by Eric K(7-9 when the release seti version was 3.35). No harm there, however he left it available on his site for more than 3 weeks after knowing it was an improper value. If it was available, you know people would download and use it, especially since it requested more than twice the normal credit. After some time it would have also been used by enough users to once again destroy the credit system. Yet again, not a very responsible act. Then he made a version which contained a "load store adjustment of 3.36" (3.35 is the official loadstore adjustment today at seti). He had to go for that little bit extra.

                                It was at this point I asked him "crunch3r? are you trying to cheat". I mean for cripes sake, seti had just implemented a system to create fair credit for all, and here he was offering products that could destroy all their hard work.

                                In December 05 he was asked about the overclaiming at Rosetta that his versions were causing. He responded "It's Rosetta's fault for not being open source". This was causing credit/flame wars at Rosetta before I had even crunched my first wu for them, or posted there.

                                His Boinc optimizations have caused Einstein to invest the time and money to implement a fixed credit system. Caused flamewars at Rosetta, and caused them to implement a "rolling average" credit system. Look across the boards and see WHO's name is attached to all the credit wars. You know who it is..the creator of 5.5.0 that's who. Ask Hendrik if he hasn't spent time/money working on this issue.

                                So, is the person responsible for the credit wars the person who made the code or the person who tries to lesten the damage to boinc caused by them? I wonder how many man hours have been spent trying to correct this??

                                He has been repeatedly advised that RRI (return result immediately) is hard on the project servers, but continues to include it in his code. Rom Walton (boinc developer with Berkeley) has written BOINC Client: The evils of 'Returning Results Immediately' and FOLLOW-UP: The evils of 'Returning Results Immediately'. Yet he continues his irresponsible actions by keeping this in his newer clients. (see 5.9.0.32 log above).

                                On the brighter side, I see cpu affinity is the lone brightspot in his builds. If he won't be responsible and inform users of the downside of his builds, and preferably stop altering the benchmarks, then someone has to.


                                [edit]Of course I won't say anything negative about Simon Zadra, he doesn't do anything to hurt boinc or destroy the credit process. Silly[/edit]

                                Profile Scooter
                                Send message
                                Joined: Nov 21 06
                                Posts: 15
                                Credit: 37,521
                                RAC: 0
                                Message 2134 - Posted 16 Feb 2007 13:20:44 UTC

                                  I see this is one of those threads thats going to give me a headache if I keep reading it!!!

                                  Logan5@SETI.USA
                                  Send message
                                  Joined: Nov 23 06
                                  Posts: 6
                                  Credit: 3,775
                                  RAC: 0
                                  Message 2138 - Posted 16 Feb 2007 15:15:30 UTC

                                    @ Tony:

                                    It's clear now that you are biased so I will say this as clearly as I can:

                                    *IF* you can find one project Developer or Scientist that agrees with you that Crunch3r's optimized application for abc@home is somehow harming the project or the integrity of the data collected, then I will be the FIRST to convert and become your disciple and go with you from project to project telling/warning people that Crunch3r is doing more harm then good..

                                    Now, if you can't find anyone who agrees with you then you should probably move on as there will be about as much support for you then as there would be now which is close to zero.

                                    I am not deluded by my fellow team mates like you tried to imply, rather the opposite as I do see thing and people for what they are which is why I chose to respond to your initial "altered" slander in the manner I did.

                                    The burden of proof's squarely with you now so you better get to emailing those devs and scientists.... :/ LOL

                                    Astro
                                    Avatar
                                    Send message
                                    Joined: Nov 21 06
                                    Posts: 102
                                    Credit: 601,226
                                    RAC: 0
                                    Message 2140 - Posted 16 Feb 2007 16:09:36 UTC

                                      Logan5, It appears you don't dispute anything I've previously presented. You haven't disputed any of them. Instead, you intend to shift the conversation to a different issue. I'd agree that the altered benchmarks won't affect ABC, the RRI will increasingly have negative affects after the project grows.

                                      What happens when/if Hendrik endorses the use of a third party Boinc CC which alters the benchmark code???? I'll tell you, others will use it and most will attach to other projects. This would further the degradation of the credit system. I have NO issue with ones that don't alter benchmarks and don't have RRI.

                                      When "third party Boinc CC's" first came out noone wanted the negative affects discussed. Discussing/mentioning them was somehow an attack. And through ignorance(lack of anyone telling them they overclaim) and greed for credit their use spread. Now you don't want anyone to inform others of potential hazards with 64b boinc CC's? It's almost like you don't want anyone to make an informed decision for themselves.

                                      Going with 64b makes sense for ABC given their use of 64b numbers. Unfortunately, this places him on the "bleeding edge" of Boincs capabilities, so he's "leading" the show. He'll be setting precedence. He needs to make informed decisions.

                                      If I were to be quiet, who'd give him the other side of the story????? It'd be the same as the 32b early Third party boinc clients "Man, you gotta get X", "you need to try this", all with NO warnings applied. Keeping secret the facts. Kinda like the note included with the Team Xtremesystems version of 5.5.0.

                                      "XTREMESYSTEMS.ORG CONFIDENTIAL

                                      Keep these updates & optimizations secret. We don't want other teams finding out, hint hint =D."



                                      Astro
                                      Avatar
                                      Send message
                                      Joined: Nov 21 06
                                      Posts: 102
                                      Credit: 601,226
                                      RAC: 0
                                      Message 2141 - Posted 16 Feb 2007 16:17:02 UTC

                                        Also, you may have missed my post about not finding Crunch3r's 5.4.11_AMD64 to alter the benchmarks. I also notice it has "cpu affinity" (a bonus), AND I don't see any text in the message log that indicates RRI is used, so guess what? I might actually use it. If I find no hidden "gotcha's", I might even endorse it. Due to the "dual core benchmark bug" (dhrystone varying from 1/2, 3/4 and full value on a seeming random basis) I ran 25-30 benchmarks with it and never saw a value with exceeds what it should have. That doesn't mean it doesn't exist, that's what I mean by more testing and looking to see if RRI is working but not listed.

                                        Profile miw
                                        Avatar
                                        Send message
                                        Joined: Jan 21 07
                                        Posts: 32
                                        Credit: 2,780,753
                                        RAC: 0
                                        Message 2146 - Posted 16 Feb 2007 18:18:09 UTC

                                          Last modified: 16 Feb 2007 18:19:01 UTC

                                          There is a lot I could say on this topic, but I'll restrict myself to this:

                                          The benchmark system has been so thoroughly rorted that the credit in projects that still rely on boinc client benchmarks for the allocation of credit has become completely debased and I rarely crunch them anymore. QMC is a good example here. It became a total joke.

                                          Much, although not all, of the blame for this lies with the use (Not just the creation and distribution) of boinc clients that produce ridiculously inflated benchmarks. The only reason for using these clients is to get more credit for doing less crunching. There is no ethical or reasonable justification.


                                          ____________
                                          --miw

                                          Astro
                                          Avatar
                                          Send message
                                          Joined: Nov 21 06
                                          Posts: 102
                                          Credit: 601,226
                                          RAC: 0
                                          Message 2149 - Posted 16 Feb 2007 20:37:58 UTC

                                            Dang, I just ran Crunch3r's 5.8.11.64-64b and got the following:

                                            2/16/2007 2:27:52 PM||Starting BOINC client version 5.8.11.64 for windows_amd64
                                            2/16/2007 2:27:52 PM||BOINC 5.8.11.64 - 64 bit Windows Edition by Crunch3r
                                            2/16/2007 2:27:52 PM||enabled features:
                                            2/16/2007 2:27:52 PM||-cpu_affinity
                                            2/16/2007 2:27:52 PM||Processor: 2 AuthenticAMD AMD Athlon(tm) 64 X2 Dual Core Processor 5200+ [AMD64 Family 15 Model 67 Stepping 2] [fpu tsc pae nx sse sse2]
                                            2/16/2007 2:27:52 PM||Version change (5.8.12 -> 5.8.11); running CPU benchmarks
                                            2/16/2007 2:28:54 PM|| 2484 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:28:54 PM|| 4170 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:32:41 PM|| 2489 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:32:41 PM|| 4174 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:33:56 PM|| 2488 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:33:56 PM|| 4179 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:35:17 PM|| 2487 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:35:17 PM|| 6362 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:38:14 PM|| 2485 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:38:14 PM|| 4175 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:39:22 PM|| 2474 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:39:22 PM|| 4038 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:46:24 PM|| 2485 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:46:24 PM|| 4168 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:47:30 PM|| 2488 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:47:30 PM|| 4178 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:54:29 PM|| 2482 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:54:29 PM|| 4174 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:55:37 PM|| 2480 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:55:37 PM|| 4020 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:56:46 PM|| 2477 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:56:46 PM|| 4109 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 2:58:43 PM|| 2472 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 2:58:43 PM|| 4063 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 3:05:40 PM|| 2487 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 3:05:40 PM|| 3207 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 3:06:58 PM|| 2485 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 3:06:58 PM|| 3194 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 3:08:17 PM|| 2489 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 3:08:17 PM|| 3215 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 3:09:29 PM|| 2485 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 3:09:29 PM|| 3174 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 3:10:45 PM|| 2489 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 3:10:45 PM|| 3190 integer MIPS (Dhrystone) per CPU
                                            2/16/2007 3:11:58 PM|| 2485 floating point MIPS (Whetstone) per CPU
                                            2/16/2007 3:11:58 PM|| 3174 integer MIPS (Dhrystone) per CPU

                                            As you can see only cpu affinity is listed, so I'd actually have to get work to know if RRI is enabled. Below are the 32b windows benchmarks across different versions. Note: this was used to confirm the existence of the "dual benchmark bug" As you can see both versions have wildly varying Dhrystone measurements. This is due to Boinc, not Crunch3r. The max 32B Dhrystone was 4800+, the max Dhrystone on Crunch3r's version was 6300+. Given the number of available SSE registers are doubled in 64b, I think the benchmarks for 5.8.11.64 appear normal.

                                            5.8.4
                                            2632/2468
                                            2631/2618
                                            2627/2472
                                            2624/2475

                                            Being curious I installed 5.8.3 and got
                                            2633/3146
                                            and
                                            2632/3146

                                            I reverted to 5.7.5
                                            2635/4867
                                            2634/3348
                                            2636/3271

                                            Then to 5.4.9
                                            2640/4859
                                            2638/4846
                                            2636/4858

                                            Then to 4.72
                                            2629/4842
                                            2633/3280
                                            2634/3286

                                            Then to 4.43
                                            2620/4845
                                            2627/2483
                                            2627/4862

                                            Then to 4.20
                                            2622/4845
                                            2623/4856
                                            2662/3284

                                            Then back to 5.8.4
                                            2630/4613
                                            2634/4674
                                            2636/4686

                                            I hope this trend continues and I can eventually stop commenting upon "third party boinc clients" with altered benchmarks. It will take time for all of them to filter through the system(I.E become dropped , unuseable via age, or obsolete).

                                            I apologize to Crunch3r for automatically assuming these versions contained altered benchmarks and will be following future developments with interest to ensure my accuracy. This marks a departure which I view favorably. I'll even be rerunning 5.9.0.64 with more samples.

                                            tony

                                            KEP
                                            Send message
                                            Joined: Jan 16 07
                                            Posts: 29
                                            Credit: 6,287
                                            RAC: 0
                                            Message 2150 - Posted 16 Feb 2007 21:03:02 UTC

                                              People people people, well at least all your postings out of frustration at least gave me a headeache, so I skipped some of them if they looked repetorious. About the credit, I think of course 25+/credit an hour for a pc normally doing 10±(10-20%) credit an hour, sounds a bit to much for me, but all lets remember that it really doesn't matter about the credit, but about the results we produce while crunching. And if the credit is based on the amount of hits, well then the credit will be fair, because then the 64bit app will yield more triplets than the 32bit, and of course then everyone doing 64bit should be granted more credit, on the other hand if the credit isn't based on the findings, then there needs to be done some sort of adjustement, so the credit is more suited for the speed of the computer, rather than for what app it crunches with :)

                                              Now finally let me remind you americans and none americans, that this is a European funded project, and up untill now I haven't seen any bigger flaming wars, and frankly I don't wish to see it now or in the future, so even though I'm no moderator, I'll gently earch all of you to reach out for each other, and make peace in stead of flaming your opponents, because in the end no one wins on doing this ;) So respect your next and his or shes oppinion, and then defend your right to free speach, if you aren't met with respect :)

                                              Regards!

                                              KEP

                                              [AF>Belgique] Jerom007
                                              Send message
                                              Joined: Feb 3 07
                                              Posts: 5
                                              Credit: 18,472
                                              RAC: 0
                                              Message 2152 - Posted 16 Feb 2007 23:48:00 UTC

                                                After quite some efforts I could get the client 5.4.11 amd64 running under Linux Ubuntu 6.10 (kernel 2.6.17.11 / client 5.8.11 not apparently, other people did?)
                                                on my C2D T7200 compatible EM64T/amd64

                                                And got thoses errors "code 131", is it normal? or Intel again lagging behind Amd !? ;-)
                                                Sat 17 Feb 2007 12:29:10 AM CET|ABC@home|Unrecoverable error for result abc_wu_38382310000_100000_1 (process exited with code 131 (0x83))
                                                Sat 17 Feb 2007 12:29:10 AM CET|ABC@home|Deferring scheduler requests for 1 minutes and 0 seconds
                                                Sat 17 Feb 2007 12:29:10 AM CET||Rescheduling CPU: application exited
                                                Sat 17 Feb 2007 12:29:10 AM CET|ABC@home|Computation for task abc_wu_38382310000_100000_1 finished
                                                Sat 17 Feb 2007 12:29:11 AM CET|ABC@home|Unrecoverable error for result abc_wu_38382510000_100000_0 (process exited with code 131 (0x83))

                                                (maybe that's just when I activated "Run based on preferences" and he halted the current WUs? they are lost in that case!?... I remember some mention to that in the forum I think...)

                                                Logan5@SETI.USA
                                                Send message
                                                Joined: Nov 23 06
                                                Posts: 6
                                                Credit: 3,775
                                                RAC: 0
                                                Message 2153 - Posted 17 Feb 2007 0:29:58 UTC - in response to Message 2149.

                                                  I apologize to Crunch3r for automatically assuming these versions contained altered benchmarks and will be following future developments with interest to ensure my accuracy.

                                                  We're all glad that this development meets with your seal of approval.

                                                  Thanks for being man enough to admit you have been wrong about this, as that takes a lot of courage to state in open forum and for that I do commend you for your willingness to end the war between yourself and Crunch3r.

                                                  You've taken the 1st steps today to repair the rift the prior & current accusations of cheating have caused in the community.

                                                  ____________

                                                  Astro
                                                  Avatar
                                                  Send message
                                                  Joined: Nov 21 06
                                                  Posts: 102
                                                  Credit: 601,226
                                                  RAC: 0
                                                  Message 2154 - Posted 17 Feb 2007 1:24:51 UTC - in response to Message 2152.

                                                    After quite some efforts I could get the client 5.4.11 amd64 running under Linux Ubuntu 6.10 (kernel 2.6.17.11 / client 5.8.11 not apparently, other people did?)
                                                    on my C2D T7200 compatible EM64T/amd64

                                                    (maybe that's just when I activated "Run based on preferences" and he halted the current WUs? they are lost in that case!?... I remember some mention to that in the forum I think...)


                                                    I'm afraid I don't understand. the client 5.4.11.64 AMD64 is a windows only boinc CC. Augustines version for linux is boinc_5.8.11_x86_64-pc-linux-gnu and boinc_5.4.11_x86_64-pc-linux-gnu. Are you saying you're attempting to run the windows version under linux?

                                                    Error code 131 reads:

                                                    ERR_FILE_TOO_BIG -131 file size too big an output file was bigger than max_nbytes

                                                    If you're trying to run it under linux then you need to install the normal linux boinc.sh from Berkeley (5.4.11_i686-pc-linux-gnu.sh, for example), then download and extract one of Augustines boincs. Untar it, then rename the boinc executable in the boinc folder to something else, and replace it with the untarred Augustine version. restart boinc, and you should be done.

                                                    I need help understanding your needs.

                                                    tony

                                                    Astro
                                                    Avatar
                                                    Send message
                                                    Joined: Nov 21 06
                                                    Posts: 102
                                                    Credit: 601,226
                                                    RAC: 0
                                                    Message 2155 - Posted 17 Feb 2007 1:55:54 UTC - in response to Message 2153.

                                                      I apologize to Crunch3r for automatically assuming these versions contained altered benchmarks and will be following future developments with interest to ensure my accuracy.

                                                      We're all glad that this development meets with your seal of approval.

                                                      Thanks for being man enough to admit you have been wrong about this, as that takes a lot of courage to state in open forum and for that I do commend you for your willingness to end the war between yourself and Crunch3r.

                                                      You've taken the 1st steps today to repair the rift the prior & current accusations of cheating have caused in the community.

                                                      Why thank you Logan5, you have always seemed to want to play the moderator. You continue to ask people to "take a different tack" or "tone down messages". To play the moderator,you need to moderate between two parties or there's nothing to moderate.

                                                      Your post leaves me feeling a little empty. Shouldn't you be posting words encouraging Crunch3rs continued recent behavior in this regard? That would be moderation, without it, it's a little one sided, don't you think?

                                                      Logan5@SETI.USA
                                                      Send message
                                                      Joined: Nov 23 06
                                                      Posts: 6
                                                      Credit: 3,775
                                                      RAC: 0
                                                      Message 2156 - Posted 17 Feb 2007 3:58:54 UTC - in response to Message 2155.

                                                        Last modified: 17 Feb 2007 4:03:22 UTC

                                                        Why thank you Logan5, you have always seemed to want to play the moderator. You continue to ask people to "take a different tack" or "tone down messages". To play the moderator,you need to moderate between two parties or there's nothing to moderate.

                                                        Your post leaves me feeling a little empty. Shouldn't you be posting words encouraging Crunch3rs continued recent behavior in this regard? That would be moderation, without it, it's a little one sided, don't you think?


                                                        This whole thread is one sided after YOU took it WAY OFF TOPIC. You accused someome of cheating the claimed credit and then when you actually bothered to try his latest optimized application you discovered that you were wrong.....

                                                        I gave you a sincere compliment for admitting publicly that you were wrong about an erroneous conclusion you had reached but you are either too stubborn or too ignorant to see it for what it really was....After reading your comments above I wonder why I even bothered to think you are a decent human being. You just proved once again what a complete and totallly insincere person you are... I regret my prior comments and wish I hadn't thought that you were actually sorry for what you had said before about Crunch3r because you have just shown that you could care less...

                                                        KEP
                                                        Send message
                                                        Joined: Jan 16 07
                                                        Posts: 29
                                                        Credit: 6,287
                                                        RAC: 0
                                                        Message 2157 - Posted 17 Feb 2007 5:06:25 UTC - in response to Message 2156.

                                                          Why thank you Logan5, you have always seemed to want to play the moderator. You continue to ask people to "take a different tack" or "tone down messages". To play the moderator,you need to moderate between two parties or there's nothing to moderate.

                                                          Your post leaves me feeling a little empty. Shouldn't you be posting words encouraging Crunch3rs continued recent behavior in this regard? That would be moderation, without it, it's a little one sided, don't you think?


                                                          This whole thread is one sided after YOU took it WAY OFF TOPIC. You accused someome of cheating the claimed credit and then when you actually bothered to try his latest optimized application you discovered that you were wrong.....

                                                          I gave you a sincere compliment for admitting publicly that you were wrong about an erroneous conclusion you had reached but you are either too stubborn or too ignorant to see it for what it really was....After reading your comments above I wonder why I even bothered to think you are a decent human being. You just proved once again what a complete and totallly insincere person you are... I regret my prior comments and wish I hadn't thought that you were actually sorry for what you had said before about Crunch3r because you have just shown that you could care less...


                                                          Apparently you didn't understand much of my previous post. Now please notice that it is the second time that your attack Tony, without him having attacked you once just yet. You were after all the first one to write in a utterly ugly language, though hiding it in shortened and covered language. Example given you replied to tonys first post with WTF, which just shows how you deterred this thread. Anyway I'm not going to get into this argument, but I think that either people should post positiv and happy posts, or else they shouldn't bother to post anything at all. After all we are all crunching for the same, and that is to see new discoveries being made, and by attacking other crunchers either because they makes mistakes and claim something wrong or simply because they post things one doesn't like, then one is hurting that cause. Now all please settle and use this thread to what it was meant for, and give Hendrik feedback about any issues relating with the 64b application, and then take your apparent war somewhere else.

                                                          Tony, I appreciate your right to be here, and since you posted because you felt there were a problem with the credit in the first way, I strongly urge you to ignore the seti.USA team, and post only responses to other posts or when/if there is a problem to report :) This should be considered a friendly advice, and is not meant to rob you your right to free speach, but it can truly help us get peace on this message board again. As mentioned earlier, this is a European Project, and they typically doesn't have the same popularity as American projects, so peace is extra wishfull at the European projects, such as we can maintain a ever growing number of participants :)

                                                          Can you all respect each other and others oppinion, and then have a really nice crunching day ;)

                                                          Regards!

                                                          KEP
                                                          ____________
                                                          On August 10. 2008 at 17:15 UTC
                                                          -------------------------------------------------
                                                          My highest WU-id: 1249971
                                                          Triplets done: 44.435.258.262.051
                                                          ABC-triplets found: 6.347.692
                                                          Current progress: 0,0088870516524102 %
                                                          Error rate: 6,107959387850 %

                                                          Logan5@SETI.USA
                                                          Send message
                                                          Joined: Nov 23 06
                                                          Posts: 6
                                                          Credit: 3,775
                                                          RAC: 0
                                                          Message 2158 - Posted 17 Feb 2007 5:43:58 UTC - in response to Message 2157.

                                                            @KEP:

                                                            I appreciate your position and respect it but you do not understand that Tony has a rather bad history at the seti@home message boards for making the same types of slanderous accusations against Crunch3r where he was accusing him of writing optimized applications that 'cheated' by artifically inflating the claimed credit.

                                                            He is doing the same thing here once again and he needs to be reminded that this behavior should and will not be tolerated in the Distributed Computing Community.

                                                            Tony's claims were proven ultimately false at the seti boards only AFTER he caused a HUGE rift in the membership by forcing people to take sides....

                                                            This is the history of the person who you think I am picking on, but if you really understood that history you would maybe not be so quick to come to his defense..

                                                            You are right KEP that this needs to end, but what needs to end is Tonys witch hunt of Crunch3r which I *thought* had ended when he publicly apologized..

                                                            [AF>Belgique] Jerom007
                                                            Send message
                                                            Joined: Feb 3 07
                                                            Posts: 5
                                                            Credit: 18,472
                                                            RAC: 0
                                                            Message 2159 - Posted 17 Feb 2007 7:42:32 UTC - in response to Message 2154.

                                                              [quote]Error code 131 reads:

                                                              ERR_FILE_TOO_BIG -131 file size too big an output file was bigger than max_nbytes

                                                              If you're trying to run it under linux then you need to install the normal linux boinc.sh from Berkeley (5.4.11_i686-pc-linux-gnu.sh, for example), then download and extract one of Augustines boincs. Untar it, then rename the boinc executable in the boinc folder to something else, and replace it with the untarred Augustine version. restart boinc, and you should be done.

                                                              Yes, it is Linux only and I could indeed find how to use Augustine client on other more detailed forums... after some time and efforts! doh :p
                                                              ("apt-get" only worked to have a working Boinc-manager, not via the boinc installer ?!...)
                                                              [windows 64bit *stable* client is not out yet anyway/beta only - other topic than this one "64bit linux in stable"]

                                                              Ok, the error code looks network related... I got it again this morning when re-running manually CPU benchmarks but couldn't isolate more specifically the source of this (smaller than I thought) issue.

                                                              Anyway, Linux is quite painfull to work with at the beginning (maybe I took an unstable new version and should have rather used ubuntu6.06 - same for wifi support...) but I think the efforts were worth it when I see the performance of the abc@home 64bits client!! ;)
                                                              Thanks, bye all + "peace&love". hehe
                                                              J.

                                                              KEP
                                                              Send message
                                                              Joined: Jan 16 07
                                                              Posts: 29
                                                              Credit: 6,287
                                                              RAC: 0
                                                              Message 2161 - Posted 17 Feb 2007 8:32:58 UTC - in response to Message 2158.

                                                                @KEP:
                                                                I appreciate your position and respect it but you do not understand that Tony has a rather bad history at the seti@home message boards for making the same types of slanderous accusations against Crunch3r where he was accusing him of writing optimized applications that 'cheated' by artifically inflating the claimed credit.


                                                                You're right, since I'm not crunching for SETI@home, that I do not know anything to this case, but that just justifys furthermore my pleads to you aswell as Tony not to bring your fuzz to this message board. Could that plead be respected and thereby peace restored?


                                                                He is doing the same thing here once again and he needs to be reminded that this behavior should and will not be tolerated in the Distributed Computing Community.


                                                                I appreciate you wanting to remind Tony, that false acusations shouldn't be brought on any message boards, but you and Tony has a history, which according to my oppinion, makes it better, to let the developers of this project either prove or fail his oppinion. Usually when the officials takes a stand, any fuzz fades away and peace gets restored, because then there is no longer colored oppinions in flict ;)


                                                                Tony's claims were proven ultimately false at the seti boards only AFTER he caused a HUGE rift in the membership by forcing people to take sides....


                                                                It might be, but then there really is no need to take the war to this message board, an official should have been contacted, and then he/she should have solved the issue, if there really indeed is an issue here to deal with...


                                                                This is the history of the person who you think I am picking on, but if you really understood that history you would maybe not be so quick to come to his defense..


                                                                Please don't feel offended, but I'm not beeing his defence, I'll not go into that sort of business, but as far as I see it, he hasn't used angry and bad language, but you have. Also no matter what one feels about what he has to say, one should reply in respect and not anger, else there will be no way to rightfully determine that a reply isn't meant as anything else than a limitation of the counter parts right to free speach. Coming from Denmark, I really knows what free speach is worth, and everyday I feel gratefull, that a civil war werent sparked by the muslims, who has no respect for free speach. So please be aware, with the image of burning embassys, that last year could have been a dreadfull year for free speach, but fortunantly we still have it despite a minority trying to take it from us. So please that's all I'm asking, use the official system, to defeat the oppinions which may be false or inflamious. Because if you, though it is hard, use the official system, and reply in honour, then the support will ultimately be with you, and then you're victorious. Because flaming posters has not much chance on bullying oe flaming, if no one reacts to the bulling and flaming, then they are unsuccesfull, just remember that. But also remember that of course what one has to stand up to, does have a limit, and then again if you feel offended, then take action through the official channels :)


                                                                You are right KEP that this needs to end, but what needs to end is Tonys witch hunt of Crunch3r which I *thought* had ended when he publicly apologized..


                                                                It would be nice, if it could end or else taken back to the boards where it started. Just remember that not everything ends with an apology, which this case is a sad example on...

                                                                Happy crunching everyone

                                                                Regards!

                                                                KEP
                                                                ____________
                                                                On August 10. 2008 at 17:15 UTC
                                                                -------------------------------------------------
                                                                My highest WU-id: 1249971
                                                                Triplets done: 44.435.258.262.051
                                                                ABC-triplets found: 6.347.692
                                                                Current progress: 0,0088870516524102 %
                                                                Error rate: 6,107959387850 %

                                                                Astro
                                                                Avatar
                                                                Send message
                                                                Joined: Nov 21 06
                                                                Posts: 102
                                                                Credit: 601,226
                                                                RAC: 0
                                                                Message 2174 - Posted 17 Feb 2007 13:37:31 UTC

                                                                  Last modified: 17 Feb 2007 14:05:59 UTC

                                                                  Logan5, I think you might need a trip to the optometrists office as seem to have trouble reading with comprehension. So, I'll respond with an outline form to allow you to take in bits at a time without losing your place, and it's a handy for you to respond to points without quoting the entire thread.

                                                                  1) I didn't take this thread off topic. Crunch3r did when he attacked me.

                                                                  My post My post was an answer to the question of Clownius. That was on topic and directly answered his question. NOTE: Crunch3r was not named.

                                                                  then

                                                                  crunch3r started the attack in this post

                                                                  2) If anything you are the one following me, not I following him. I proved I'm not following him in this post. You stated it was unfortunate your "first post" to these boards had to be the way it is. Well, here's your stats:
                                                                  Logan5@SETI.USA
                                                                  Joined: Nov 23, 2006
                                                                  Posts: 5
                                                                  ID: 272
                                                                  Credit: 0
                                                                  RAC: 0
                                                                  You have been a member for months, but haven't contributed ONE wu, and you haven't posted to help people either. Until Crunch3r attacked me that is, then you posted like you always do, and that's to defend your team mate without the use of facts. The facts sure make it look like you're following me, or atleast not really sorry you posted until now.

                                                                  3) Read my apology again. You seem to think it's "broad based", or all encompassing. I apologized for "assuming" all his clients had altered benchmarks. NOTHING more than that. Heck, let's face it when someone does something a dozen times in a row, you kind of have to assume he'll do it the 13th time. In my previous posts to this thread. I never stated "they are all overclaiming". read my comments again and see. However, in my mind I assumed they were and was apologizing for that and that alone. I do hope he continues this trend, and if so, It will abate my comments about how his cc's have destroyed the credibility of the boinc credit system. It won't happen overnite, as his "third party clients" are still in use and affecting projects, but as I said, they will lessen and eventually stop after all his clients are not used for one reason or another. I truly do hope this trend continues.

                                                                  4) In the seti 'incident', I never stated "crunch3r IS cheatiing", I called the thread "crunch3r, are you trying to cheat?". Not the same thing, and throughout that thread (and it's secondaries) I presented facts to support my belief that he was and let the people choose. You seem to think there wasn't enough meat to my presentation. Others would disagree. The main purpose of the thread to was bring to light the fact that he "crunch3r" was once again attempting to destroy the brand spanking new credit system which was "in part" designed to stop the overclaiming of his previous boinc CC's.

                                                                  Fact A) Crunch3r built 5.11 for use on seti beta and even did testing there with it, so he knew what "load store adjustment" was in it.

                                                                  Fact B) I waited and watched his account from day one when seti "enhanced" was released, and he was using 5.11 on the main seti project on his own computers, even though it was known that the "load store adjustment" of the release official app was 3.35, not the 7.00 or 9.00 that was in 5.11.

                                                                  Fact C) I watched his website to see if he'd pulled 5.11 from the download page, as any responsible compiler would.

                                                                  Fact D) After 3 weeks 5.11 was still available for download and still being used on atleast one of his puters. So that's when I started the thread.

                                                                  I will spare the audience the full bit and stop here. He used 5.11 on his own computers even knowing that they claimed high. Nuff said. Case closed, point proven.

                                                                  The really important bit is that OTHERs were using his 5.11, and as prevously stated, that would once again destroy the boinc credit system as it's use grew. Luckily, he took all his toys and went home, pulling his site down, and effectively "saving" the credit system. ironic huh?

                                                                  5) all this is much to familiar. I present 'FACTS' and back them with proof. You an others make claims, distort the truth, and just plain make stuff up, without evidence. You never answer anything and just make more claims, ask question, or change the subject.

                                                                  6) Noone's answered my questions:
                                                                  a) why doesn't he optimized the benchmark code on ALL his Boinc CC's (here's the new improved part)except his 64b boinc clients 5.9.0.64, released Jan 8 2007, 5.4.11_AMD64 released Feb 11 2007, and 5.8.11_AMD64 released Feb 16 2007??? What's the reasoning/rational for them??

                                                                  PS all proof can be found in the old thread and the links within them. If you want them and don't wanna look, I'll present them for you. Some threads were deleted, but they were saved to computer beforehand, so they are still available.

                                                                  Astro
                                                                  Avatar
                                                                  Send message
                                                                  Joined: Nov 21 06
                                                                  Posts: 102
                                                                  Credit: 601,226
                                                                  RAC: 0
                                                                  Message 2177 - Posted 17 Feb 2007 13:50:27 UTC

                                                                    Last modified: 17 Feb 2007 13:52:31 UTC

                                                                    On a side note, which is more on topic, Rom Walton has informed me that a 64b Boinc for windows IS in the works. yeah No timeframe was asked for or given though.

                                                                    Profile Scooter
                                                                    Send message
                                                                    Joined: Nov 21 06
                                                                    Posts: 15
                                                                    Credit: 37,521
                                                                    RAC: 0
                                                                    Message 2178 - Posted 17 Feb 2007 14:50:42 UTC

                                                                      I dont do alot of posting but decided I wanted to make a comment or two...my personal opinion is that this whole conversation is getting a bit on the silly side....if this project isn't using those benchmarks to calculate the credits why should whether otr not they are inflated matter?....if Hendrick approved the use of Crunch3r's client then that should be the end of itshouldn't it?...I will be using it for testing in beta and then when it is all ready in the stable....I think it is a good thing that Crunch3r gave us the opportunity to be able to help advance the work being done and should be thanked for it...if anyone disagrees with me I don't really care but none of the argument said in this thread are going to change my opinion ...and on a side note I think some people here should act more there age...thank you very much and have a nice day!!!!!

                                                                      Dagorath
                                                                      Send message
                                                                      Joined: Jan 7 07
                                                                      Posts: 384
                                                                      Credit: 3,374,905
                                                                      RAC: 0
                                                                      Message 2179 - Posted 17 Feb 2007 14:53:35 UTC - in response to Message 2177.

                                                                        On a side note, which is more on topic, Rom Walton has informed me that a 64b Boinc for windows IS in the works. yeah No timeframe was asked for or given though.


                                                                        Is there anything wrong with using Crunch3r's 64-bit BOINC for Windows? Apparently it neither alters benchmarks nor returns results immediately so it seems to be compliant with existing standards.



                                                                        Astro
                                                                        Avatar
                                                                        Send message
                                                                        Joined: Nov 21 06
                                                                        Posts: 102
                                                                        Credit: 601,226
                                                                        RAC: 0
                                                                        Message 2181 - Posted 17 Feb 2007 15:13:34 UTC - in response to Message 2179.

                                                                          Last modified: 17 Feb 2007 15:40:37 UTC

                                                                          On a side note, which is more on topic, Rom Walton has informed me that a 64b Boinc for windows IS in the works. yeah No timeframe was asked for or given though.


                                                                          Is there anything wrong with using Crunch3r's 64-bit BOINC for Windows? Apparently it neither alters benchmarks nor returns results immediately so it seems to be compliant with existing standards.




                                                                          I downloaded and ran 5.8.11_amd64, and attached to Rieselsieve. It downloaded work and completed 2 wus. They were uploaded, but NOT reported immediately, so I assume RRI is not enabled.

                                                                          I don't know who to blame on the following. when the next two started the cpu time didn't seem to move even though both showed as "running". I looked in Taskmanager and only ONE app was running, system idle process was getting the other 50%. I suspended both those wus and the next two started and ran 30 minutes with NO accumulation in cpu time. I ditched the rest and went back to 32b regular boinc.

                                                                          As I read Augustines posts, the only projects supporting 64b windows is rieselsieve and ABCathome. ABCathome is not currently issuing work. Rieselsieve is, but it's only 32b running in WOW64.

                                                                          I see no issues running 5.8.11_amd64 myself(note: this is not an endorsement, and won't be until further testing is done when ABC has win64 work), except there's no work to be done(see Rieselsieve comment). Perhaps you could take it for a test run and see if rieselsieve performs the same for you. Also, due to the "dual core benchmark bug"(the bug is boincs fault and exists in every version back to 4.05), not Crunch3r's fault), My readings can't be trusted. I wish someone with a single core would do a comparison. A dhrystone increase of 30-50% would seem explainable because of the doubling of the SSE registers.

                                                                          tony

                                                                          [edit]I forgot this, at one point I exitted boinc to see if I could get the cpu time display moving again upon restart. After exitting, BOTH rieselsieve apps were still running and taking 50% of the cpu.

                                                                          Dagorath
                                                                          Send message
                                                                          Joined: Jan 7 07
                                                                          Posts: 384
                                                                          Credit: 3,374,905
                                                                          RAC: 0
                                                                          Message 2188 - Posted 17 Feb 2007 16:46:26 UTC - in response to Message 2181.


                                                                            I'm pretty sure Chess960 sends a 64-bit science app though I must admit I've never tried it. I did try RieselSieve a week ago but could not get it to work, maybe later when I have more time and they offer a 64-bit app. Anyway, it sounds like you have RieselSieve working. Malaria also sends a 32-bit app to 64-bit systems, automatically like Riesel. If Riesel gives you more trouble later then maybe try Malaria.

                                                                            Logan5@SETI.USA
                                                                            Send message
                                                                            Joined: Nov 23 06
                                                                            Posts: 6
                                                                            Credit: 3,775
                                                                            RAC: 0
                                                                            Message 2190 - Posted 17 Feb 2007 17:26:13 UTC - in response to Message 2178.

                                                                              {/snip} if this project isn't using those benchmarks to calculate the credits why should whether otr not they are inflated matter?....if Hendrick approved the use of Crunch3r's client then that should be the end of itshouldn't it?...I will be using it for testing in beta and then when it is all ready in the stable....I think it is a good thing that Crunch3r gave us the opportunity to be able to help advance the work being done and should be thanked for it...if anyone disagrees with me I don't really care but none of the argument said in this thread are going to change my opinion ...and on a side note I think some people here should act more there age...thank you very much and have a nice day!!!!!


                                                                              Thank you for this, it is nice to see support for Crunch3rs work to better the community. He volunteers his time & spends great amounts of his monies for the compilers & software to make his apps and he asks for nothing in return.

                                                                              Profile Scooter
                                                                              Send message
                                                                              Joined: Nov 21 06
                                                                              Posts: 15
                                                                              Credit: 37,521
                                                                              RAC: 0
                                                                              Message 2191 - Posted 17 Feb 2007 18:26:11 UTC - in response to Message 2190.



                                                                                Thank you for this, it is nice to see support for Crunch3rs work to better the community. He volunteers his time & spends great amounts of his monies for the compilers & software to make his apps and he asks for nothing in return.


                                                                                Just saying what I think !!!

                                                                                Dagorath
                                                                                Send message
                                                                                Joined: Jan 7 07
                                                                                Posts: 384
                                                                                Credit: 3,374,905
                                                                                RAC: 0
                                                                                Message 2194 - Posted 17 Feb 2007 19:28:46 UTC - in response to Message 2178.

                                                                                  Last modified: 17 Feb 2007 19:39:47 UTC

                                                                                  I dont do alot of posting but decided I wanted to make a comment or two...my personal opinion is that this whole conversation is getting a bit on the silly side....if this project isn't using those benchmarks to calculate the credits why should whether otr not they are inflated matter?


                                                                                  Well, it's beginning to look like Crunch3r's BOINC for Windows does not inflate the benchmarks, though I don't know for sure because I use Linux, not Windows. But let's pretend for a moment, just for the sake of discussion, that it does inflate benchmarks. In that case, it would matter to this project because some people crunch several projects and some of those projects still use the benchmarks. So if people use a benchmark inflating BOINC to crunch ABC then they'll be cheating at some other projects. They don't want to cheat so they won't run the cheater BOINC which means they can't crunch 64-bit here at ABC if they run Windows. That hurts ABC. See? No man lives in total isolation. No project lives in total isolation. Everything we do ripples out and touches everyone and everything. Open your eyes and you will see this is so.

                                                                                  ....if Hendrick approved the use of Crunch3r's client then that should be the end of itshouldn't it?


                                                                                  Has Hendrik actually approved Crunch3r's client? I don't think so. Unless I've missed one of his posts it looks like he has neither approved nor disapproved. And when you think about it, isn't it best for him to just stay neutral? Now please, some people would say that constitutes a non-verbal agreement .... he didn't say we shouldn't use it so that means we should... but that's not really responsible and honest, is it?

                                                                                  ...I will be using it for testing in beta and then when it is all ready in the stable....I think it is a good thing that Crunch3r gave us the opportunity to be able to help advance the work being done and should be thanked for it...


                                                                                  Well, like I said above, the issue is probably a non-issue because it appears it doesn't inflate benchmarks. If that's true then I add my thanks to your thanks and applaud Crunch3r for helping the community.

                                                                                  if anyone disagrees with me I don't really care but none of the argument said in this thread are going to change my opinion ...


                                                                                  That's a very closed minded attitude and it probably won't help you in life in general but it's up to you. Just be careful how many people you hurt with that attitude.

                                                                                  and on a side note I think some people here should act more there age...


                                                                                  Indeed they should!!

                                                                                  EDIT: Added I Am A Rock

                                                                                  Profile Scooter
                                                                                  Send message
                                                                                  Joined: Nov 21 06
                                                                                  Posts: 15
                                                                                  Credit: 37,521
                                                                                  RAC: 0
                                                                                  Message 2201 - Posted 17 Feb 2007 21:09:00 UTC

                                                                                    I think you didnt read it correctly It clearly stated IF Hendrick approved it. IF being the key word. So you might need to be the one to open your eyes and read before responding.

                                                                                    Dagorath
                                                                                    Send message
                                                                                    Joined: Jan 7 07
                                                                                    Posts: 384
                                                                                    Credit: 3,374,905
                                                                                    RAC: 0
                                                                                    Message 2207 - Posted 17 Feb 2007 22:20:39 UTC - in response to Message 2201.

                                                                                      I think you didnt read it correctly It clearly stated IF Hendrick approved it. IF being the key word. So you might need to be the one to open your eyes and read before responding.


                                                                                      No, I read it correctly. That's why I phrased it as a question. I didn't say you said he approved it. Go ahead and argue that I implied it because then I'll just argue that you implied he said it and that your IF was extraneous.

                                                                                      Furthermore, my suggestion to open your eyes had nothing to do with what Hendrik did or did not say. It was about the harm that may befall this project, an idea Webmaster Yoda mentioned already in this thread. At first I was a little baffled how you missed that so I tried gently to lead you to it. I should have known better than to try to help, it's clear you really don't want or need any, your mind already made up and surely made up right.

                                                                                      STE\/E [BlackOps]
                                                                                      Send message
                                                                                      Joined: Nov 21 06
                                                                                      Posts: 25
                                                                                      Credit: 7,588,293
                                                                                      RAC: 0
                                                                                      Message 2211 - Posted 17 Feb 2007 22:39:53 UTC

                                                                                        This Thread reminds me of the Good Old Days over @ in the Seti Forum's ... NOT

                                                                                        I stopped running Seti a couple of years ago just because of this sort of 3 Ring Circus Atmosphere, I guess I can add ABC to the Circus List now that things seem to be degrading to the same level as Seti ... Cheer's

                                                                                        Dagorath
                                                                                        Send message
                                                                                        Joined: Jan 7 07
                                                                                        Posts: 384
                                                                                        Credit: 3,374,905
                                                                                        RAC: 0
                                                                                        Message 2214 - Posted 17 Feb 2007 23:44:49 UTC - in response to Message 2211.

                                                                                          This Thread reminds me of the Good Old Days over @ in the Seti Forum's ... NOT

                                                                                          I stopped running Seti a couple of years ago just because of this sort of 3 Ring Circus Atmosphere, I guess I can add ABC to the Circus List now that things seem to be degrading to the same level as Seti ... Cheer's


                                                                                          That's what happens when people hold grudges over past events and won't let go. Maybe Crunch3r was wrong in the past, maybe he was right, I don't care anymore for reasons I won't explain here lest some nitwit with a grudge use it as proof their crusade was The One True and Holy Crusade.

                                                                                          The problem here, today, in this thread, is that people have made false accusations. One has apologised, one has not. And a third has attempted to jump aboard a bandwagon that left town yesterday spoiling for an argument, thinking he too can march out of oblivion and into the dream with mugwump word tricks.

                                                                                          Profile Scooter
                                                                                          Send message
                                                                                          Joined: Nov 21 06
                                                                                          Posts: 15
                                                                                          Credit: 37,521
                                                                                          RAC: 0
                                                                                          Message 2215 - Posted 18 Feb 2007 0:33:23 UTC - in response to Message 2214.



                                                                                            The problem here, today, in this thread, is that people have made false accusations. One has apologised, one has not. And a third has attempted to jump aboard a bandwagon that left town yesterday spoiling for an argument, thinking he too can march out of oblivion and into the dream with mugwump word tricks.



                                                                                            you seem like you are the one who wants an argument and also seem to be very negative about anything anyone has an opinion about ....on that note im done dealing with your responses as it is giving me a headache!!!

                                                                                            Dagorath
                                                                                            Send message
                                                                                            Joined: Jan 7 07
                                                                                            Posts: 384
                                                                                            Credit: 3,374,905
                                                                                            RAC: 0
                                                                                            Message 2217 - Posted 18 Feb 2007 1:03:49 UTC - in response to Message 2215.



                                                                                              The problem here, today, in this thread, is that people have made false accusations. One has apologised, one has not. And a third has attempted to jump aboard a bandwagon that left town yesterday spoiling for an argument, thinking he too can march out of oblivion and into the dream with mugwump word tricks.



                                                                                              you seem like you are the one who wants an argument and also seem to be very negative about anything anyone has an opinion about ....on that note im done dealing with your responses as it is giving me a headache!!!



                                                                                              I can be very negative about opinions that hold no water and I will not be beaten back by mugwump word tricks taught by second rate high school debate team coaches. If you can't stand that kind of heat then hang out in the "Last One To Post Here Wins" thread.

                                                                                              If you feel I am negative about everything then you obviously haven't read everything I have to say. Be tenacious, loquacious and segacious, all three. The first 2 alone just won't do, you'll get a headache every time..

                                                                                              Profile Webmaster Yoda
                                                                                              Avatar
                                                                                              Send message
                                                                                              Joined: Dec 31 06
                                                                                              Posts: 81
                                                                                              Credit: 4,544,249
                                                                                              RAC: 0
                                                                                              Message 2218 - Posted 18 Feb 2007 1:27:03 UTC

                                                                                                Last modified: 18 Feb 2007 1:28:01 UTC

                                                                                                If you want to run ABC@home's 64 bit client, you will have to use one of the third party BOINC clients, as Berkeley do not (yet) have a 64 bit client. Period.

                                                                                                Yes, the 64bit client I use now (installed for use at AB@home) does appear to inflate benchmarks but the work is done much faster too and I still get more than claimed. When all is said and done, benchmarks are irrelevant on this project.

                                                                                                Let's stop this endless, nonsensical, off-topic bickering.

                                                                                                Dagorath
                                                                                                Send message
                                                                                                Joined: Jan 7 07
                                                                                                Posts: 384
                                                                                                Credit: 3,374,905
                                                                                                RAC: 0
                                                                                                Message 2219 - Posted 18 Feb 2007 2:01:05 UTC - in response to Message 2218.

                                                                                                  Last modified: 18 Feb 2007 2:02:38 UTC

                                                                                                  If you want to run ABC@home's 64 bit client, you will have to use one of the third party BOINC clients, as Berkeley do not (yet) have a 64 bit client. Period.

                                                                                                  Yes, the 64bit client I use now (installed for use at AB@home) does appear to inflate benchmarks but the work is done much faster too and I still get more than claimed. When all is said and done, benchmarks are irrelevant on this project.

                                                                                                  Let's stop this endless, nonsensical, off-topic bickering.



                                                                                                  I'm all for stopping the bickering but I fear that won't happen until we get the facts straight. So, Webmaster Yoda, about those apparently inflated benchmarks you're seeing...

                                                                                                  I'm not sure exactly what you're seeing but I do know that on my 64-bit Linux system the IOPS benchmark is much higher with Augustine's 64-bit BOINC client compared to when I had 32-bit Linux and standard BOINC on that machine. Now it's 6252 MIOPS, before was 4700 MIOPS. The MFLOPS is about the same. It's an Athlon 64, not C2D, so this isn't due to that issue some people say happens with C2D.

                                                                                                  What are you seeing, Webmaster Yoda?

                                                                                                  I'm thinking maybe my MIOPS is way up because the integer portion of the benchmarks code executes faster. But last time I looked at that code it used 32 bit ints not 64 bit so I'm wondering.... how can my IOPS be so much higher? Did Augustine modify the benchmark code to use 64-bit ints? If so then that seems fair to me because the science apps will be using 64-bit ints too. And if Augustine modified it then maybe Crunch3r modified it too?

                                                                                                  Profile Webmaster Yoda
                                                                                                  Avatar
                                                                                                  Send message
                                                                                                  Joined: Dec 31 06
                                                                                                  Posts: 81
                                                                                                  Credit: 4,544,249
                                                                                                  RAC: 0
                                                                                                  Message 2220 - Posted 18 Feb 2007 2:11:37 UTC - in response to Message 2219.

                                                                                                    Last modified: 18 Feb 2007 2:22:07 UTC

                                                                                                    What I am seeing is benchmarks that are inflated by a factor of about two. I don't care what causes it, as there is no official 64 bit BOINC client. I don't care whether it's Augustine's or Crunch3r's or anyone else's 64 bit client. There is no official 64 bit BOINC client

                                                                                                    And as I said, it does not matter here. Nothing to argue about. Let it rest! it only matters if you use it on projects that rely on the benchmarks for granting credit (which I don't with this computer)

                                                                                                    My computers are not hidden so if you want to see what I'm seeing, look them up. Host 6446 is overclocked by almost 50% and uses one of the 64 bit clients. It used to be host 5822 with 32bit Linux and official 32bit BOINC client 5.4.11
                                                                                                    ____________


                                                                                                    Join the #1 Aussie Alliance on ABC@Home

                                                                                                    Astro
                                                                                                    Avatar
                                                                                                    Send message
                                                                                                    Joined: Nov 21 06
                                                                                                    Posts: 102
                                                                                                    Credit: 601,226
                                                                                                    RAC: 0
                                                                                                    Message 2221 - Posted 18 Feb 2007 2:15:59 UTC

                                                                                                      Last modified: 18 Feb 2007 2:28:52 UTC

                                                                                                      My Dhrystone went from a max of 4800 with standard boinc for windows, Crunch3r's 5.8.11_amd64 maxed out at 6300 and change. It looks to be very similar in proportions. 64b has twice the SSE registers and I'm thinking that it simply works faster (IE completes more loops in the run time). There IS a way to tell. (NOTE: this only works with 5.8.10 and above, so you'd have to have the latest boinc official, the latest Crunch3r, and Augustine)

                                                                                                      By creating a cc_config.xml file and placing it in the boinc folder, then starting boinc and running the benchmarks (advanced-run benchmarks).

                                                                                                      the cc_config.xml should contain the following:

                                                                                                      <cc_config>
                                                                                                      <log_flags>
                                                                                                      <benchmark_debug />
                                                                                                      </log_flags>
                                                                                                      </cc_config>

                                                                                                      It will output something similar to this:

                                                                                                      2/17/2007 5:19:02 PM||Running CPU benchmarks
                                                                                                      2/17/2007 5:19:02 PM||Suspending computation - running CPU benchmarks
                                                                                                      2/17/2007 5:19:04 PM||[benchmark_debug] Starting floating-point benchmark
                                                                                                      2/17/2007 5:19:24 PM||[benchmark_debug] Ended floating-point benchmark
                                                                                                      2/17/2007 5:19:39 PM||[benchmark_debug] Starting integer benchmark
                                                                                                      2/17/2007 5:19:59 PM||[benchmark_debug] Ended integer benchmark
                                                                                                      2/17/2007 5:20:02 PM||[benchmark_debug] Ended benchmark
                                                                                                      2/17/2007 5:20:03 PM||[benchmark_debug] CPU 0 has finished
                                                                                                      2/17/2007 5:20:03 PM||[benchmark_debug] CPU 1 has finished
                                                                                                      2/17/2007 5:20:03 PM||[benchmark_debug] 2 out of 2 CPUs done
                                                                                                      2/17/2007 5:20:03 PM||[benchmark_debug] CPU 0: fp 2604068857.589984 int 3145259831.507770 intloops 109920000.000000 inttime 19.890625
                                                                                                      2/17/2007 5:20:03 PM||[benchmark_debug] CPU 1: fp 2637636080.870917 int 3186892757.171830 intloops 111200000.000000 inttime 19.859375
                                                                                                      2/17/2007 5:20:03 PM||Benchmark results:
                                                                                                      2/17/2007 5:20:03 PM|| Number of CPUs: 2
                                                                                                      2/17/2007 5:20:03 PM|| 2621 floating point MIPS (Whetstone) per CPU
                                                                                                      2/17/2007 5:20:03 PM|| 3166 integer MIPS (Dhrystone) per CPU
                                                                                                      2/17/2007 5:20:04 PM||Resuming computation

                                                                                                      It shows the runtime and loops. The avg of fp and int for cpu0+cpu1 is used in the displayed output everyone normally sees.

                                                                                                      The problem for me is that the dual core benchmark bug makes it to unreliable for me to use, however someone with a single core could test it. I could test 32b linux vs 32b win vs 64b linux.

                                                                                                      The question would probably still remain as to what caused it to do more intloops/second. But even this is speculation without testing

                                                                                                      Dagorath
                                                                                                      Send message
                                                                                                      Joined: Jan 7 07
                                                                                                      Posts: 384
                                                                                                      Credit: 3,374,905
                                                                                                      RAC: 0
                                                                                                      Message 2222 - Posted 18 Feb 2007 2:28:49 UTC - in response to Message 2220.

                                                                                                        What I am seeing is benchmarks that are inflated by a factor of two. I don't care what causes it, as there is no official 64 bit BOINC client. I don't care whether it's Augustine's or Crunch3r's or anyone else's 64 bit client. There is no official 64 bit BOINC client

                                                                                                        And as I said, it does not matter here. Nothing to argue about. Let it rest!



                                                                                                        There are many reasons arguments erupt. One reason is that sometimes people just don't know the facts. I am not trying to prolong any argument, I simply want to get a few facts straight. Why are my MIOPS higher on Augustine's 64-bit BOINC client for Linux? And why are Webmaster Yoda's benchmarks apparently higher on Crunch3r's 64-bit BOINC for Windows?

                                                                                                        As Webmaster Yoda said... inflated benchmarks don't make any difference on ABC so I'm not going to call anyone a cheater or anything like that. I personally don't care anymore because I absolutely refuse to crunch projects that use benchmarks for credit calculations. I just want to know why the benchmarks are higher, that's all.

                                                                                                        Astro
                                                                                                        Avatar
                                                                                                        Send message
                                                                                                        Joined: Nov 21 06
                                                                                                        Posts: 102
                                                                                                        Credit: 601,226
                                                                                                        RAC: 0
                                                                                                        Message 2223 - Posted 18 Feb 2007 2:34:25 UTC

                                                                                                          Last modified: 18 Feb 2007 2:40:35 UTC

                                                                                                          Quick note: if you run winme or win98 don't bother with 5.8.12 or 5.8.13, it won't install and gives an error. 5.8.14 does work however.

                                                                                                          Profile Scooter
                                                                                                          Send message
                                                                                                          Joined: Nov 21 06
                                                                                                          Posts: 15
                                                                                                          Credit: 37,521
                                                                                                          RAC: 0
                                                                                                          Message 2224 - Posted 18 Feb 2007 2:36:13 UTC

                                                                                                            on my AMD Athlon 64 3400+ OC'd to 2.36

                                                                                                            32 bit 5.4.11

                                                                                                            2/17/2005 9:27:35 AM||Running CPU benchmarks
                                                                                                            2/17/2005 9:28:34 AM||Benchmark results:
                                                                                                            2/17/2005 9:28:34 AM|| Number of CPUs: 1
                                                                                                            2/17/2005 9:28:34 AM|| 2148 floating point MIPS (Whetstone) per CPU
                                                                                                            2/17/2005 9:28:34 AM|| 3939 integer MIPS (Dhrystone) per CPU
                                                                                                            2/17/2005 9:28:34 AM||Finished CPU benchmarks

                                                                                                            Crunch3r's 64 bit 5.8.11


                                                                                                            2/17/2005 9:31:00 AM||Running CPU benchmarks
                                                                                                            2/17/2005 9:32:01 AM||Benchmark results:
                                                                                                            2/17/2005 9:32:01 AM|| Number of CPUs: 1
                                                                                                            2/17/2005 9:32:01 AM|| 2050 floating point MIPS (Whetstone) per CPU
                                                                                                            2/17/2005 9:32:01 AM|| 5435 integer MIPS (Dhrystone) per CPU
                                                                                                            2/17/2005 9:32:02 AM||Resuming computation

                                                                                                            Dagorath
                                                                                                            Send message
                                                                                                            Joined: Jan 7 07
                                                                                                            Posts: 384
                                                                                                            Credit: 3,374,905
                                                                                                            RAC: 0
                                                                                                            Message 2225 - Posted 18 Feb 2007 2:36:20 UTC - in response to Message 2221.

                                                                                                              By creating a cc_config.xml file and placing it in the boinc folder, then starting boinc and running the benchmarks (advanced-run benchmarks).


                                                                                                              Yah, I can do that. Have to go out for a while right now but I'll get on it as soon as I return.


                                                                                                              Dagorath
                                                                                                              Send message
                                                                                                              Joined: Jan 7 07
                                                                                                              Posts: 384
                                                                                                              Credit: 3,374,905
                                                                                                              RAC: 0
                                                                                                              Message 2226 - Posted 18 Feb 2007 5:37:55 UTC

                                                                                                                Last modified: 18 Feb 2007 5:41:10 UTC

                                                                                                                OK, Astro, here's what I have so far. First is Augustine's BOINC, second is standard 5.8.11. The 2 log excerpts are from twin machines (same mobo, same CPU, same video, same RAM, same everything, no OC). Both have the cc_config.xml in BOINC dir as per your request.

                                                                                                                Note Augustine's client doesn't report the loop count like Crunch3r's does. I don't know if these tests tell you what you want to know. If not then let me know if you have other ideas, because I can try other tests too. I don't really want to but I could put WinXP 64bit on a spare drive and test Crunch3r's client too.

                                                                                                                Also, the difference in int ops is even greater than I said a few posts up. I was working from memory then and recalled the standard int ops incorrectly.

                                                                                                                ************************************************
                                                                                                                2007-02-16 19:16:56 [---] Starting BOINC client version 5.8.11 for x86_64-pc-linux-gnu
                                                                                                                2007-02-16 19:16:57 [---] Processor: 1 AuthenticAMD AMD Athlon(tm) 64 Processor 3300+ [fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 syscall nx mmxext lm 3dnowext 3dnow]
                                                                                                                2007-02-16 19:16:57 [---] Memory: 875.53 MB physical, 1.72 GB virtual

                                                                                                                2007-02-17 21:11:07 [---] [benchmark_debug] Starting floating-point benchmark
                                                                                                                2007-02-17 21:11:27 [---] [benchmark_debug] Ended floating-point benchmark
                                                                                                                2007-02-17 21:11:43 [---] [benchmark_debug] Starting integer benchmark
                                                                                                                2007-02-17 21:11:05 [---] Running CPU benchmarks
                                                                                                                2007-02-17 21:12:02 [---] [benchmark_debug] Ended integer benchmark
                                                                                                                2007-02-17 21:12:05 [---] [benchmark_debug] Ended benchmark
                                                                                                                2007-02-17 21:12:06 [---] [benchmark_debug] CPU 0 has finished
                                                                                                                2007-02-17 21:12:06 [---] [benchmark_debug] 1 out of 1 CPUs done
                                                                                                                2007-02-17 21:12:06 [---] [benchmark_debug] CPU 0: fp 2253916518.037568 int 6276753887.305083
                                                                                                                2007-02-17 21:12:06 [---] Benchmark results:
                                                                                                                2007-02-17 21:12:06 [---] Number of CPUs: 1
                                                                                                                2007-02-17 21:12:06 [---] 2254 floating point MIPS (Whetstone) per CPU
                                                                                                                2007-02-17 21:12:06 [---] 6277 integer MIPS (Dhrystone) per CPU
                                                                                                                2007-02-17 21:12:07 [---] Resuming computation

                                                                                                                ************************************************

                                                                                                                2007-02-17 18:23:22 [---] Starting BOINC client version 5.8.11 for windows_intelx86
                                                                                                                2007-02-17 18:23:22 [---] Processor: 1 AuthenticAMD AMD Athlon(tm) 64 Processor 3300+ [x86 Family 15 Model 12 Stepping 0] [fpu tsc pae nx sse sse2 3dnow mmx]
                                                                                                                2007-02-17 18:23:22 [---] Memory: 895.48 MB physical, 2.76 GB virtual

                                                                                                                2007-02-17 22:20:02 [---] Running CPU benchmarks
                                                                                                                2007-02-17 22:20:03 [---] Suspending computation - running CPU benchmarks
                                                                                                                2007-02-17 22:20:05 [---] [benchmark_debug] Starting floating-point benchmark
                                                                                                                2007-02-17 22:20:25 [---] [benchmark_debug] Ended floating-point benchmark
                                                                                                                2007-02-17 22:20:40 [---] [benchmark_debug] Starting integer benchmark
                                                                                                                2007-02-17 22:21:00 [---] [benchmark_debug] Ended integer benchmark
                                                                                                                2007-02-17 22:21:03 [---] [benchmark_debug] Ended benchmark
                                                                                                                2007-02-17 22:21:04 [---] [benchmark_debug] CPU 0 has finished
                                                                                                                2007-02-17 22:21:04 [---] [benchmark_debug] 1 out of 1 CPUs done
                                                                                                                2007-02-17 22:21:04 [---] [benchmark_debug] CPU 0: fp 2278481012.658228 int 4194962692.401509
                                                                                                                2007-02-17 22:21:04 [---] Benchmark results:
                                                                                                                2007-02-17 22:21:04 [---] Number of CPUs: 1
                                                                                                                2007-02-17 22:21:04 [---] 2278 floating point MIPS (Whetstone) per CPU
                                                                                                                2007-02-17 22:21:04 [---] 4195 integer MIPS (Dhrystone) per CPU
                                                                                                                2007-02-17 22:21:05 [---] Resuming computation
                                                                                                                ************************************************

                                                                                                                Profile clownius
                                                                                                                Forum moderator
                                                                                                                Avatar
                                                                                                                Send message
                                                                                                                Joined: Jan 14 07
                                                                                                                Posts: 242
                                                                                                                Credit: 6,347,407
                                                                                                                RAC: 0
                                                                                                                Message 2229 - Posted 18 Feb 2007 10:05:40 UTC

                                                                                                                  One of the projects i do run (RCN) does rely on benchmarks. Its the only project i still crunch that does for similar reasons to many. I know it doesn't come with a 64bit app but i can always do the app info thing to keep crunching.
                                                                                                                  I really want to find a BOINC Client that doesn't inflate benchmarks in any way. Can anyone tell me if there is one? Im Linux only so Crunchers is out straight away.
                                                                                                                  Do i need to compile my own is that the only way?

                                                                                                                  Thanks to anyone that can help me as i would hate to have to stay on 32bit and not crunch faster ABC just because i cant find a good BOINC client
                                                                                                                  ____________

                                                                                                                  fubared
                                                                                                                  Send message
                                                                                                                  Joined: Jan 11 07
                                                                                                                  Posts: 17
                                                                                                                  Credit: 561,575
                                                                                                                  RAC: 0
                                                                                                                  Message 2231 - Posted 18 Feb 2007 10:14:54 UTC - in response to Message 2229.

                                                                                                                    Last modified: 18 Feb 2007 10:15:27 UTC


                                                                                                                    Do i need to compile my own is that the only way?


                                                                                                                    Probably the only way. I too have noticed that integer benchmark is alot higher (Augustine's client) but I'm not worried. My favourite projects are fixed credit anyway.
                                                                                                                    ____________
                                                                                                                    Team Boinc@Australia.

                                                                                                                    Profile clownius
                                                                                                                    Forum moderator
                                                                                                                    Avatar
                                                                                                                    Send message
                                                                                                                    Joined: Jan 14 07
                                                                                                                    Posts: 242
                                                                                                                    Credit: 6,347,407
                                                                                                                    RAC: 0
                                                                                                                    Message 2233 - Posted 18 Feb 2007 10:34:18 UTC

                                                                                                                      Well i decided im taking the plunge. It may take a while though as im finishing off all of the work on the machine first and that includes a pair of CPDN WU's as im afraid of loosing things i already have hundred of hours put into. Expect my output to drop off as only the pair of Celeries will remain crunching until i clear the CPDN's and then ill install 64bit and try my own compile of the BOINC CC. If i manage it ill make it available to others
                                                                                                                      ____________

                                                                                                                      Dagorath
                                                                                                                      Send message
                                                                                                                      Joined: Jan 7 07
                                                                                                                      Posts: 384
                                                                                                                      Credit: 3,374,905
                                                                                                                      RAC: 0
                                                                                                                      Message 2235 - Posted 18 Feb 2007 11:12:57 UTC - in response to Message 2231.


                                                                                                                        Do i need to compile my own is that the only way?


                                                                                                                        Probably the only way. I too have noticed that integer benchmark is alot higher (Augustine's client) but I'm not worried. My favourite projects are fixed credit anyway.


                                                                                                                        fubared and Clownius, you've missed the point. Perhaps the benchmarks are higher because the benchmark calculations run faster in 64-bit, not because Augustine and Crunch3r are trying to cheat. Perhaps Augustine's and Crunch3r's benchmarks are exactly what a 64-bit BOINC from Berkely would produce. Think about it.

                                                                                                                        Also, if you're going to compile your own 64-bit BOINC then where are you going to getthe source code to compile? From Berkeley? Nope, they don't have such a critter yet. Hmmm? Are we thinking yet? Or just jerking knees?

                                                                                                                        Profile clownius
                                                                                                                        Forum moderator
                                                                                                                        Avatar
                                                                                                                        Send message
                                                                                                                        Joined: Jan 14 07
                                                                                                                        Posts: 242
                                                                                                                        Credit: 6,347,407
                                                                                                                        RAC: 0
                                                                                                                        Message 2237 - Posted 18 Feb 2007 11:35:55 UTC

                                                                                                                          We are trying to come up with a solution :) I have some Linux experience but none with 64bit so its going to be a learning experience. That may well be the reason for the higher benchmarks but i don't know and to avoid hassles like being called a cheat ( Im far from a fan of altered benchmarks ) im quite willing to play around with compiling from source (i actually though that i could compile for 64bit considering source is just that source and generally compilable on multiple platforms but i may be wrong (i hope not)).
                                                                                                                          If i find its running higher benchmarks then i will wholehearted support these other clients. If i i find it isn't i will make my binary available to others.
                                                                                                                          That said ive been know to completely nerf installs and compiles but i get them right at times too so we will see.
                                                                                                                          Crunchers client is unfortunately useless to me as im a Linux user so my only comparison will be with Augustine's client.

                                                                                                                          If anyone knows im wrong about something speak up as compiling from source is no picnic nor is installing the Intel compiler (maybe im better off using an open source compiler). Im always open to accepting help or ideas.

                                                                                                                          Clownius
                                                                                                                          ____________

                                                                                                                          fubared
                                                                                                                          Send message
                                                                                                                          Joined: Jan 11 07
                                                                                                                          Posts: 17
                                                                                                                          Credit: 561,575
                                                                                                                          RAC: 0
                                                                                                                          Message 2239 - Posted 18 Feb 2007 12:10:30 UTC - in response to Message 2235.


                                                                                                                            fubared and Clownius, you've missed the point. Perhaps the benchmarks are higher because the benchmark calculations run faster in 64-bit, not because Augustine and Crunch3r are trying to cheat. Perhaps Augustine's and Crunch3r's benchmarks are exactly what a 64-bit BOINC from Berkely would produce. Think about it.


                                                                                                                            No you missed mine. I never said anything about cheating/inflating of credits. It was just an observation I made. I already said I wasn't worried about the difference in benchmark. I used the optimised clients in the past and now I'm suddenly worried? Unlikely.
                                                                                                                            ____________
                                                                                                                            Team Boinc@Australia.

                                                                                                                            Astro
                                                                                                                            Avatar
                                                                                                                            Send message
                                                                                                                            Joined: Nov 21 06
                                                                                                                            Posts: 102
                                                                                                                            Credit: 601,226
                                                                                                                            RAC: 0
                                                                                                                            Message 2240 - Posted 18 Feb 2007 12:13:04 UTC

                                                                                                                              Last modified: 18 Feb 2007 12:21:40 UTC

                                                                                                                              Hmmm, I'm looking at the reading Dagorath presented. I don't see the "intloops and inttime". It must be that they only work with 5.8.13 and above. The <benchmark_debug/> was created in 5.8.10 so they might be able to track down the "dual core benchmark bug". They don't have any dual cores with winxp so they can't find it themselves. I see in the changelog for 5.8.13 that he made additions to the benchmark debug capabilities.

                                                                                                                              Change Log:

                                                                                                                              - core client: don't print misleading error messages when
                                                                                                                              reference
                                                                                                                              site access succeeds on initial startup.
                                                                                                                              - core client: a few more messages for <benchmark_debug>
                                                                                                                              - core client: fix bug where if a task is aborted
                                                                                                                              (e.g. because it exceeds CPU limit)
                                                                                                                              it's restarted on the next enforce_schedule().
                                                                                                                              The problem: we're deleting the ACTIVE_TASK,
                                                                                                                              but the result is still in the ordered_scheduled_results list.
                                                                                                                              The solution: call request_schedule_cpus() in
                                                                                                                              handle_finished_apps() when an ACTIVE_TASK is deleted.
                                                                                                                              - WININSTALL: Use the Unicode version of ShutdownBOINCManager on
                                                                                                                              NT based
                                                                                                                              platforms and use the ANSI version on 9X based platforms.


                                                                                                                              So the extra info comes from the version, not the benchmark debug.

                                                                                                                              According to the message log both ran fp for 20 seconds, but augustines int ran 22 seconds, as opposed to 20 seconds for the original. Since the loops are counted in CPU time, NOT wall clock this is meaningless without the extra info in 5.8.13, specifically "inttime" and "intloops". It's probable, but unproveable, that Augustines did run for 20 seconds cpu time, and some other process got the other two seconds.

                                                                                                                              FP is supposed to run 20 seconds, then pause 15 seconds, then INT run for 20 seconds.

                                                                                                                              I think anyone else running this is useless ATM (until comparable versions over 5.8.13 are available). Not to mention that we'd still most likely be left wondering why the higher "intloops/time"

                                                                                                                              Dagorath
                                                                                                                              Send message
                                                                                                                              Joined: Jan 7 07
                                                                                                                              Posts: 384
                                                                                                                              Credit: 3,374,905
                                                                                                                              RAC: 0
                                                                                                                              Message 2254 - Posted 18 Feb 2007 19:53:15 UTC - in response to Message 2240.

                                                                                                                                Last modified: 18 Feb 2007 19:56:16 UTC

                                                                                                                                I think anyone else running this is useless ATM (until comparable versions over 5.8.13 are available). Not to mention that we'd still most likely be left wondering why the higher "intloops/time"


                                                                                                                                I am confident that in the end you will find the higher intloops/time is due to the fact the 64-bit code executes faster. We'll never know for sure until we see the source code or have a standard from Berkeley to compare credit claims against. It's not a mission/crusade/project I have the time to get involved in, let alone the motivation.

                                                                                                                                In the meantime there is an easy workaround for anybody who wants to run Crunch3r's or Augustine's 64-bit BOINC and wants to make absolutely sure they won't overclaim. The solution is to set the bechmarks as low as you think they should be in client_state.xml and then adjust the p_calc number to a high value to prevent BOINC from setting the benchmarks to what it thinks they should be.

                                                                                                                                Anyway, I'm running Augustine's BOINC and I don't care if it overclaims. I think it is not overclaiming and that's all I have to say on the matter, as if anybody cares what I say, lol.

                                                                                                                                Augustine
                                                                                                                                Avatar
                                                                                                                                Send message
                                                                                                                                Joined: Nov 22 06
                                                                                                                                Posts: 36
                                                                                                                                Credit: 231,225
                                                                                                                                RAC: 0
                                                                                                                                Message 2411 - Posted 1 Mar 2007 15:20:35 UTC - in response to Message 2078.

                                                                                                                                  I run other projects, and so won't run any 64b Boinc which has "altered" benchmarks, but that seems to be the only source available so far. That means I won't be doing any 64b with windows.

                                                                                                                                  Well, as someone who libels others, the burden of proof is on you: provide evidence that the benchmarks were "altered" in either the Linux or the Windows AMD64 clients.

                                                                                                                                  ____________

                                                                                                                                  Post to thread

                                                                                                                                  Message boards : Number crunching : 64bit linux in stable


                                                                                                                                  Return to ABC@home main page


                                                                                                                                  Copyright © 2014 University of Leiden