Posts by Brian Priebe

21) Message boards : Number crunching : Milkyway N-Body uses all BOINC time (Message 51284) Posted 3 Oct 2011 by Brian Priebe Post: Yes, in fact it takes 10 minutes to complete the WU... but with an estimated time over 100 hours Boinc runs Milkyway N-Body in high priority as soon as a WU is downloaded. Have you tried resetting the MW project to reset the BOINC Duration Correction Factor?
22) Message boards : Number crunching : HD5870 dead - help with replacement! (Message 50840) Posted 26 Aug 2011 by Brian Priebe Post: Test: Leave 20 cores unused and see how the Cayman performs now. It should now be fully utilized. Even with BOINC allowed to use only 10% of the cores, the GPU was still running around 75%. Utterly bizarre. One thing that is noticeably different with this machine is the NUMA memory access mode is enabled by default in the BIOS. I tried disabling it. Now everything is running fine! With BOINC allowed to use 95% of the processor cores, I get 98% GPU usage with 22 CPU WU's running. At 96% of the processor cores, GPU drops to 94% usage with 23 CPU WU's running. At 100% of the processor cores, GPU drops to 65% (!).
23) Message boards : Number crunching : HD5870 dead - help with replacement! (Message 50812) Posted 23 Aug 2011 by Brian Priebe Post: Why would you want to? Several people have claimed on here that leaving a whole core available for MW ATI app results in improved throughput. I'm not seeing it...
24) Message boards : Number crunching : HD5870 dead - help with replacement! (Message 50811) Posted 23 Aug 2011 by Brian Priebe Post: DL Process Lasso, install it and configure the default priority of MW to "Above normal". MW should now run at 98-99%... Have previously tried setting MW ATI app to "above normal" or "high" with Task Manager and no visible improvement at all.
25) Message boards : Number crunching : HD5870 dead - help with replacement! (Message 50804) Posted 23 Aug 2011 by Brian Priebe Post: A 6950 will consistently outperform - when measured across a broad range of tasks - a 5870 by around 10%. The margin for a 6970 is around 20%. In theory this should be true. However, I have two workstations executing identical series WU's and the HD5870 (Win/XP) is consistently running away from the HD6970 (WIN7 64-bit). MW seems to have a hard time keeping the HD6970 more than 75% busy even if I prevent BOINC from using more than 90% of the processor cores available. The HD5870 is running about 96% busy with no restrictions on processor core use. It is a puzzlement...
26) Message boards : Number crunching : Tasks not quitting when computer becomes active (Message 50631) Posted 10 Aug 2011 by Brian Priebe Post: Sometimes it takes up to 5sec for the task to pause. It won't let me edit my own message so I'll reply to it... If one looks at the BOINC manager screen while typing in keystrokes or moving the mouse, you can see the GPU task is indeed pausing within about 1sec. If done outside of BOINC and the GPU lockout setting is too short, like 1-2sec, the GPU task starts right up again before the application has had a chance to respond to your input. So this responsiveness issue is just a variant of the other I reported. Asking for a 5sec lockout on the GPU appears to greatly increase apparent responsiveness. Unfortunately, this greatly stretches the elapsed time to return GPU WU results.
27) Message boards : Number crunching : Tasks not quitting when computer becomes active (Message 50630) Posted 10 Aug 2011 by Brian Priebe Post: Since these GPU tasks are making my WIN/XP machine unusable for work, I have also tried that automatic override to temporarily halt the tasks. (This machine has an ATI HD5870 so the problem is not limited to NVidia.) Sometimes it works and sometimes it doesn't. Sometimes it takes up to 5sec for the task to pause. As I type in this message, the screen is visibly unable to keep up with the keystrokes, halting for seconds at a time. Elsewhere it has been posted that these GPU tasks are supposed to limit themselves to about 1/30sec of run time. Based on the symptoms in front of my eyes, I don't believe that is the case.
28) Message boards : Number crunching : GPU Requirements (Message 49862) Posted 4 Jul 2011 by Brian Priebe Post: I'm not seeing that on my machines (3850/4870/4870/5970). Then I wonder what the difference is. The HD5870 is running under XP SP3. Since the new MW ATI apps, the system has been nearly unusable due to this UI lag.
29) Message boards : Number crunching : GPU Requirements (Message 49827) Posted 3 Jul 2011 by Brian Priebe Post: This makes sure if you are doing some other work on the PC (Mail, Office, WEB-Browsing, etc) the GPU chruches. And you need to be aware that having MW crunch using the GPU all the time can make using IE8, MS/Word, Excel, or even scrolling through a long list of files in a folder a painfully slow experience. With an HD5870 crunching, I'm frequently reaching for the Snooze GPU option just to do basic work.
30) Message boards : News : ATI application updated to 0.60 (Message 47781) Posted 13 Apr 2011 by Brian Priebe Post: Brian, on two of your machines all your tasks are showing as 0.50 SSE2 when we are now on 0.62? The third machine (ADMIN2) is the XP machine with the HD5870. The other two have only HD5770's that can't be used for MW.
31) Message boards : News : ATI application updated to 0.60 (Message 47774) Posted 13 Apr 2011 by Brian Priebe Post: No screen lag, responsive on mouse and keyboard. Can open browsers etc on top of it fine. Unfortunately, my experience under WinXP with Catalyst 11.3 is just the opposite. 0.62 slows the screen down so much it renders the machine almost unusable. What changed since 0.59?
32) Message boards : News : ATI application updated again (Message 47714) Posted 12 Apr 2011 by Brian Priebe Post: are you concurrently running any CPU tasks? Yes. BOINC is running the CPU WU's at Low priority and the MW 0.59 ATI client at BelowNormal.
33) Message boards : News : ATI application updated again (Message 47709) Posted 12 Apr 2011 by Brian Priebe Post: ...Catalyst 11.3 drivers might still be worth a shot, as Windows XP 32-bit users have generally been having luck with these drivers. I can confirm that 11.3 has mostly solved the UI responsiveness problem but it's still on the order of 0.5sec sometimes to echo a character to the screen. It also looks like it's dropped the % activity on the GPU when MW is running. Will have to monitor for awhile.
34) Message boards : News : ATI application updated again (Message 47620) Posted 11 Apr 2011 by Brian Priebe Post: I am also seeing the customary negligible CPU time being used for GPU WU's even for version 0.59. However, what does appear new is an up to 2sec delay at the UI to show keystrokes echoed to the screen. This problem stops when the GPU WU ends and then starts up again when a new WU starts. This is happening with an ATI HD5870 under Windows/XP running Catalyst 10.12. Oddly enough, Catalyst claims GPU usage is less than 90% when this is happening.
35) Message boards : Number crunching : More of those 17_3s's... (Message 44169) Posted 24 Nov 2010 by Brian Priebe Post: The last one I ran generated "exit code -1073740940 (0xc0000374)" Sounds like something is very wrong with version 0.45. Your error indicates the dynamic memory heap has become corrupted.
36) Message boards : Number crunching : More of those 17_3s's... (Message 44162) Posted 24 Nov 2010 by Brian Priebe Post: 50% longer running (GPU)and proper credit :) True for the GPU WU's. But all of the 17_3s's I've seen processed by CPU app 0.45 (SSE2) fail instantly on an access violation.
37) Message boards : Number crunching : Torrents of Invalid WU's (Message 44113) Posted 23 Nov 2010 by Brian Priebe Post: Are you guys using optimized applications? These are 0.45 SSE2 plain vanilla applications failing here.
38) Message boards : Number crunching : boinc --allow_multiple_clients (Message 43990) Posted 20 Nov 2010 by Brian Priebe Post: For some reason its not seeing the HT on the 980 and only gets the 6 each for the main CPUs I had this problem with a XEON 5540 when I first enabled HT. The message log may claim it sees N cores but dispatches for only N/2. Or if you turn off HT, it claims it sees N/2 cores but still dispatches for N. For a Windows client, use the Advanced menu and do Run CPU Benchmarks. This was sufficient to reset whatever BOINC was using internally to correct the core count. You may have to do this every time you change the enable/disable state of HT. There is apparently a command line "--run_benchmarks" switch but whether this quits after running it I don't know.
39) Message boards : Number crunching : Torrents of Invalid WU's (Message 43980) Posted 19 Nov 2010 by Brian Priebe Post: I think I've been handed a batch of Invalids 'Completed Marked as Invalid'...CPU Onlys MW 0.45 sse2's. I'm getting the same thing after some have run for 10+ hours. It is a puzzlement...
40) Message boards : Number crunching : super charged? (Message 43699) Posted 10 Nov 2010 by Brian Priebe Post: Could it just be because we got the scheduler to correctly send out SSE2 applications, which are over twice as fast as the ones without SSE2? Unfortunately, the results scroll from the task view in very short order. The ATI GPU WU's seem to be generating a normal 2.38-ish credits/second for my HD5870. (Absolute max of 205,632/day possible.) The 9hr-long 0.45 SSE2 "de_separation_82_3s" WU that just got verified was awarded a normal-looking 213.78 credits. The two N-body WU's I can see got considerably less credit proportional to run time. (To my recollection, 0.19 was also turning in WU's in the 9-10hr time frame so is the SSE2 faster?) My RAC is now 232,600 and still climbing. So where is the extra 27,000 credits/day coming from? Even if both machines running CPU WU's were monopolized by MW, which they are not, they couldn't possibly do more than 85 WU/day.

Previous 20 · Next 20