Message boards :
Number crunching :
FM1 LLano
Message board moderation
Author | Message |
---|---|
garfield Send message Joined: 4 Jan 02 Posts: 45 Credit: 7,409,265 RAC: 65 |
Hi, yesterday I assembled my new system, an A8 4Core with onboard HD 6550D Since I want to use this system also for Albert@home, I installed BM 7.0.11, latest available CCC from the AMD Homepage. I installed Lunatice optimised apps, using the settings for HD5XXX cards. My system has a HD5830 and the On-chip HD6550. A single Open CL wu finished yesterday, now all wu's start, work for ~15 sec's, and then stop. No error, they start again after a short time, but they do not crunch. The message log says 'Restarting task ... using setiathome_enhanced version 610(ati13ati)' No special settings to run more than one wu at a time. Any idea where I should start to look at? Alexander |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
That sounds like you have your computer settings for run while Computer and use GPU while computer is in use clicked. Check your computing settings on your account page In a rich man's house there is no place to spit but his face. Diogenes Of Sinope |
garfield Send message Joined: 4 Jan 02 Posts: 45 Credit: 7,409,265 RAC: 65 |
Hi skildude, sorry, I checked the settings, they seem to be ok, and Albert@home is running fine. The seti wu's try to start and switch back to 'verdrängt' after 13 - 15 secs. |
ML1 Send message Joined: 25 Nov 01 Posts: 20372 Credit: 7,508,002 RAC: 20 |
Not enough GPU memory available for the s@h tasks?... Happy fast crunchin', Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
... The stderr.txt file produced by the application should have some interesting entries. It can be found in the slot directory, and if you select a task in BOINC Manager then click "Properties" the information shown should include a line indicating which slot directory is in use for that task. Joe |
garfield Send message Joined: 4 Jan 02 Posts: 45 Credit: 7,409,265 RAC: 65 |
Windows optimized S@H Enhanced application by Alex Kan Version info: v8b2-SSE3x (AMD/Intel, Core 2-optimized v8-nographics) V5.13 by Alex Kan SSE3x Win32 Build 386 , Ported by : Jason G, Raistmer, JDWhale CPUID: AMD A8-3870 APU with Radeon(tm) HD Graphics Speed: 1 x 3000 MHz Cache: L1=64K L2=1024K Features: MMX SSE SSE2 SSE3 Work Unit Info: ............... Credit multiplier is : 2.85 WU true angle range is : 0.431849 boinc_exit(): requesting safe worker shutdown -> boinc_exit(): worker didn't respond to exit request within 2 seconds, exiting anyway. Restarted at 3.85 percent. boinc_exit(): requesting safe worker shutdown -> boinc_exit(): worker didn't respond to exit request within 2 seconds, exiting anyway. Will this help? |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
Windows optimized S@H Enhanced application by Alex Kan That looks like it's from the CPU AK_v8b2 application, so doesn't say much about the OpenCL application for ATI GPU. Your 'Restarting task ... using setiathome_enhanced version 610(ati13ati)' note implies boinc_temporary_exit() is being used. I was hoping a stderr.txt from the OpenCL app would also show why the app wasn't able to run. The BOINC wiki page assumes not enough memory available and that's the most likely thing, but other kinds of initialization problems may cause the application to ask BOINC to try again a little later. Joe |
garfield Send message Joined: 4 Jan 02 Posts: 45 Credit: 7,409,265 RAC: 65 |
Joe, many thx for your efford to help. I cannot believe that mem is a issue, more likely some other process. I've seen a short message saying 'scheduler wait' in the status-column, diappearing after a couple of seconds. I've tried to put the other AMD-GPU project to hold; did not change anything. Albert@home wu's run fine. 01.02.2012 13:41:12 | | Processor: 4 AuthenticAMD AMD A8-3870 APU with Radeon(tm) HD Graphics [Family 18 Model 1 Stepping 0] 01.02.2012 13:41:12 | | Processor: 1.00 MB cache 01.02.2012 13:41:12 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 htt pni cx16 syscall nx lm svm sse4a osvw ibs skinit wdt page1gb rdtscp 3dnowext 3dnow 01.02.2012 13:41:12 | | OS: Microsoft Windows 7: Home Premium x64 Edition, Service Pack 1, (06.01.7601.00) 01.02.2012 13:41:12 | | Memory: 7.50 GB physical, 14.99 GB virtual 01.02.2012 13:41:12 | | Disk: 111.69 GB total, 73.43 GB free 01.02.2012 13:41:12 | | Local time is UTC +1 hours 01.02.2012 13:41:12 | | VirtualBox version: 4.1.8 01.02.2012 13:41:12 | | NVIDIA GPU 0: GeForce 9500 GT (driver version 285.62, CUDA version 4.10, compute capability 1.1, 512MB, 473MB available, 130 GFLOPS peak) 01.02.2012 13:41:12 | | ATI GPU 0: ATI Radeon HD 5800 series (Cypress) (CAL version 1.4.1664, 1024MB, 991MB available, 2688 GFLOPS peak) 01.02.2012 13:41:12 | | ATI GPU 1: AMD SUPERSUMO (CAL version 1.4.1664, 512MB, 479MB available, 960 GFLOPS peak) 01.02.2012 13:41:12 | | OpenCL: NVIDIA GPU 0: GeForce 9500 GT (driver version 285.62, device version OpenCL 1.0 CUDA, 512MB, 473MB available) 01.02.2012 13:41:12 | | OpenCL: ATI GPU 0: Cypress (driver version CAL 1.4.1664 (VM), device version OpenCL 1.1 AMD-APP (851.4), 1024MB, 991MB available) 01.02.2012 13:41:12 | | OpenCL: ATI GPU 1: BeaverCreek (driver version CAL 1.4.1664 (VM), device version OpenCL 1.1 AMD-APP (851.4), 512MB, 479MB available) 01.02.2012 13:41:12 | | NVIDIA GPU is OpenCL-capable 01.02.2012 13:41:12 | | ATI GPU is OpenCL-capable Alexander |
garfield Send message Joined: 4 Jan 02 Posts: 45 Credit: 7,409,265 RAC: 65 |
I've updated now to BM 7.0.12 not exact the same behaviour, one wu has now a very long stderr.txt here is a link to it: http://dl.dropbox.com/u/50246791/stderr.txt And it contains the line ERROR: OpenCL kernel/call 'Enqueueing kernel:PC_find_spike_kernel_cl' call failed (-55) in file ..\analyzeFuncs.cpp near line 3106. Hope this helps! Alexander |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
I've updated now to BM 7.0.12 Hmm, the reported error is fairly inscrutable. Raistmer has pointed out elsewhere that -55 is the numerical code for CL_INVALID_WORK_ITEM_SIZE but the selected ATI GPU has "Max work group size: 256" so that should not be the problem. It's curious that so little information is presented for the BeaverCreek GPU within the APU, but that doesn't seem likely to be the cause. I'm out of my depth on this, but I'll note that others who have had unexplained -55 errors have cured them by changing to different drivers. And I'll also note that with both ATI and NVIDIA GPUs in the system, ATI drivers after NV drivers is the install order which users have found actually works. Beyond that, I hope those with some practical OpenCL experience will look for clues in the stderr.txt you captured. If the task happened to finish, BOINC would only report the last 64KB. That would actually be fairly complete, the over 500KB has 126 iterations of the same cycle, but it's better to know than have to guess it's all repetitive. Joe |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
-55 is the invalid workgroup size error indeed and in this case can be caused mixed ATi/NV setup. To confirm this try to remove NV GPU from PC and restart that task. Will error appear? EDIT: one possibility would be to switch GPUs inside case (physically) to get different reporting order for OpenCL enumeration. Also, reinstalling AMD driver after NV one could give different order too. The switch to non-HD5 build would not help but switching to LHD4K build could be helpful (though this variant definitely will lower host performance for ATi GPUs so better to switch cards for now). Hope this issue will be removed in upcoming build. |
garfield Send message Joined: 4 Jan 02 Posts: 45 Credit: 7,409,265 RAC: 65 |
THX for all the help. I've set my nVidia project to no new work. When the remaining tasks are done I'll try to switch the cards or even remove the nvidia. BTW, the drivers were installed in the posted order, nVidia first and AMD afterwards. Alexander |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
THX for all the help. I'm testing build with fix for your issue so maybe GPU removal would be not needed soon... |
garfield Send message Joined: 4 Jan 02 Posts: 45 Credit: 7,409,265 RAC: 65 |
well, removing the nvidia card did the trick ... Exchanging the pci-e slots is a problem due to available space. So, what is the source of the problem and how can that be solved? Different settings in the lunatics installer or is it a scheduler-problem, that needs to be postet in the boinc-forum? Alexander Edit: a couple of seconds too late, Raistmer already posted an answer ... |
garfield Send message Joined: 4 Jan 02 Posts: 45 Credit: 7,409,265 RAC: 65 |
I've got the new built, it's working now. http://setiathome.berkeley.edu/show_host_detail.php?hostid=6392383 THX a lot for the help! Alexander |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
If someone met the same problem, here is the link to updated binaries: http://dl.dropbox.com/u/60381958/MB7_r426hd5.7z Bugfix will be included in next ordinary release of course. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.