1)
Message boards :
Number crunching :
Will these GPUs crunch Seti
(Message 1160859)
Posted 10 Oct 2011 by ![]() Post: Hey guys thanks for the help. The lunatics app runs a bit higher, 75-85% with single instance. After some initial hiccups trying the 0.5 count- about 10-15 cuda instances tried to run at once and it ate up my system memory (a side effect of having more gpu memory than main memory) and the computer slowed to a crawl even though the display is attached to the AMD card. I reset the project and reinstalled the lunatics and this time successfully ran two cuda_fermi apps at once with 93-99% gpu. The thing is, right now I got a mix of cuda_fermi and cuda23 apps. I think the problem was that sometimes the gpu underclocks semipermanently to 270mhz core and 540 shader, then maybe boinc thinks its slow so it downloads those apps or is this normal? evga precision doesnt do too well with tesla driver- sometimes when the idle time is too long it gets buggy and never returns to full load speed. Is it ok to edit the appinfo.xml to exclude all but the two (x86 and x86_64) cuda_fermi apps like this? ...cpu stuff left intact... <app_version> <app_name>setiathome_enhanced</app_name> <version_num>610</version_num> <platform>windows_intelx86</platform> <plan_class>cuda_fermi</plan_class> <avg_ncpus>0.040000</avg_ncpus> <max_ncpus>0.040000</max_ncpus> <coproc> <type>CUDA</type> <count>0.5</count> </coproc> <file_ref> <file_name>Lunatics_x38g_win32_cuda32.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>cudart32_32_16.dll</file_name> </file_ref> <file_ref> <file_name>cufft32_32_16.dll</file_name> </file_ref> </app_version> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>610</version_num> <platform>windows_x86_64</platform> <plan_class>cuda_fermi</plan_class> <avg_ncpus>0.040000</avg_ncpus> <max_ncpus>0.040000</max_ncpus> <coproc> <type>CUDA</type> <count>0.5</count> </coproc> <file_ref> <file_name>Lunatics_x38g_win32_cuda32.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>cudart32_32_16.dll</file_name> </file_ref> <file_ref> <file_name>cufft32_32_16.dll</file_name> </file_ref> </app_version> or should I delete even the x86 one and see if it forces the x86_64 cuda_fermi to run exclusively? I've been experimenting but it's just that I've got a mix of every seti app class running at once and it's hard to see anything conclusive on the nvidia card. Every time I'm looking something different is running and some of the times the gpu driver is stuck on low clocks. I'm just glad I got a perfect pair of WU to run together this one instance :\ The SSE3x app sure helped, the cpu runtime was cut more than half. Again thanks for the help, it's much appreciated. |
2)
Message boards :
Number crunching :
Will these GPUs crunch Seti
(Message 1160764)
Posted 9 Oct 2011 by ![]() Post: I've noticed that the gpu utilization for the 6.10 (cuda_fermi) is 60-70% so I was trying to figure out how to make two run at once on the same card. I don't see an appinfo.xml and was wondering if its because a stock app doesn't have that? If so then I will create one, provided there's some examples out there that I can copy- standard settings but just two at once so I can see if it works out first. I've tried running two units of other projects on an AMD gaming card on stock clocks and high fan and throughout many driver updates still would freeze the system. Hopefully this low-clocked tesla is a little more stable so it can actually run two at once. |
3)
Message boards :
Number crunching :
No GPU WU's
(Message 1159931)
Posted 7 Oct 2011 by ![]() Post: Ah ok, thanks for the help. I guess I'll just set both cpu and gpu allowed now and just let it sort itself out. |
4)
Message boards :
Number crunching :
No GPU WU's
(Message 1159927)
Posted 7 Oct 2011 by ![]() Post: Is there something I need to do to make sure I get gpu units with stock app? I set the preferences on the account page to cpu - no and nvidia gpu - yes, it wouldn't give out work. If I set both to yes I would get some cpu work. Should I just try the lunatics installer 0.38 and see if it gets any units for either gpu (fermi and a cayman) or is this still a lack of workunits to send out thing? Thanks. |
©2025 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.