Seti@Home optimized science apps and information
Optimized Seti@Home apps => Windows => GPU crunching => Topic started by: codeman05 on 18 Apr 2009, 12:24:26 am
-
I have a new rig that I just put together with some spare parts:
AMD 5600+ Dual Core
1x 8800GT
Windows XP
Boinc 6.6.20
Link to host details (http://setiathome.berkeley.edu/show_host_detail.php?hostid=4886222) (only been running CUDA for about 36hrs)
Crunching MB only on both
I noticed that the CUDA WU completion times seemed very slow...about 2-3x slower than 8800gs cards in another rig and 25-50% slower than the CPU MB units.
So I did some investigating...It looks like I have two CPU AK proccesses running, but only one is allocated CPU time...take a look at the screenshot below...and for some reason two CUDA MB proccess?
I have a dual core e5200 running with a pair of 8800GS GPUs. On that rig, both AK processes are utilized 100% (50% each with CUDA fighting for a few %) and CUDA WU's are completing in ~15-20 mins/gpu...slower cards.
Any idea on what may be going on?
Thanks
[attachment deleted by admin]
-
here is a copy of the app_info.xml
<app_info>
<app>
<name>astropulse</name>
</app>
<file_info>
<name>ap_5.00r103_SSE3.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse</app_name>
<version_num>500</version_num>
<flops>6321525975</flops>
<file_ref>
<file_name>ap_5.00r103_SSE3.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>astropulse_v5</name>
</app>
<file_info>
<name>ap_5.03r112_SSE3.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse_v5</app_name>
<version_num>503</version_num>
<flops>7304874460</flops>
<file_ref>
<file_name>ap_5.03r112_SSE3.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>AK_v8b_win_SSE3_AMD_GPU_CPU_team_V10.exe</name>
<executable/>
</file_info>
<file_info>
<name>MB_6.08_mod_CUDA_V10.exe</name>
<executable/>
</file_info>
<file_info>
<name>cudart.dll</name>
<executable/>
</file_info>
<file_info>
<name>cufft.dll</name>
<executable/>
</file_info>
<file_info>
<name>libfftw3f-3-1-1a_upx.dll</name>
<executable/>
</file_info>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>603</version_num>
<platform>windows_intelx86</platform>
<flops>4916742425</flops>
<file_ref>
<file_name>AK_v8b_win_SSE3_AMD_GPU_CPU_team_V10.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.127970</avg_ncpus>
<max_ncpus>0.127970</max_ncpus>
<flops>13000000000</flops>
<plan_class>cuda</plan_class>
<file_ref>
<file_name>MB_6.08_mod_CUDA_V10.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart.dll</file_name>
</file_ref>
<file_ref>
<file_name>cufft.dll</file_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
</file_ref>
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>
</app_version>
</app_info>
-
Yuo should use "teamed" AK_v8 with app_info, supplied wt the pack.
Or you could use some another app_info but ordinal AK_v8.
And check if you set cc_config correctly if you continue with V10 pack.
-
ok, to be sure I haven't messed anything up, I reloaded your "Raistmer's_opt_pack_V10a_x86_SSE3_AMD_MB_CPU_GPU_team.rar" bone stock, no modifications to the app_info and the following in the cc_config:
<cc_config>
<options>
<ncpus>3</ncpus>
</options>
</cc_config>
so far it looks like it's running correctly. 3 WU's crunch simultaneously and I now see 2 AK processes and a MB process...all being utilized...in the task manager.
Weird thing is that all of the workunits in BOINC show to be enhanced 6.08, no mention of CUDA.
CUDA is obviously running though because now one of the 3 processes is crunching WU's in 7 minutes vs ~50min for the other 2 (cpu). Muuuuccchh better.
Thanks for all of your work on this and your patience with all the questions ;)