SETI MB CUDA for Linux

[Claggy]

I have another question:
CPU utilization is still almost 100%  for SETI and temp. on the GK stays low (50 C).

Task Status shows " Running (0.04% CPUs, 1 CUDA)", progress proceeds very slowly.

Is it possible that CPU is too weak to feed Graphic Card (Athlon X2 3800+) ?

It sounds as if the app is running in CPU fallback mode if it's running at 100%,
normally happens if GPU doesn't meet the memory requirements, when the first one completes, check your results.
I've been testing 6.08 (on windows) on Seti Beta with a 8400M GS, since it's only got 128Mb,
it goes into CPU fallback mode, and takes 4 to 7 hours per WU, a lot slower than stock 6.03, let alone AK_V8.
You could try different resolutions/colour depths to see if it makes a difference.

Claggy

[Claggy]

I have another question:
CPU utilization is still almost 100%  for SETI and temp. on the GK stays low (50 C).

Task Status shows " Running (0.04% CPUs, 1 CUDA)", progress proceeds very slowly.

Is it possible that CPU is too weak to feed Graphic Card (Athlon X2 3800+) ?

See ML1's post http://setiathome.berkeley.edu/forum_thread.php?id=51556&nowrap=true#858930

Claggy

[koschi]

The CUDA Linux app on my Q6600 with 9800GT (512MB) is also consuming 100% (so one full core) CPU time :-/
It is not working in fallback mode or like that, the times are 2-3 times faster than with the V8 app and GPU temperature is the same as when running GPUGRID.

http://setiathome.berkeley.edu/results.php?hostid=4425787&offset=200

[sunu]

The CUDA Linux app on my Q6600 with 9800GT (512MB) is also consuming 100% (so one full core) CPU time :-/

Yes, for performance reasons (both cuda app itself and general desktop responsiveness), Crunch3r made it that way .

[koschi]

Thanks, if that was intended I didn't say anything ;-)

[Raistmer]

The CUDA Linux app on my Q6600 with 9800GT (512MB) is also consuming 100% (so one full core) CPU time :-/

Yes, for performance reasons (both cuda app itself and general desktop responsiveness), Crunch3r made it that way .

Just wonder how to force CUDA MB to use 100% of CPU ??     
REally more likely it's CPU fallback mode...

[sunu]

The CUDA Linux app on my Q6600 with 9800GT (512MB) is also consuming 100% (so one full core) CPU time :-/

Yes, for performance reasons (both cuda app itself and general desktop responsiveness), Crunch3r made it that way .

Just wonder how to force CUDA MB to use 100% of CPU ??     
REally more likely it's CPU fallback mode...

Raistmer look at the pre-release section, thread about linux cuda, second page, last post (from Crunch3r) (don't think I should post a direct link here )

[koschi]

If its fallback mode, then the CUDA app has CPU code that is 2-3 times faster than AK V8... 
Check my results, WUs with angle range of ~0.43 take just ~19 minutes, those took close to 1h before on my machine...

VLARs I'm killing with a cron script 

[Raistmer]

Ok, I see.

Check my results, WUs with angle range of ~0.43 take just ~19 minutes, those took close to 1h before on my machine...

In task results I can see only CPU time, not wallclock time. 19min is big CPU time indeed for CUDA MB. It's interesting to know wallclock times for this build. Does additional usage of CPU lead to significant shrinkage wall clock time ?
Could anyone run PG*.wu test set with this build ?

[sunu]

Raistmer, this is what Crunch3r said:
P.S. yes, the app needs a full core, cuz i've removed that part of the code that messes around with that on windows.
It also slows down the gpu if you'd run a 4+1 config.

Since it uses a full core, cpu time is actually wall time. It's not like in windows where cpu time << wall time.

[Raistmer]

Raistmer, this is what Crunch3r said:
P.S. yes, the app needs a full core, cuz i've removed that part of the code that messes around with that on windows.
It also slows down the gpu if you'd run a 4+1 config.

Since it uses a full core, cpu time is actually wall time. It's not like in windows where cpu time << wall time.

OK, replace walltime by CPU time in my prev post. Anyway I interested inthat numbers.

[koschi]

I attached the CUDA start and stop messages from the log, run times are ~ the same I think...
Where to get this PG*.wu test set?

[attachment deleted by admin]

[Raistmer]

I attached the CUDA start and stop messages from the log, run times are ~ the same I think...
Where to get this PG*.wu test set?

Thik in download area of this site, in benchmarking sector.

[Claggy]

I had a look at Hefto99's http://setiathome.berkeley.edu/results.php?hostid=4774614host today, he's now completed two Cuda WU's with Chrunch3r's 608 r06 Cuda app,
both have completed on CPU fallback after 'CUDA runtime ERROR in device memory allocation' in about 15K secs (4hrs),
no credit granted yet.

Claggy

[Raistmer]

I had a look at Hefto99's http://setiathome.berkeley.edu/results.php?hostid=4774614host today, he's now completed two Cuda WU's with Chrunch3r's 608 r06 Cuda app,
both have completed on CPU fallback after 'CUDA runtime ERROR in device memory allocation' in about 15K secs (4hrs),
no credit granted yet.

Claggy

Hm.. As I expected... Still interesting in (clear) CUDA work with CPU 100% allocation times. I don't understand for what 100% CPU usage is needed for CUDA app indeed. That's why I want to look at elapsed times for tasks I know, not some "wild" task from server...