Message boards :
Number crunching :
Driver difference?
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 26 Jun 09 Posts: 815 Credit: 1,470,385,294 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
I have Nathan WU's, short run (cuda 3.1) on a GTX285, driver 310.90 and takes about 16000-17000 seconds to complete. Another Nathan WU (same type) also short run (cuda 3.1) on a GTX550Ti , driver 306.97 takes about 41600 seconds to complete. I know the GTX285 is a better card than the newer GTX550Ti but I didn't expect this difference. Moreover as with Einstein@home (cuda) run time is almost the same. Could this be the result of the driver? The GTX550Ti did the beta's just within an hour. Greetings from TJ |
dskagcommunitySend message Joined: 28 Apr 11 Posts: 463 Credit: 958,266,958 RAC: 34 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
The 41xxx secs sounds ways too long. But you hide your systems so i cant look into you general Stats. DSKAG Austria Research Team: http://www.research.dskag.at
|
microchipSend message Joined: 4 Sep 11 Posts: 110 Credit: 326,102,587 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
I think it is because of Compute Capability thus the GTX 285 can be fully used by GPUGRID while the GTX 550Ti can't... Correct me if I'm wrong. Team Belgium |
skgivenSend message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
The 4.2app is about twice as fast as the 3.1app on a GTX550Ti. Don't use the 3.1app on a GTX550Ti; run WU's on the 4.2app. For the GTX285 the 3.1app is faster - it being an older card, relying on older code. FAQ's HOW TO: - Opt out of Beta Tests - Ask for Help |
|
Send message Joined: 26 Jun 09 Posts: 815 Credit: 1,470,385,294 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
The 4.2app is about twice as fast as the 3.1app on a GTX550Ti. Thanks for your response skgiven. I followed your advice and have for that pc set to accept beta and cuda 4.2. This has resulted in the following messages: 20/02/2013 15:03:23 GPUGRID Requesting new tasks for CPU and GPU 20/02/2013 15:03:24 GPUGRID Scheduler request completed: got 0 new tasks 20/02/2013 15:03:24 GPUGRID Message from server: No tasks sent 20/02/2013 15:03:24 GPUGRID Message from server: No tasks are available for Short runs (2-3 hours on fastest card) 20/02/2013 15:03:24 GPUGRID Message from server: No tasks are available for ACEMD beta version 20/02/2013 15:03:24 GPUGRID Message from server: No tasks are available for the applications you have selected. Is this due to the 306.97 driver? It runs Einstein@home without any issues as well GPUGRID, with good system response. Greetings from TJ |
microchipSend message Joined: 4 Sep 11 Posts: 110 Credit: 326,102,587 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
The 4.2app is about twice as fast as the 3.1app on a GTX550Ti. No, it's not a driver issue. It's a short queue issue. I get the same reported when trying to fetch from the short 4.2 queue. Server status says plenty of tasks but it's lying so don't trust it. Team Belgium |
|
Send message Joined: 26 Jun 09 Posts: 815 Credit: 1,470,385,294 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
No, it's not a driver issue. It's a short queue issue. I get the same reported when trying to fetch from the short 4.2 queue. Server status says plenty of tasks but it's lying so don't trust it. That's weird. I will make a new thread then and see if we get helpful responses. Greetings from TJ |
|
Send message Joined: 26 Jun 09 Posts: 815 Credit: 1,470,385,294 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
It could be driver issue, as I have now updated to nVidia 314.07, I got a cuda42 short run for the GTX550Ti. Greetings from TJ |
|
Send message Joined: 26 Jun 09 Posts: 815 Credit: 1,470,385,294 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
The cuda42 short runs, run happily on my 550TI in 25xxx seconds, compared to 46xxx seconds for cuda31 WU's. Greetings from TJ |
©2025 Universitat Pompeu Fabra