| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Beta Testing
Forum: Beta Test Support Forum
Thread: OpenPandemics GPU Beta Test - Feb 18 2021 [Issues Thread] |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 272
|
|
| Author |
|
|
goben_2003
Advanced Cruncher Joined: Jun 16, 2006 Post Count: 146 Status: Offline Project Badges: 
|
Thanks Grumpy_Swede. I was comparing just CPU to CPU+GPU total performance. The CPU definitely slowed down, but the GPU more than made up for it. I did not check it versus the GPU running by itself though. I'll have to do that once we have work units again. I wonder if the results are still the same, or if it has changed in the last several generations(That was Ivy Bridge(3rd gen), 2 of my computers are Kaby Lake Refresh(8th gen), 1 is Coffee Lake Refresh(9th gen)).
---------------------------------------- |
||
|
|
Rickjb
Veteran Cruncher Australia Joined: Sep 17, 2006 Post Count: 666 Status: Offline Project Badges: 
|
Thanks, Grumpy Swede.
Now for a thought bubble for the techs: If they want to increase the diversity of GPU devices tested while they have limited numbers of beta tasks available, perhaps they could increase the quorum size, eg beyond the usual max of 2. |
||
|
|
Klaus-Dieter Vossberg
Cruncher Germany Joined: Apr 28, 2007 Post Count: 10 Status: Offline Project Badges:
|
This is aninformation for the WCG team:
I just recognized a strange behaviour for two of the WUs being calculated by my laptop. Despite being calculated soon after being received it says "Too late". I did not find a reason for this behaviour. Maybe it is interesting for you to know about this issue. Here a copy of the result status: BETA_OPN1_0020017_00115_0 LAPTOP-0GAITBQ3 Too Late 2/20/21 16:22:17 2/21/21 04:07:26 0.35 / 1.50 72.8 / 0.0 BETA_OPN1_0020004_04999_0 LAPTOP-0GAITBQ3 Too Late 2/19/21 15:33:41 2/19/21 17:28:07 0.02 / 0.10 72.8 / 0.0 All the best Klaus |
||
|
|
Grumpy Swede
Master Cruncher Svíþjóð Joined: Apr 10, 2020 Post Count: 2278 Status: Offline Project Badges: 
|
Yeah, I got one of those too (run on my iGPU HD4600). It's was a bunch of computers that errored out on the same WU, and I'm the only one who finished it, and I finished it first. Despite that, my WU came out as "Too late"
---------------------------------------- (should probably have read "too many errors" instead.) Yeah well, that's the things Beta is for, to find all the glitches.https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=538631913 [Edit 4 times, last edit by Grumpy Swede at Feb 27, 2021 2:15:07 PM] |
||
|
|
goben_2003
Advanced Cruncher Joined: Jun 16, 2006 Post Count: 146 Status: Offline Project Badges:
|
Edit: I have been running some Beta tasks here on my iGPU HD4600, and it works as expected. I can't say more than that, since you can't directly compare the larger GPU tasks with many more "jobs", to the CPU tasks with just a few "jobs". More WU's are needed before any more conclusions can be made. I was using jobs/hr as a comparison. My Intel UHD 620s averaged ~233 jobs/hr for the 30 work units that I recorded(with CPU fully loaded too). The same computers range from approximately .33 - 1 jobs/hr for a cpu work unit(no gpu running). They are i7s with 4 core / 8 thread, so 2.67 - 8 jobs/hr total. Now the problem is that jobs have variable number of dockings, and each docking is variable as well. So it is hard to do an exact comparison. That being said, I still think it is safe to say that the GPU still runs way faster than the CPU does. I will definitely have to try running just the GPU next time. It would only have to run a few % faster to make it so that more science is done than using the CPU + GPU. Of course, that would not help collect badges  , but would help science.  |
||
|
|
goben_2003
Advanced Cruncher Joined: Jun 16, 2006 Post Count: 146 Status: Offline Project Badges:
|
Be aware of, that running OpenCL WU's on your iGPU/APU, and at the same time running WU's on the CPU, will severely slow down tasks on both the iGPU, and the CPU. That was something we discovered on SETI@home years ago. Unless the OpenCL apps here, works differently, it's not a production increasing thing to run many (if any) WU's on the CPU, if you will crunch on the iGPU/APU also. See here enormous amounts of data from one very experienced OpenCL programmer who made many OpenCL apps for SETI@home, the posts are oldish, but afaik the conclusions are still very much valid: http://lunatics.kwsn.info/index.php/topic,1735.0.html You piqued my interest Grumpy_Swede. I am currently running those tests. I modified the scripts due to being able to run 8 CPU units simultaneously instead of 4. If people are interested I will post the results(in a new thread of course). I wonder if the AstroPulse 7 workloads are similar enough to openPandemics workloads for the results to be representative? |
||
|
|
[AF>Libristes]Maeda
Cruncher Joined: Sep 1, 2011 Post Count: 43 Status: Offline Project Badges: 
|
Yeah, I got one of those too (run on my iGPU HD4600). It's was a bunch of computers that errored out on the same WU, and I'm the only one who finished it, and I finished it first. Despite that, my WU came out as "Too late" (should probably have read "too many errors" instead.) Yeah well, that's the things Beta is for, to find all the glitches.https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=538631913 Indeed, all WU's failing on my HD4600 (on Linux), but good that it is working for somebody else. |
||
|
|
TribbleRED
Cruncher United States Joined: Dec 20, 2019 Post Count: 4 Status: Offline Project Badges: 
|
Any idea on the next batch of betas?
---------------------------------------- |
||
|
|
comancheace
Cruncher Joined: Aug 1, 2013 Post Count: 14 Status: Offline Project Badges: 
|
Hi! Good evening! I am only getting about 0.4 credits per gpu task. Is this normal or there is a problem with my computer? Can you explain? Thank you! Kind regards!
----------------------------------------  |
||
|
|
Vester
Senior Cruncher USA Joined: Nov 18, 2004 Post Count: 325 Status: Offline Project Badges:
|
uplinger said in the current Beta topic, " As a plus, one of the main developers of AutoDock has started to explore the points issue :)"
---------------------------------------- |
||
|
|
|