| 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 - March 5 2021 [ Issues Thread ] |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 200
|
|
| Author |
|
|
Bryn Mawr
Senior Cruncher Joined: Dec 26, 2018 Post Count: 345 Status: Offline Project Badges: 
|
6 WUs between the two GT710s all successful in between 2.23 hours and 3.59 hours which equates to a range of 4.6 and 5.4 minutes per job internally.
I’ll probably drop out of the GPU testing at this point as it hits my CPU processing quite badly. |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
Good morning everyone,
I'm seeing lots of posts about setting up custom plan classes for this to run multiple beta workunits at a single time. This to me sounds like everything is going well for this version. I have not looked at the stats, but I did check and apparently I did not load everything in yesterday. I show about 20k work units still left to load. I'm going to kick off the create work program to load more in now. There is a good chance that I will need to do this again later in the day. These workunits and results are going to be similar but give me more data to compare and review. Also, I have been seeing the posts about points and other reporting issues in the beta threads. At the moment, my primary goal is to get solid beta tests working and GPU out the door as soon as possible. WHEN I get some time, I will work towards fixing the credit issue and CPU time issue. Both of these should be granted and set from the validator, so it will be something on the backend. I am not planning on fixing credit previously granted. I wanted to up front and open with everyone on my plans towards points and time. I knew there would be some tweaks with that and hopefully I will be able to address it soon and before we go to production. Having said all of that, my first priority listed from the chats above is to figure out where the single and double precision errors are coming from. If I'm able to figure that out, then that would decrease the overall error rate for this application even more. For those customizing their settings for plan classes, this project will be released with shortname of opng. You are seeing the beta as beta29 right now. Thanks, -Uplinger |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges:
|
After loading in the next group, there are about 9k work units that still need to be loaded. I plan on doing that later today. I unfortunately do not have an exact timeline for that and I apologize.
Thank you again to everyone for your help and feedback, it is all very helpful! -Uplinger |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: 
|
Excellent. The best way to fix the credit system is to eliminate it entirely. That will eliminate all past and future confusion on the issue.
|
||
|
|
Richard Haselgrove
Senior Cruncher United Kingdom Joined: Feb 19, 2021 Post Count: 360 Status: Offline Project Badges: 
|
Good morning everyone, Good afternoon ;-)For those customizing their settings for plan classes, this project will be released with shortname of opng. You are seeing the beta as beta29 right now. Thanks, -Uplinger Got my first couple of beta29 (on Linux, this time). I'm going to experiment with 0.5 CPU, 0.5 GPU to start with. So, with 2 GPUs, I should run 4 tasks, using 2 CPU cores in support. There are two questions: 1) How much CPU does BOINC allocate for support? 2) How much CPU does your app need for support? 1 - BOINC's allocation (guesswork) is crazy. It takes no account of the actual programming language used, sync methods, or anything related to the actual work. It's usually in the high 90% range, which is neither fish nor fowl. Either the app responds well to having a free core, or it doesn't. 2 - I was seeing fairly high CPU usage for this particular app (NVidia version), but my gut feeling is that, while high in %age terms, it won't actually be driving that particular core very hard. Hence the 50% CPU test. |
||
|
|
ThreadRipper
Veteran Cruncher Sweden Joined: Apr 26, 2007 Post Count: 1321 Status: Offline Project Badges: 
|
Thanks uplinger for the update!
----------------------------------------Got some new tasks now. Running them on RX590, a new card I got for this project and I just installed it. So far it crunches away with AMD driver version 27.20.14535.3005  Join The International Team: https://www.worldcommunitygrid.org/team/viewTeamInfo.do?teamId=CK9RP1BKX1 AMD TR2990WX @ PBO, 64GB Quad 3200MHz 14-17-17-17-1T, RX6900XT @ Stock AMD 3800X @ PBO AMD 2700X @ 4GHz |
||
|
|
Grumpy Swede
Master Cruncher Svíþjóð Joined: Apr 10, 2020 Post Count: 2167 Status: Offline Project Badges: 
|
Thanks Uplinger. I just started my GPU cruncher, and got NVIDIA work first. Had buffer settings of 0,25 days +0,01, and nothing for the iGPU, despite is also asking for work.
----------------------------------------Lowered it to 0,05 +0,01, and the iGPU immediately grabbed one task. And as said yesterday, there's an issue with not wanting CPU tasks (set to NO). That setting seems to block receiving GPU tasks, or at least iGPU tasks. But the apps both NVIDIA and iGPU (HD4600) works as they should, at least for me. (I won't repeat the strange checkpoint behaviour though of the iGPU, or the hammering of the HD with checkpoints for the Nvidia., I think I've said enough of that, and that could be a later fix) Edit: And since the validator isn't running and validating the Beta tasks, the tasks does not adjust their expected runtime to the real runtime. My Nividia does a WU in 3-5 minutes, but the expected runtime for each downloaded task is still over an hour. That makes it diffcult to adjust the buffer to something that's even close to what you want.  Edit2: @Uplinger: Is there a limit set in Beta to how many WU's in total you can have in your cache? [Edit 4 times, last edit by Grumpy Swede at Mar 6, 2021 3:02:35 PM] |
||
|
|
Rickjb
Veteran Cruncher Australia Joined: Sep 17, 2006 Post Count: 666 Status: Offline Project Badges: 
|
Thanks, KU. I totally understand about credit awards being very low in your priorities.
I just scored some v1.26 WUs for my 3770K HD4000 iGPUs and 2 have started umm ..."running". They are behaving in the same manner as the v1.25 betas, ie they run the AutoGrid part and then hang when the GPU starts. Here's the stderr.txt file from the slot directory of one of them: ----- projects/www.worldcommunitygrid.org/wcgrid_beta29_autodockgpu_7.26_windows_x86_64__opencl_intel_gpu_102 -jobs OPNG_0022067_00211.job -input OPNG_0022067_00211.zip -seed 924393226 -wcgruns 1750 -wcgdpf 35 INFO: Using gpu device from app init data 0 INFO:[00:58:45] Start AutoGrid... autogrid4: Successful Completion. INFO:[00:59:41] End AutoGrid... INFO:[00:59:41] Start AutoDock for ZINC000846432955_RX1--6y84_002_gln110-rot--CYS156.dpf(Job #0)... OpenCL device: Intel(R) HD Graphics 4000 ----- EOF ------ Now time is 01:44, no further progress. I will let the 2 that are "running" sit there for a while, but I'll pass the unstarted others on to someone else. HTH - Rick PS: You are hereby awarded a koala stamp (rare) for dedication to the cause |
||
|
|
zombie67 [MM]
Senior Cruncher USA Joined: May 26, 2006 Post Count: 228 Status: Offline Project Badges: 
|
uplinger,
----------------------------------------Is there an app/plan class/whatever for "intel gpu + MacOS"? My Mac mini has not received any GPU tasks so far. I am just trying to understand if the issue is on my end, or if there is just no app.  |
||
|
|
torma99
Cruncher Hungary Joined: Mar 30, 2020 Post Count: 12 Status: Offline Project Badges: 
|
25 work units. 24 completed, one unfortunately run on error. First error of the beta tests.
Result Name: BETA_ OPNG_ 0022072_ 00192_ 1-- <core_client_version>7.16.11</core_client_version> <![CDATA[ <message> (unknown error) - exit code 4294967295 (0xffffffff)</message> <stderr_txt> projects/www.worldcommunitygrid.org/wcgrid_beta29_autodockgpu_7.26_windows_x86_64__opencl_nvidia_102 -jobs OPNG_0022072_00192.job -input OPNG_0022072_00192.zip -seed 855605790 -wcgruns 2600 -wcgdpf 52 INFO: Using gpu device from app init data 0 INFO:[15:25:19] Start AutoGrid... autogrid4: Successful Completion. INFO:[15:25:41] End AutoGrid... INFO:[15:25:42] Start AutoDock for ZINC001455425393_RX1--6y84_002_gln110-rot--CYS156.dpf(Job #0)... OpenCL device: GeForce RTX 2070 INFO:[15:25:47] End AutoDock... INFO:[15:25:47] Start AutoDock for ZINC001347255446_RX1--6y84_002_gln110-rot--CYS156.dpf(Job #1)... OpenCL device: GeForce RTX 2070 </stderr_txt> ]]> |
||
|
|
|