Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 
World Community Grid Forums
  Category: Active Research
  Forum: OpenPandemics - COVID-19 Project
  Thread: OpenPandemics - GPU Stress Test

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 781
Posts: 781   Pages: 79   [ Previous Page | 52 53 54 55 56 57 58 59 60 61 | Next Page ]
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 946260 times and has 780 replies Next Thread
spRocket
Senior Cruncher
Joined: Mar 25, 2020
Post Count: 280
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test
So..... Do you need to reserve 1 thread for each task run? I'm still a little confused by how much I should be reserving.


I'm doing 2 GPU + 14 CPU on Linux. In BOINC's computing preferences, I'm explicitly limiting CPU tasks to 14 by using 87% of the CPUs 100% of the time.

My /var/lib/boinc-client/projects/www.worldcommunitygrid.org/app_config.xml:
<app_config>
   <app>
      <name>opng</name>
      <max_concurrent>2</max_concurrent>
      <gpu_versions>
          <gpu_usage>.5</gpu_usage>
          <cpu_usage>.25</cpu_usage>
      </gpu_versions>
   </app>
</app_config>


The max_concurrent value is probably unnecessary, but it will hard-limit the system to 2 GPU tasks. The gpu_usage value divvies up the per-task GPU usage, so for two tasks, .5 is what you want. The cpu_usage value is for BOINC bookkeeping in determining how many CPUs are used by the GPU tasks; I found the .25 value by trial and error, but the end result is 14 CPU + 2 GPU on my Ryzen 7.

IMO, the configuration ought to be more straightforward than this, but that's something for the BOINC developers to worry about.
[May 1, 2021 1:09:35 AM]   Link   Report threatening or abusive post: please login first  Go to top 
True54Blue
Advanced Cruncher
Joined: Nov 17, 2004
Post Count: 97
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test
I'm now running 16 GPU tasks with 1 CPU thread each. BOINC has automatically paused the CPU tasks for those threads. When we had the server issue and GPU WUs ran out for me on 4/30 it switched back to CPU tasks and also managed to get some iGPU tasks. It's now running 16 GPU tasks and 1 iGPU task but the iGPU is slow as the CPU is pretty busy feeding the RTX3060Ti.
----------------------------------------
[May 1, 2021 1:35:11 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Pandelta
Advanced Cruncher
Joined: Jun 24, 2012
Post Count: 55
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test
@Uplinger
I hope you are getting some well deserved sleep :) You all have done an awesome job and this has been a blast! When you get a chance can you give an update on how much of the stress test we have completed?
----------------------------------------
[Edit 1 times, last edit by Pandelta at May 1, 2021 2:56:40 AM]
[May 1, 2021 2:55:04 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Jessica Keyes
Cruncher
Joined: Mar 25, 2020
Post Count: 1
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test
Do these units require single or double precision?
[May 1, 2021 2:58:46 AM]   Link   Report threatening or abusive post: please login first  Go to top 
poppinfresh99
Cruncher
Joined: Feb 29, 2020
Post Count: 49
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test
Do these units require single or double precision?

I have a couple Intel GPUs that do not support hardware double-precision. They run and complete tasks just fine!
----------------------------------------
[Edit 1 times, last edit by poppinfresh99 at May 1, 2021 5:00:41 AM]
[May 1, 2021 5:00:10 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test
Wondering if this issue is related to the GPU Stress Test?

I've noticed an issue an issue since 04/27/2021 with some OPN units awarding outrageously large point totals vs points claimed during the last several days. I first noticed in on 04/27/2021 when my daily Points Generated total increased 2-fold from typical values, and it is continuing through today 05/01/2021 (1:00 AM EDT)


Typical result:
OPN1 00434009 018690 1.97/3.88 80.8 claimed/82.7 awarded

Results with huge points awarded incorrectly:
OPNG 0020127 000880 0.15/3.13 1.0 claimed/1,655.9 awarded
OPNG 0023654 001141 0.15/2.84 0.9 claimed/1,639.9 awarded

I'm running World Community Grid BOINC on two cores of a quad-core Intel i5 -2400S (Sandy Bridge) in a mid-2011 iMac with a AMD Radeon 6750M GPU and macOS 10.13.x High Sierra.
[May 1, 2021 5:22:24 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Falconet
Master Cruncher
Portugal
Joined: Mar 9, 2009
Post Count: 3315
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test
Wondering if this issue is related to the GPU Stress Test?

I've noticed an issue an issue since 04/27/2021 with some OPN units awarding outrageously large point totals vs points claimed during the last several days. I first noticed in on 04/27/2021 when my daily Points Generated total increased 2-fold from typical values, and it is continuing through today 05/01/2021 (1:00 AM EDT)


Typical result:
OPN1 00434009 018690 1.97/3.88 80.8 claimed/82.7 awarded

Results with huge points awarded incorrectly:
OPNG 0020127 000880 0.15/3.13 1.0 claimed/1,655.9 awarded
OPNG 0023654 001141 0.15/2.84 0.9 claimed/1,639.9 awarded

I'm running World Community Grid BOINC on two cores of a quad-core Intel i5 -2400S (Sandy Bridge) in a mid-2011 iMac with a AMD Radeon 6750M GPU and macOS 10.13.x High Sierra.




Those points are fine. Each GPU work unit has dozens or hundreds of jobs inside it while a CPU work unit only has 1 or 2.
----------------------------------------


- AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W
- AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W
- AMD Ryzen 7 7730U 8C/16T 3.0 GHz
[May 1, 2021 7:02:31 AM]   Link   Report threatening or abusive post: please login first  Go to top 
zdnko
Senior Cruncher
Joined: Dec 1, 2005
Post Count: 235
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test 
Result Name: OPNG_ 0014723_ 00094_ 0--
<core_client_version>7.6.33</core_client_version>
<![CDATA[
<stderr_txt>
projects/www.worldcommunitygrid.org/wcgrid_opng_autodockgpu_7.28_windows_x86_64__opencl_intel_gpu_102 -jobs OPNG_0014723_00094.job -input OPNG_0014723_00094.zip -seed 1915770179 -wcgruns 4650 -wcgdpf 93
INFO: Using gpu device from app init data 0
INFO:[09:27:14] Start AutoGrid...

autogrid4: Successful Completion.
INFO:[09:37:16] End AutoGrid...
INFO:[09:37:17] Start AutoDock for OB3ZINC000174426751--7jji_002_mgltools--TYR380_inert.dpf(Job #0)...
OpenCL device: Intel(R) HD Graphics 620
11 warnings generated.
warning: Linking two modules of different data layouts: '' is 'e-m:e-i64:64-f80:128-n8:16:32:64-S128' whereas '<origin>' is 'e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-n8:16:32' 

warning: Linking two modules of different target triples: ' is 'x86_64-pc-windows-msvc' whereas '<origin>' is 'igil_64_GEN9'
 
INFO:[09:38:58] End AutoDock...
INFO:[09:38:59] Start AutoDock for OB3ZINC000846549099_2--7jji_002_mgltools--TYR380_inert.dpf(Job #1)...
OpenCL device: Intel(R) HD Graphics 620
11 warnings generated.
warning: Linking two modules of different data layouts: '' is 'e-m:e-i64:64-f80:128-n8:16:32:64-S128' whereas '<origin>' is 'e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-n8:16:32' 

warning: Linking two modules of different target triples: ' is 'x86_64-pc-windows-msvc' whereas '<origin>' is 'igil_64_GEN9'
...

What mean that warnings?
Is it a problem with the wu or the computer?

In this moment the wu status is: 1 Invalid, 1 Error and 1 PV.
https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=640390549

PS: the warnings are in the PV log (Win10 Enterprise x64 Edition, (10.00.18363.00))
[May 1, 2021 7:31:30 AM]   Link   Report threatening or abusive post: please login first  Go to top 
squid
Advanced Cruncher
Germany
Joined: May 15, 2020
Post Count: 56
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test
Try the newest World Community Grid Client 7.14.3.
[May 1, 2021 8:22:46 AM]   Link   Report threatening or abusive post: please login first  Go to top 
siu77
Cruncher
Russia
Joined: Mar 12, 2012
Post Count: 23
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: OpenPandemics - GPU Stress Test
So..... Do you need to reserve 1 thread for each task run?
The processor must keep up with feeding the video card. The leftovers can be used painlessly for CPU-only jobs.

I have 4-core i5 with HD4600 and an R7 260x.
<app_config>
<app>
<name>opng</name>
<gpu_versions>
<gpu_usage>.33</gpu_usage>
<cpu_usage>.49</cpu_usage>
</gpu_versions>
</app>
</app_config>
.49 * 6 ( 3 GPU-tasks for AMD and 3 GPU-tasks for Intel ) = 2.94 - i.e., 4 - 2 = 2 cores can do CPU-only tasks.
.5 * 6 = 3 - i.e., 4 - 3 = 1 core can do CPU-only task.

It depends on your hardware.
[May 1, 2021 9:16:47 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Posts: 781   Pages: 79   [ Previous Page | 52 53 54 55 56 57 58 59 60 61 | Next Page ]
[ Jump to Last Post ]
Post new Thread