COVID-19 Project Not Sending GPU Tasks - My Config or a Project Issue?

Message boards : GPUs : COVID-19 Project Not Sending GPU Tasks - My Config or a Project Issue?
Message board moderation

To post messages, you must log in.

AuthorMessage
danielito76

Send message
Joined: 22 Oct 21
Posts: 2
United States
Message 105831 - Posted: 22 Oct 2021, 18:13:21 UTC

Howdy. I am looking to get some understanding on why my BOINC client isn't picking up any GPU tasks. I've tried to use local prefs and web prefs to specify the use of GPUs for work. I saw the latest update from the project saying there is work:

https://www.worldcommunitygrid.org/about_us/article.s?articleId=742

...so I believe work for my GPU is available. Below are the first few lines of my client log. I have an OS X machine and an AMD 6900 XT GPU (along with an internal one being called out as being ignored). Is there a problem with my client or should I check with the COVID-19 project folks for suggestions? Thanks in advance!

Fri Oct 22 11:07:19 2021 | | Starting BOINC client version 7.16.19 for x86_64-apple-darwin
Fri Oct 22 11:07:19 2021 | | log flags: file_xfer, task, task_debug
Fri Oct 22 11:07:19 2021 | | Libraries: libcurl/7.73.0 OpenSSL/1.1.0l zlib/1.2.11 c-ares/1.13.0
Fri Oct 22 11:07:19 2021 | | Data directory: /Library/Application Support/BOINC Data
Fri Oct 22 11:07:19 2021 | | OpenCL: AMD/ATI GPU 0: AMD Radeon RX 6900 XT Compute Engine (driver version 1.2 (Aug 30 2021 06:56:17), device version OpenCL 1.2, 16368MB, 16368MB available, 17024 GFLOPS peak)
Fri Oct 22 11:07:19 2021 | | OpenCL: AMD/ATI GPU 1 (ignored by config): AMD Radeon Pro Vega 48 Compute Engine (driver version 1.2 (Aug 30 2021 06:56:38), device version OpenCL 1.2, 8176MB, 8176MB available, 4608 GFLOPS peak)
Fri Oct 22 11:07:19 2021 | | OpenCL CPU: Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz (OpenCL driver vendor: Apple, driver version 1.1, device version OpenCL 1.2)
Fri Oct 22 11:07:19 2021 | | Host name: iMac.local
Fri Oct 22 11:07:19 2021 | | Processor: 16 GenuineIntel Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz [x86 Family 6 Model 158 Stepping 13]
Fri Oct 22 11:07:19 2021 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clfsh ds acpi mmx fxsr sse sse2 ss htt tm pbe pni pclmulqdq dtes64 mon dscpl vmx smx est tm2 ssse3 fma cx16 tpr pdcm sse4_1 sse4_2 x2apic movbe popcnt aes pcid xsave osxsave seglim64 tsctmr avx rdrand f16c
Fri Oct 22 11:07:19 2021 | | OS: Mac OS X 11.6 (Darwin 20.6.0)
Fri Oct 22 11:07:19 2021 | | Memory: 32.00 GB physical, 563.13 GB virtual
Fri Oct 22 11:07:19 2021 | | Disk: 1.55 TB total, 545.37 GB free
Fri Oct 22 11:07:19 2021 | | Local time is UTC -7 hours
Fri Oct 22 11:07:19 2021 | World Community Grid | General prefs: from World Community Grid (last modified 15-Oct-2021 09:46:58)
Fri Oct 22 11:07:19 2021 | World Community Grid | Computer location: work
Fri Oct 22 11:07:19 2021 | | General prefs: using separate prefs for work
Fri Oct 22 11:07:19 2021 | | Reading preferences override file
Fri Oct 22 11:07:19 2021 | | Preferences:
Fri Oct 22 11:07:19 2021 | | max memory usage when active: 32768.00 MB
Fri Oct 22 11:07:19 2021 | | max memory usage when idle: 32768.00 MB
Fri Oct 22 11:07:19 2021 | | max disk usage: 10.00 GB
Fri Oct 22 11:07:19 2021 | | (to change preferences, visit a project web site or select Preferences in the Manager)
Fri Oct 22 11:07:19 2021 | | Setting up project and slot directories
Fri Oct 22 11:07:19 2021 | | Checking active tasks
Fri Oct 22 11:07:19 2021 | World Community Grid | URL http://www.worldcommunitygrid.org/; Computer ID 6690094; resource share 100
Fri Oct 22 11:07:19 2021 | | Setting up GUI RPC socket
Fri Oct 22 11:07:19 2021 | | Checking presence of 590 project files
Fri Oct 22 11:07:19 2021 | World Community Grid | [task] ACTIVE_TASK::start(): forked process: pid 2716
Fri Oct 22 11:07:19 2021 | World Community Grid | [task] task_state=EXECUTING for OPN1_0081916_02755_1 from start
ID: 105831 · Report as offensive
Profile Joseph Stateson
Volunteer tester
Avatar

Send message
Joined: 27 Jun 08
Posts: 611
United States
Message 105832 - Posted: 22 Oct 2021, 19:11:58 UTC

GPU work is intermittent. Maybe a coincidence but I had the same question myself and just went to their site and poked around looking. In the mean time I picked up a few ATI tasks from them with not having to do anything.

I assume you don't want to use your Vega for crunching, only the other, slower card.
Currently, I have 10 cpu tasks and 2 ati tasks running WCG with 7 other ati tasks read to run. A few minutes ago there were no ati tasks.

My default profile for the above system is nvidia,ati: yes; intel: no; cpu: yes
I only do the covid


Some projects will not send tasks to system with cards that are too slow and cannot complete before the deadline. I do not know if WCG does that.
ID: 105832 · Report as offensive
danielito76

Send message
Joined: 22 Oct 21
Posts: 2
United States
Message 105836 - Posted: 24 Oct 2021, 0:29:16 UTC - in response to Message 105832.  

Thanks for the tip! I will try playing with the settings,
ID: 105836 · Report as offensive
Goesnet

Send message
Joined: 16 Dec 21
Posts: 1
United States
Message 106437 - Posted: 16 Dec 2021, 3:02:57 UTC

Alright you really smart people out there:

I have been Boincing since the 90's but mostly astrophysics (I really miss Seti@home). I want to give SiDock some love, but that leaves me with 25 Tesla cards sitting on their butt. These cards a great at the space stuff, but suck at folding@home and the low precision stuff.

WCG rarely kick out any gpu work - certainly not enough to feed these cards, and the rest of the projects are cpu only.

Is there any magic out there that will let these cards emulate a cpu to help with the covid effort?
ID: 106437 · Report as offensive
robsmith
Volunteer tester
Help desk expert

Send message
Joined: 25 May 09
Posts: 1056
United Kingdom
Message 106439 - Posted: 16 Dec 2021, 8:35:39 UTC - in response to Message 106437.  

No such magic exists
ID: 106439 · Report as offensive
Ian&Steve C.

Send message
Joined: 24 Dec 19
Posts: 198
United States
Message 106447 - Posted: 17 Dec 2021, 16:13:25 UTC - in response to Message 106439.  
Last modified: 17 Dec 2021, 16:14:45 UTC

No such magic exists


and if it did, it would probably be slower than just running on a CPU.

GPU architecture is fundamentally different than a CPU. you need to code for them a totally different way.

Goesnet, if you have Tesla's with decent double precision performance, Milkyway will be the best home for them.

they might also do well on Einstein@home, which has tons of work available all the time.
ID: 106447 · Report as offensive

Message boards : GPUs : COVID-19 Project Not Sending GPU Tasks - My Config or a Project Issue?

Copyright © 2022 University of California. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.