GPUGrid is in need of CPU Volunteers

Message boards : Number crunching : GPUGrid is in need of CPU Volunteers
Message board moderation

To post messages, you must log in.

AuthorMessage
PappaLitto
Volunteer tester

Send message
Joined: 27 Jul 15
Posts: 8
Credit: 1,379,260
RAC: 0
United States
Message 1934364 - Posted: 8 May 2018, 11:11:34 UTC

Hello, GPUGrid is a largely GPU based Biomedical research BOINC project but more recently they have integrated CPU support which uses a different application that computes a different type of research called Quantum Chemistry. Currently, this application only runs on Linux of all distributions. They are very short on CPU volunteers and they would love your help.
ID: 1934364 · Report as offensive
Profile tullio Project Donor
Volunteer moderator
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 6893
Credit: 2,015,012
RAC: 645
Italy
Message 1934391 - Posted: 8 May 2018, 13:16:49 UTC
Last modified: 8 May 2018, 14:40:58 UTC

I joined them but they have sent me also a GPU task which is running on my GTX 750 Ti and a 2 CPU task which is waiting to start after a LHC@home task running on one CPU finishes.I wonder if the GPU task and the 2 cores CPU task can coexist, since my Opteron 1210 has only two cores. SuSE Linux Leap 42.3.
Tullio
The Quantum Chemistry 2 CPU task has started and is running as python alongside the GPU task. The LHC task is waiting.
The @ CPU task has erred after about 5 minutes. I think you need VirtualBox installed, which I have for LHC, since I see the LHC wrapper in Properties. I shall check as soon as a new task starts.
No, you don't need VirtualBox. But why that wrapper?
ID: 1934391 · Report as offensive
PappaLitto
Volunteer tester

Send message
Joined: 27 Jul 15
Posts: 8
Credit: 1,379,260
RAC: 0
United States
Message 1934459 - Posted: 9 May 2018, 0:25:52 UTC - in response to Message 1934391.  

These WUs don't need virtualbox. There is more need for CPU computation at the moment than GPU so in your GPUGrid account I would disable Long and Short tasks as these are for GPU. If you run one CPU task at a time using 2 threads it will run best for you. The default is for each CPU work unit to use 4 threads, but as you don't have 4 threads it will use 2.
ID: 1934459 · Report as offensive
Profile tullio Project Donor
Volunteer moderator
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 6893
Credit: 2,015,012
RAC: 645
Italy
Message 1934470 - Posted: 9 May 2018, 0:52:28 UTC

CPU tasks all fail, but I am not alone according to the message boards. Long GPU task runs fine on my GTX 750 Ti.
Tullio
ID: 1934470 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 5314
Credit: 354,648,683
RAC: 1,089,909
United States
Message 1934476 - Posted: 9 May 2018, 1:06:33 UTC - in response to Message 1934470.  

CPU tasks run fine . . . as long as you don't let two cpu tasks start within a couple of seconds of each other. There are two separate threads on the site detailing the problem, one in News and the other in Multi-CPU forum. If you do start two task concurrently, they will error out and waste bandwidth. Easiest solution is to set <project_max_concurrent>1</project_max_concurrent> in an app_config or app_info so you would never start two tasks at the same time.

That is the only weakness in the application. They have tons of tasks and not enough volunteers to crunch them.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1934476 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 4659
Credit: 293,483,599
RAC: 465,278
United States
Message 1934480 - Posted: 9 May 2018, 1:11:40 UTC - in response to Message 1934476.  

I posted this over at Einstein since he started a thread there as well.

Only let 1 start at a time, limit it to 4 cores max

<app_config>
  <app>
    <name>QC</name>
    <max_concurrent>1</max_concurrent>
  </app>
  <app_version>
    <app_name>QC</app_name>
    <plan_class>mt</plan_class>
    <avg_ncpus>4</avg_ncpus>
  </app_version>
</app_config>

ID: 1934480 · Report as offensive
juan BFP Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 7078
Credit: 403,079,876
RAC: 276,022
Panama
Message 1934509 - Posted: 9 May 2018, 2:50:29 UTC

The credit gived for the CPU app is very low, compared with the GPU. They need to equilize that.
ID: 1934509 · Report as offensive
Profile tullio Project Donor
Volunteer moderator
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 6893
Credit: 2,015,012
RAC: 645
Italy
Message 1934521 - Posted: 9 May 2018, 5:11:48 UTC

Only GPU tasks complete and validate. CPU tasks fail both on the Opteron 1210 and the AMD E-450, both running SuSE Linux Leap 42.3.
Tullio
ID: 1934521 · Report as offensive
Profile tullio Project Donor
Volunteer moderator
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 6893
Credit: 2,015,012
RAC: 645
Italy
Message 1934556 - Posted: 9 May 2018, 12:39:48 UTC

I gained 96,300 credits on a single GPU task.That is about 28 times what I get at Einstein for a GPU task.
Tullio
ID: 1934556 · Report as offensive
juan BFP Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 7078
Credit: 403,079,876
RAC: 276,022
Panama
Message 1934581 - Posted: 9 May 2018, 13:40:58 UTC - in response to Message 1934556.  
Last modified: 9 May 2018, 13:50:30 UTC

I gained 96,300 credits on a single GPU task.That is about 28 times what I get at Einstein for a GPU task.
Tullio

I not talk about diferent projects, inside the GPUGrid project a single GPU WU could paid more than 150 K.

Anyway is not what i could find for the CPU WU, from the top crunchers an example:

http://www.gpugrid.net/result.php?resultid=17545942

CPU Time 31,335.67
Credit	932.88

Maybe i'm looking on the wrong site. Please correct me if i'm wrong with the link. Thanks
ID: 1934581 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 5314
Credit: 354,648,683
RAC: 1,089,909
United States
Message 1934703 - Posted: 9 May 2018, 20:20:23 UTC - in response to Message 1934581.  
Last modified: 9 May 2018, 20:33:34 UTC

I crunched cpu tasks when they first appeared. But then the credit dropped down to non-existent (6 credits for an hours work). I complained about it and the developers said there was nothing they can do because they use CreditScrew. So I dropped any idea of cpu crunching for GPUGrid.net. I still crunch with gpus when that work is available.

I wonder if that cpu credit example is just from a new cruncher. We all know that CreditScrew gives great credit for first time appearances of a host and then drops credit precipitously after a while due its levelling/anti-cheating mechanism. If that cpu task got that much credit for an established host, them cpu work might need to be looked at again.

[Edit] Either that or the molecule size was much larger than what was used in the first experimental tasks. But I got discouraged when for example I got 111 credits for 3292 seconds of run time while another user got 1361 credits for the exact same 3292 seconds of run time.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1934703 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 4659
Credit: 293,483,599
RAC: 465,278
United States
Message 1934706 - Posted: 9 May 2018, 20:29:41 UTC - in response to Message 1934703.  

The only reason I would do it is if I had extra CPU cores that I'm not using at that time. Like i do fo Einstein, GPUGrid on the cards and any left over cores run Einstein CPU work units.

But, since they don't have Windows CPU apps at GPUGrid, it's a mute point for me...hahaha....
ID: 1934706 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 5314
Credit: 354,648,683
RAC: 1,089,909
United States
Message 1934709 - Posted: 9 May 2018, 20:35:20 UTC - in response to Message 1934706.  

That is in the works. There is an experimental Windows10 cpu application that they have been trying to get debugged. And the developers have stated they are very interested in getting Windows users onboard with cpu crunching.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1934709 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 4659
Credit: 293,483,599
RAC: 465,278
United States
Message 1934713 - Posted: 9 May 2018, 20:44:49 UTC - in response to Message 1934709.  

Saw that.. Will wait and see what happens. As long as the don't affect the GPU work, I might consider doing a few but limit to which CPUs and how many at a time that run.
ID: 1934713 · Report as offensive
Profile tullio Project Donor
Volunteer moderator
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 6893
Credit: 2,015,012
RAC: 645
Italy
Message 1935188 - Posted: 11 May 2018, 17:22:08 UTC
Last modified: 11 May 2018, 17:23:12 UTC

All my CPU tasks have failed on GPUGRID using SuSE Leap 42.3 on 2 AMD CPUs, an Opteron 1210 and a E-450, both with two cores. They are inviting CPU users but the results are not encouraging to spend time and electricity on their tasks.
Tullio
ID: 1935188 · Report as offensive

Message boards : Number crunching : GPUGrid is in need of CPU Volunteers


 
©2018 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.