How to run more than one GPU WU on a Nvidia 650 Ti

Message boards : Number crunching : How to run more than one GPU WU on a Nvidia 650 Ti
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile robl

Send message
Joined: 8 Dec 12
Posts: 47
Credit: 2,587,572
RAC: 0
United States
Message 1436602 - Posted: 1 Nov 2013, 18:58:16 UTC
Last modified: 1 Nov 2013, 19:01:05 UTC

I had been sharing this node (Ubuntu 12.04 with NVIDIA drivers 304.88) with S&H and E&H. Both projects could run 1,2, or 3 GPU tasks in any order/combination. I recently decide to "rearrange" a couple of nodes with respect to project assignments. This is one such node. It now supports Rosetta and S&H.

When it ran S&H and E&H I had a file called /var/lib/boinc-client/projects/setiathome.berkeley.edu/app_config.xml. Its contents looked like the following:

<app_config>
<app>
<name>astropulse_v6</name>
<max_concurrent>6</max_concurrent>
<gpu_versions>
<gpu_usage>0.33</gpu_usage>
<cpu_usage>0.2</cpu_usage>
</gpu_versions>
</app>
</app_config>

I reinstalled this same file but can only get one GPU WU going. It looks like this in boinc-manager:

seti@home some% Running (0.317 CPUs + 1 NVIDIA GPU)

I have looked around and tried different gpu_usage/cpu_usage values but none seem to make a difference. I did restart boinc-client after installing/changing these values. It seems that this file is not being read.

I tried the same app_config.xml file without the concurrent. No difference.
http://www.usefulramblings.org
ID: 1436602 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1436605 - Posted: 1 Nov 2013, 19:04:25 UTC - in response to Message 1436602.  
Last modified: 1 Nov 2013, 19:06:13 UTC

From the wiki:

It is available with 7.0.40+ client versions.

Client configuration

Make sure on the host you're trying to use the app_confg.xml that it has at least that, 7.0.27 is too old.

Claggy
ID: 1436605 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1436608 - Posted: 1 Nov 2013, 19:07:17 UTC - in response to Message 1436602.  

You have a Linux computer running BOIBC v7.0.27

According to the documentation, the minimum version needed to recognise an app_config.xml file is v7.0.40
ID: 1436608 · Report as offensive
Profile robl

Send message
Joined: 8 Dec 12
Posts: 47
Credit: 2,587,572
RAC: 0
United States
Message 1436673 - Posted: 1 Nov 2013, 21:12:37 UTC - in response to Message 1436608.  
Last modified: 1 Nov 2013, 21:13:06 UTC

I have now installed version 7.0.65. No GPU jobs are running.

The event log is showing:

"app astropulse_v6 not found in app_config.xml"

Here is my app_config.xml file's contents:

<app_config>
<app>
<name>astropulse_v6</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.04</cpu_usage>
</gpu_versions>
</app>
</app_config>

I have downloaded one AstroPulse WU but it remains in a "Ready to Start" state.

My app_config.xml file is located at:
/var/lib/boinc-client/projects/setiathome.berkeley.edu
ID: 1436673 · Report as offensive
Profile Fred E.
Volunteer tester

Send message
Joined: 22 Jul 99
Posts: 768
Credit: 24,140,697
RAC: 0
United States
Message 1436687 - Posted: 1 Nov 2013, 21:54:24 UTC

I have now installed version 7.0.65. No GPU jobs are running.

The event log is showing:

"app astropulse_v6 not found in app_config.xml"

Here is my app_config.xml file's contents:

<app_config>
<app>
<name>astropulse_v6</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.04</cpu_usage>
</gpu_versions>
</app>
</app_config>

I have downloaded one AstroPulse WU but it remains in a "Ready to Start" state.

My app_config.xml file is located at:
/var/lib/boinc-client/projects/setiathome.berkeley.edu

The app_config.xml file looks okay, and BOINC read it. The message probably means that BOINC can't find the application mentioned in the app_config.xml file, but the wording is confusing. Looks like you have some abandoned tasks from the upgrade - did you change project directory location? Or pehaps uninstalled before you installed the new version? Something must have zapped the application files.

All of the tasks you have on hand are cpu tasks, so the message is not a problem. The project will resend the application files when it sends the first gpu task. The AP you have is also a cpu task, so it's waiting its turn to run.

New tasks are hard to get this afternoon, so it may take a while to get the gpu tasks. You will probably want to add a section for S@H v7 gpu tasks, but maybe one thing at a time is best.
Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.
ID: 1436687 · Report as offensive
Profile robl

Send message
Joined: 8 Dec 12
Posts: 47
Credit: 2,587,572
RAC: 0
United States
Message 1436987 - Posted: 2 Nov 2013, 13:06:59 UTC - in response to Message 1436687.  

Thanks for the input all. I changed my app_config.xml file to look like this:

<app_config>
<app>
<name>astropulse_v6</name>
<max_concurrent>3</max_concurrent>
<gpu_versions>
<gpu_usage>0.33</gpu_usage>
<cpu_usage>0.2</cpu_usage>
</gpu_versions>
</app>
</app_config>

I now have 3 GPU WUs crunching away. The main confusion was over the version of boinc/boinc-manager. Also the "max_concurrent" attribute seems required to run more than 1 GPU.

FYI for newbies: If you change the app_config.xml file you need to get BOINC to re-read the file. Do this by selecting "advanced" ---> "read config file".
No need to restart boinc-client.
ID: 1436987 · Report as offensive
Profile Fred E.
Volunteer tester

Send message
Joined: 22 Jul 99
Posts: 768
Credit: 24,140,697
RAC: 0
United States
Message 1436994 - Posted: 2 Nov 2013, 13:46:13 UTC

I now have 3 GPU WUs crunching away. The main confusion was over the version of boinc/boinc-manager. Also the "max_concurrent" attribute seems required to run more than 1 GPU.

No, the max concurrent line is not required. The previous app_info.xml would have worked once you got some gpu tasks. The count value determines the number of tasks from that application to run on the gpu. The max concurrent value determines the total number of tasks with the same short name that can run at the same time on the cpu and gpu, so this value of 3 won't allow any AP cpu tasks to run while 3 AP gpu tasks are running. You could increase the max concurrent to 4 to allow 1 AP cpu task to also run while 3 gpu tasks are running. Otherwise they will run while a S@H v7 task is running on the cpu.

If you want to run more than one S@H v7 task at a time, add a section for it. If you want to use app_info.xml for your other BOINC projects, you'll need to make one in that project's data directory. I use the max concurrent setting even if it is a cpu only project to help enforce my resource share setting.
Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.
ID: 1436994 · Report as offensive
Profile robl

Send message
Joined: 8 Dec 12
Posts: 47
Credit: 2,587,572
RAC: 0
United States
Message 1436999 - Posted: 2 Nov 2013, 14:33:12 UTC - in response to Message 1436994.  

Fred,

Thanks for explaining the "max_concurrent" attribute. I did remove it. I had an AP cpu job in a wait state and it never ran when cpus became available. Now I know why.
ID: 1436999 · Report as offensive
billy ewell 1931 Project Donor
Volunteer tester

Send message
Joined: 1 Apr 03
Posts: 23
Credit: 24,295,322
RAC: 2
United States
Message 1450015 - Posted: 4 Dec 2013, 21:07:12 UTC - in response to Message 1436994.  
Last modified: 4 Dec 2013, 21:32:39 UTC

Fred: This is the app I constructed to run multiple SETI tasks on my i7 and GTX 550ti. I have tried various combinations of "name" within the app and different file names in which to place the app and I cannot get more than one task to run on the gpu. I have been running POEM on three different gpus for the past month using the name "poemcl" on the beginning name line and "poempp" at the ending name line and placed the app_config.xml as follows: C;\programdata\BOINC\projects\boinc.fzk.de_poem. With this configuration I was able to run as many as 10 tasks concurrently. Some guidance please as this 82 year old brain is not that of a rocket scientist. Bill

PS I don't know why I am listed as a volunteer tester because I have serious doubts as to my being qualified.

<app_config>
<app>
<name>setiathome_enhanced</name>
<gpu_versions>
<gpu_usage>0.25</gpu_usage>
<cpu_usage>0.5</cpu_usage>
</gpu_versions>
</app>
<app>
<name>setiathome_enhanced</name>
<max_concurrent>10</max_concurrent>
</app>
</app_config>
ID: 1450015 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1450028 - Posted: 4 Dec 2013, 21:43:46 UTC - in response to Message 1450015.  

Fred: This is the app I constructed to run multiple SETI tasks on my i7 and GTX 550ti. I have tried various combinations of "name" within the app and different file names in which to place the app and I cannot get more than one task to run on the gpu.

See this thread for the number of setiathome_enhanced workunits left to complete:

http://setiathome.berkeley.edu/forum_thread.php?id=72383

The type of work that replaced setiathome_enhanced is called setiathome_v7

Here is a suitable app_config.xml:

<app_config>
 <app>
 <name>astropulse_v6</name>
 <max_concurrent>2</max_concurrent>
 <gpu_versions>
 <gpu_usage>0.5</gpu_usage>
 <cpu_usage>0.05</cpu_usage>
 </gpu_versions>
 </app>
 <app>
 <name>setiathome_v7</name>
 <max_concurrent>10</max_concurrent>
 <gpu_versions>
 <gpu_usage>0.25</gpu_usage>
 <cpu_usage>0.05</cpu_usage>
 </gpu_versions>
 </app>
</app_config>


Claggy
ID: 1450028 · Report as offensive
Profile Fred E.
Volunteer tester

Send message
Joined: 22 Jul 99
Posts: 768
Credit: 24,140,697
RAC: 0
United States
Message 1450044 - Posted: 4 Dec 2013, 22:05:41 UTC

Claggy beat me as usual. Would suggest a count of .5 for a 550ti, the .25 you're using is too aggressive to start with.
Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.
ID: 1450044 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22158
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1450047 - Posted: 4 Dec 2013, 22:10:45 UTC

PS I don't know why I am listed as a volunteer tester because I have serious doubts as to my being qualified.


You have done some work for SETI@home-Beta and this gives you the "volunteer tester" status.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1450047 · Report as offensive
billy ewell 1931 Project Donor
Volunteer tester

Send message
Joined: 1 Apr 03
Posts: 23
Credit: 24,295,322
RAC: 2
United States
Message 1450180 - Posted: 5 Dec 2013, 5:24:26 UTC

Thanks for the excellent advice and input that now has my 550ti crunching with a load factor of 91-93% at a temp of 64 degrees.

Bill: AKA 82YOVT
ID: 1450180 · Report as offensive

Message boards : Number crunching : How to run more than one GPU WU on a Nvidia 650 Ti


 
©2024 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.