Strategies for managing mixed GPU WU balance.

Message boards : Number crunching : Strategies for managing mixed GPU WU balance.
Message board moderation

To post messages, you must log in.

AuthorMessage
_
Avatar

Send message
Joined: 15 Nov 12
Posts: 299
Credit: 9,037,618
RAC: 0
United States
Message 1500771 - Posted: 6 Apr 2014, 21:55:33 UTC

Hello Everyone,

Thanks to all the great people on this forum, I was finally able to get my cruncher back and running with an NVIDIA card and an ATI card running side by side.

Over the years, has anyone developed working strategies for maintaining a WU balance between the two types of WUs needed?

For example, I have heard that keeping the "Maintain enough tasks to keep busy for at least..." field in the account preferences very low will help. I've not had any luck in this department however.

If anyone has had any luck in this department, I'd love to know!
ID: 1500771 · Report as offensive
Wedge009
Volunteer tester
Avatar

Send message
Joined: 3 Apr 99
Posts: 451
Credit: 431,396,357
RAC: 553
Australia
Message 1500778 - Posted: 6 Apr 2014, 22:26:39 UTC

The issue stems from the task limits that continue to be in place. Reducing the work cache is the only way that I know of to balance task counts between ATI and NV GPUs. You will need to experiment with what numbers work best for your host: I generally find 1 day is fine for when AstroPulse is being distributed, but when only MultiBeam is available then I will need to drop it to 0.2-0.3, depending on the server's expected work-unit time for your GPUs, plus whether or not it's sending lots of small tasks AKA 'shorties'. In the latter case, I've found I've had to drop the cache to as low as 0.1.
Soli Deo Gloria
ID: 1500778 · Report as offensive
_
Avatar

Send message
Joined: 15 Nov 12
Posts: 299
Credit: 9,037,618
RAC: 0
United States
Message 1500818 - Posted: 7 Apr 2014, 0:42:56 UTC - in response to Message 1500778.  

Thanks very much for your reply! I was trying .5 days, but I will lower it to .2 and see what happens.
ID: 1500818 · Report as offensive
_
Avatar

Send message
Joined: 15 Nov 12
Posts: 299
Credit: 9,037,618
RAC: 0
United States
Message 1501084 - Posted: 7 Apr 2014, 19:04:47 UTC - in response to Message 1500818.  

Does this statistic actually do anything? :) Half joking...

I've set the value to .1, and my machine still keeps a 100 GPU WU cache. (With no ATI WU, to boot!)
ID: 1501084 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 1501097 - Posted: 7 Apr 2014, 19:42:15 UTC - in response to Message 1501084.  

Does this statistic actually do anything? :) Half joking...

I've set the value to .1, and my machine still keeps a 100 GPU WU cache. (With no ATI WU, to boot!)

If you're changing the web preference setting with no effect, it may be because any local preferences override web prefs. For the version of BOINC Manager I'm using, the Advanced menu -> Preferences brings up those local prefs. You can either set the minimum and extra values there, or clear them so the web prefs will be in effect.
                                                                   Joe
ID: 1501097 · Report as offensive
_
Avatar

Send message
Joined: 15 Nov 12
Posts: 299
Credit: 9,037,618
RAC: 0
United States
Message 1501102 - Posted: 7 Apr 2014, 19:47:48 UTC - in response to Message 1501097.  

Does this statistic actually do anything? :) Half joking...

I've set the value to .1, and my machine still keeps a 100 GPU WU cache. (With no ATI WU, to boot!)

If you're changing the web preference setting with no effect, it may be because any local preferences override web prefs. For the version of BOINC Manager I'm using, the Advanced menu -> Preferences brings up those local prefs. You can either set the minimum and extra values there, or clear them so the web prefs will be in effect.
                                                                   Joe


Thanks Joe! I'll be sure to take a look at this when I get home!
ID: 1501102 · Report as offensive
Wedge009
Volunteer tester
Avatar

Send message
Joined: 3 Apr 99
Posts: 451
Credit: 431,396,357
RAC: 553
Australia
Message 1504359 - Posted: 15 Apr 2014, 12:16:18 UTC

It'll also take some time for your system to work through its cache, if it's already hit 100 WUs. Once your cache setting is low enough that less than 100 WUs is enough to satisfy one GPU, then the server will start delivering WUs for your other GPU.
Soli Deo Gloria
ID: 1504359 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34744
Credit: 261,360,520
RAC: 489
Australia
Message 1504361 - Posted: 15 Apr 2014, 12:31:21 UTC
Last modified: 15 Apr 2014, 12:33:06 UTC

I think that you're a bit late with that post Wedge as he is getting work for both cards.

Mherr, you may want to check you hard drive for free space or adjust the amount of disk space that BOINC can use as you're now generating errors with, "SETI@home error -3 Can't write to file -- disk full?", "SETI@home error -1 Can't create file -- disk full?", "(unknown error) - exit code -1073741819 (0xc0000005)" or "couldn't start app: CreateProcess() failed - Insufficient system resources exist to complete the requested service."

[edit] Maybe even check the hard drive for errors.

Cheers.
ID: 1504361 · Report as offensive
Profile BilBg
Volunteer tester
Avatar

Send message
Joined: 27 May 07
Posts: 3720
Credit: 9,385,827
RAC: 0
Bulgaria
Message 1504591 - Posted: 16 Apr 2014, 5:54:58 UTC - in response to Message 1504361.  

Or his antivirus is interfering:
http://boincwiki.mundayweb.com/index.php?title=Add_the_BOINC_Data_directory_to_the_exclusions_of_my_antivirus_program
 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 
ID: 1504591 · Report as offensive
_
Avatar

Send message
Joined: 15 Nov 12
Posts: 299
Credit: 9,037,618
RAC: 0
United States
Message 1505030 - Posted: 17 Apr 2014, 2:15:56 UTC - in response to Message 1504361.  
Last modified: 17 Apr 2014, 2:16:29 UTC

Thanks for your statement Wedge. Thinks are going pretty well with my cahce at the moment once I have it hovering under 100 WUs.

I think that you're a bit late with that post Wedge as he is getting work for both cards.

Mherr, you may want to check you hard drive for free space or adjust the amount of disk space that BOINC can use as you're now generating errors with, "SETI@home error -3 Can't write to file -- disk full?", "SETI@home error -1 Can't create file -- disk full?", "(unknown error) - exit code -1073741819 (0xc0000005)" or "couldn't start app: CreateProcess() failed - Insufficient system resources exist to complete the requested service."

[edit] Maybe even check the hard drive for errors.

Cheers.


Thanks for the advice Wiggo. This happens usually after the computer is running for a few days straight. BOINC has plenty of disk space to use.

Thanks bill for mentioning the antivirus. That could be it, but since it only seems to happen after the computer has been on for a while, you would think that the antivirus would interfere long before that.

One problem I am having however, and I am considering making a thread about it if I can't find any information, is that my NVIDIA WUs are freezing now and then.

I'll come home from work only to find that a NVIDIA WU has been going for 8 hours and hasn't even gotten above 10%. If I suspend that WU, and resume it, it can usually finish up. It seems to happen pretty often. I've checked the computers memory and that doesn't seem to be an issue.
ID: 1505030 · Report as offensive

Message boards : Number crunching : Strategies for managing mixed GPU WU balance.


 
©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.