Are WU cuda specific?

Questions and Answers : GPU applications : Are WU cuda specific?
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Searcher
Avatar

Send message
Joined: 26 Jun 99
Posts: 139
Credit: 9,063,168
RAC: 15
United States
Message 891224 - Posted: 4 May 2009, 16:44:28 UTC

Are the WU that download specific to the GPU?
I had all 4 processors and the cuda process running.
I set the down load to not download any more wu last week and everything was busy still.
Today I check and the gpu is running but the cpus are idle with 82 wu still in the queue.
So this makes me think that the wu are different eg only for the gpu.

-searcher
ID: 891224 · Report as offensive
OzzFan Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Apr 02
Posts: 15691
Credit: 84,761,841
RAC: 28
United States
Message 891258 - Posted: 4 May 2009, 19:05:11 UTC

When your computer requests more tasks from the server, it downloads them onto your sysstem and automatically links them to an application to be processed. CUDA WUs are techinically the same SETI@Home MultiBeam workunits used on the CPU, but have been linked with the CUDA app instead of the CPU app. It is possible, but very difficult, to change the association from the CUDA app to the GPU app and vice versa.

So its not technically the WU that's specific, but the app that it has been linked to that is specific.
ID: 891258 · Report as offensive
Profile Searcher
Avatar

Send message
Joined: 26 Jun 99
Posts: 139
Credit: 9,063,168
RAC: 15
United States
Message 891364 - Posted: 5 May 2009, 0:08:56 UTC

Thanks! Seems odd that is it not just another cpu. Sort of like switching seti version while you have some processing. The execution app is inside the slots.

Thanks again.

-searcher
ID: 891364 · Report as offensive
PHoganDive

Send message
Joined: 18 May 99
Posts: 3
Credit: 55,625,949
RAC: 0
United States
Message 900413 - Posted: 28 May 2009, 9:40:48 UTC

I have a question along similar lines...I have an Intel quad core, and an nVidia 260. The SETI program is set up to use 5 CPUs. It downloads two types of files, ver 6.03 for the Intel CPU, and ver 6.08 for the GPU. 4 CPUs run the 6.03 files just fine, however the 5th CPU that runs seems to be totally random, sometimes it will correctly run a 6.08 file on the GPU, but frequently it runs a 5th instance of a 6.03 file, and spends a lot of time swapping tasks running 5 WUs on the 4 Intel cores, while the GPU sits idle. Is there any fix for this? I keep hoping the next new version will fix it, but no joy so far, running latest ver 6.6.28.

Thanks,
Peter
phogandive@aol.com
ID: 900413 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 900421 - Posted: 28 May 2009, 10:12:45 UTC - in response to Message 900413.  

Is there any fix for this? I keep hoping the next new version will fix it, but no joy so far, running latest ver 6.6.28.

You added <ncpus>5</ncpus> to a cc_config.xml file in your BOINC Data directory. Take out that line from the cc_config.xml or if it's the only line in there that's in use, remove the cc_config.xml file and restart BOINC.

BOINC doesn't come with a cc_config.xml file. You must have added this yourself, probably for the 6.4.5 to 6.4.7 versions which had no internal GPU scheduler. Since BOINC 6.6.20 and 6.6.28 do have a GPU scheduler, you no longer need the line in the cc_config.xml file.
ID: 900421 · Report as offensive
Zebra3
Avatar

Send message
Joined: 22 Oct 01
Posts: 186
Credit: 13,658,148
RAC: 0
Canada
Message 900574 - Posted: 28 May 2009, 17:45:52 UTC - in response to Message 900421.  

Is there any fix for this? I keep hoping the next new version will fix it, but no joy so far, running latest ver 6.6.28.

You added <ncpus>5</ncpus> to a cc_config.xml file in your BOINC Data directory. Take out that line from the cc_config.xml or if it's the only line in there that's in use, remove the cc_config.xml file and restart BOINC.

BOINC doesn't come with a cc_config.xml file. You must have added this yourself, probably for the 6.4.5 to 6.4.7 versions which had no internal GPU scheduler. Since BOINC 6.6.20 and 6.6.28 do have a GPU scheduler, you no longer need the line in the cc_config.xml file.


Sooo....I have installed a moderate 9600GSO card in my quad that I was given and have also installed the current drivers. I have also installed the BETA unified installer. I find that I am getting very few CUDA 6.08 WU's. but tons of 6.03's Is it a situation that there are very few CUDA's to be found? I do find that the CUDA WU's that have been validated take about 28 minutes to process...does this seem to be about correct for this card?

thanks in advance...
http://www.novascotia.com
ID: 900574 · Report as offensive
Profile Westsail and *Pyxey*
Volunteer tester
Avatar

Send message
Joined: 26 Jul 99
Posts: 338
Credit: 20,544,999
RAC: 0
United States
Message 900845 - Posted: 29 May 2009, 5:13:20 UTC

I think that Raismer's GPU>CPU rebranding project is a wonderful idea. The premise being CUDA cards have quite a time with the VLAR tasks...so, the script will detect and rebrand those as CPU units. This way one host can handle all kinds of work units with maximum efficiency. Very promising indeed. Thanks Raistmer!

Lunatics CPU <-> GPU rebranding perl script
"The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov
ID: 900845 · Report as offensive
Profile Westsail and *Pyxey*
Volunteer tester
Avatar

Send message
Joined: 26 Jul 99
Posts: 338
Credit: 20,544,999
RAC: 0
United States
Message 900847 - Posted: 29 May 2009, 5:18:17 UTC - in response to Message 900574.  


Sooo....I have installed a moderate 9600GSO card in my quad that I was given and have also installed the current drivers. I have also installed the BETA unified installer. I find that I am getting very few CUDA 6.08 WU's. but tons of 6.03's Is it a situation that there are very few CUDA's to be found? I do find that the CUDA WU's that have been validated take about 28 minutes to process...does this seem to be about correct for this card?
thanks in advance...


What version Boinc you have? 6.6.28 should fetch and brand things properly.?. If you are getting MB CPU work they are the same units that CUDA use. Is card starving or there just not as many in list? As long as card not starving, maybe Boinc will sort it out in time.
Also, that is correct processing time for that card.
Good luck and keep us posted!
"The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov
ID: 900847 · Report as offensive
Zebra3
Avatar

Send message
Joined: 22 Oct 01
Posts: 186
Credit: 13,658,148
RAC: 0
Canada
Message 900914 - Posted: 29 May 2009, 11:56:09 UTC - in response to Message 900847.  


Sooo....I have installed a moderate 9600GSO card in my quad that I was given and have also installed the current drivers. I have also installed the BETA unified installer. I find that I am getting very few CUDA 6.08 WU's. but tons of 6.03's Is it a situation that there are very few CUDA's to be found? I do find that the CUDA WU's that have been validated take about 28 minutes to process...does this seem to be about correct for this card?
thanks in advance...


What version Boinc you have? 6.6.28 should fetch and brand things properly.?. If you are getting MB CPU work they are the same units that CUDA use. Is card starving or there just not as many in list? As long as card not starving, maybe Boinc will sort it out in time.
Also, that is correct processing time for that card.
Good luck and keep us posted!


Thanks for the response...my CUDA Wu's have arrived in hordes...I guess it was just Seti catching up...all is well again...cheers
http://www.novascotia.com
ID: 900914 · Report as offensive
Zebra3
Avatar

Send message
Joined: 22 Oct 01
Posts: 186
Credit: 13,658,148
RAC: 0
Canada
Message 902875 - Posted: 2 Jun 2009, 12:17:54 UTC
Last modified: 2 Jun 2009, 12:18:35 UTC

I have also noticed that since I installed this card...see post below...that my task duration correction factor has almost doubled to 0.142923 now where it used to be a steady .067000 or so and also my non CUDA WU's are taking longer to complete.

Is this due to the card using more processor time of which in turn is slowing down the processing of my regular AP's and MB WU's?

Just another burning question I have...lol.
http://www.novascotia.com
ID: 902875 · Report as offensive

Questions and Answers : GPU applications : Are WU cuda specific?


 
©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.