Advanced search

Message boards : News : Resuming long WU supply: ABxx

Author Message
Toni
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 9 Dec 08
Posts: 602
Credit: 4,273,184
RAC: 0
Level
Ala
Scientific publications
watwatwatwat
Message 21057 - Posted: 25 Apr 2011 | 15:36:03 UTC
Last modified: 25 Apr 2011 | 15:47:43 UTC

I am also resuming the supply on the long queue. They are related to amyloid proteins, involved in diseases including Alzheimer's. The new tasks took 11h on a test GTX 275.

Make sure you disable the "acemdlong" application from you GPUGRID preferences if you have an older card.

Happy Easter!

Profile robertmiles
Send message
Joined: 16 Apr 09
Posts: 402
Credit: 170,432,446
RAC: 275,714
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21145 - Posted: 2 May 2011 | 23:46:50 UTC
Last modified: 2 May 2011 | 23:49:41 UTC

Does a GTS 450 count as an older card? My computer usually finishes the long workunits in about 24 hours, but then waits for me to return to the keyboard and tell it to report the latest finished result.

I've already sent the BOINC developers a few suggestions on adjusting the timing for GPU workunits - for example, allow the queue for GPU workunits to be set much shorter than for CPU workunits. Are the GPUGRID developers interested in finding those suggestions and expressing interest in following them, since this should allow them to get reported results a little faster?

Also, are the GPUGRID developers interested in seeing web links to articles related to amyloid proteins related to at least one form of diabetes?

Profile skgiven
Volunteer moderator
Project tester
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,837,071,099
RAC: 365,113
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21148 - Posted: 3 May 2011 | 13:34:29 UTC - in response to Message 21145.

Robert, use report tasks immediately in cc_config.xml.

How to in the FAQ: Best configurations for GPUGRID

As long as it reports within 24h it's fast enough to run long tasks and get full credit. It might also be fine if it finishes between 1day and 2days, but after that it's not recommended. If tasks start failing just crunch the short tasks.

Toni
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 9 Dec 08
Posts: 602
Credit: 4,273,184
RAC: 0
Level
Ala
Scientific publications
watwatwatwat
Message 21149 - Posted: 3 May 2011 | 15:08:41 UTC - in response to Message 21148.

Hi Robert, I second skgiven's suggestion.

Protein fibrillation is a pretty ubiquitous phenomenon (wikipedia has a nice summary list). Of course you are welcome to post links to scientific papers (but there are chances we are already aware of them :-)

Profile robertmiles
Send message
Joined: 16 Apr 09
Posts: 402
Credit: 170,432,446
RAC: 275,714
Level
Ile
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21150 - Posted: 3 May 2011 | 16:21:20 UTC - in response to Message 21148.
Last modified: 3 May 2011 | 16:50:33 UTC

Is a variation of that available to apply it to GPUGRID tasks, but not to tasks from certain other BOINC projects with server capacity limited enough that they want tasks reported only several at a time?

On posting links: Even if you're already aware of them, many of the participants here aren't, and would like to know more about what conditions your research could help.

Profile skgiven
Volunteer moderator
Project tester
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,837,071,099
RAC: 365,113
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 21151 - Posted: 3 May 2011 | 19:05:37 UTC - in response to Message 21150.

I am not aware of a way to report GPUGrid tasks but not CPU tasks; it can be done for all tasks only using cc_config, but I don't think it can be done using app_info (project specific). This is something for the Boinc developers.

The only way you could do this on a project only basis would be to use a virtual system on which you have CPU projects, and do a separate installation of Boinc on the virtual platform - far too much for most crunchers.

In the future (possibly this year) Boinc might better facilitate GPU crunching. Other projects are looking into using GPU's. Perhaps the emphasis will move slightly from CPU to GPU.

Post to thread

Message boards : News : Resuming long WU supply: ABxx