Advanced search

Message boards : Graphics cards (GPUs) : Problems with Boinc 6.6.2

Author Message
rapt0r
Send message
Joined: 4 Sep 08
Posts: 16
Credit: 9,366,617
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 5812 - Posted: 20 Jan 2009 | 10:35:18 UTC

Currently one Wu is running. And I want to have a second one. CPU-Usage is at 100 percent for my Dual-Core.

20.01.2009 11:30:12|GPUGRID|Sending scheduler request: Requested by user.
20.01.2009 11:30:12|GPUGRID|CPU work request: 885600.00 seconds, 0 instances
20.01.2009 11:30:12|GPUGRID|CUDA work request: 738000.00 seconds, 0 instances
20.01.2009 11:30:17|GPUGRID|Scheduler request completed: got 0 new tasks
20.01.2009 11:30:17|GPUGRID|Message from server: No work sent
20.01.2009 11:30:17|GPUGRID|Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
20.01.2009 11:30:17|GPUGRID|Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.
20.01.2009 11:30:17|GPUGRID|Message from server: (won't finish in time) BOINC runs 99.9% of time, computation enabled 99.1% of that

MarkJ
Volunteer moderator
Volunteer tester
Send message
Joined: 24 Dec 08
Posts: 738
Credit: 200,909,904
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5813 - Posted: 20 Jan 2009 | 10:46:27 UTC - in response to Message 5812.

Currently one Wu is running. And I want to have a second one. CPU-Usage is at 100 percent for my Dual-Core.

20.01.2009 11:30:12|GPUGRID|Sending scheduler request: Requested by user.
20.01.2009 11:30:12|GPUGRID|CPU work request: 885600.00 seconds, 0 instances
20.01.2009 11:30:12|GPUGRID|CUDA work request: 738000.00 seconds, 0 instances
20.01.2009 11:30:17|GPUGRID|Scheduler request completed: got 0 new tasks
20.01.2009 11:30:17|GPUGRID|Message from server: No work sent
20.01.2009 11:30:17|GPUGRID|Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
20.01.2009 11:30:17|GPUGRID|Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.
20.01.2009 11:30:17|GPUGRID|Message from server: (won't finish in time) BOINC runs 99.9% of time, computation enabled 99.1% of that


The last line would suggest that it doesn't think you'll be able to finish it in time. Do you have a slow graphics card and/or a big cache setting?
____________
BOINC blog

rapt0r
Send message
Joined: 4 Sep 08
Posts: 16
Credit: 9,366,617
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 5814 - Posted: 20 Jan 2009 | 10:59:00 UTC - in response to Message 5813.
Last modified: 20 Jan 2009 | 11:01:07 UTC

I use a 8800GTS512@760/1900/100. Fast enough. So what do you mean with big cache setting?

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5819 - Posted: 20 Jan 2009 | 11:41:12 UTC - in response to Message 5814.

I use a 8800GTS512@760/1900/100. Fast enough. So what do you mean with big cache setting?


How big is your work cache setting?

Your work buffer settings as set in preferences. Resource share can also play a part because BOINC Manager may not be correctly distinguishing between GPU resource allocations and CPU allocations.

This is supposed to be one of the fixes in the 6.6.x series, but, it is a work in progress. To this point, I think most of us have settled in on 6.5.0 as being the best compromise ... at least, I have seen no major complaints and it is the version I am using ... You may have to delve into the archive to get this version as it was not one of the recommended versions.
____________

rapt0r
Send message
Joined: 4 Sep 08
Posts: 16
Credit: 9,366,617
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 5821 - Posted: 20 Jan 2009 | 12:33:26 UTC - in response to Message 5819.
Last modified: 20 Jan 2009 | 12:33:52 UTC

Maintain enough work for an additional
Enforced by version 5.10+ 10 days
Resourceshare 500

http://www.gpugrid.net/show_host_detail.php?hostid=8106

The CPU ist a Opteron 165 with 2.4 GHz each Core.

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5825 - Posted: 20 Jan 2009 | 19:05:03 UTC

Is the projected run time of your one GPU-WU way off?

MrS
____________
Scanning for our furry friends since Jan 2002

rapt0r
Send message
Joined: 4 Sep 08
Posts: 16
Credit: 9,366,617
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 5827 - Posted: 20 Jan 2009 | 19:20:46 UTC - in response to Message 5825.
Last modified: 20 Jan 2009 | 19:21:42 UTC

No, it takes 6.8 Hours to finish it. While 6 Hours are elapsed to reach 55 %.

MarkJ
Volunteer moderator
Volunteer tester
Send message
Joined: 24 Dec 08
Posts: 738
Credit: 200,909,904
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5864 - Posted: 22 Jan 2009 | 10:31:48 UTC - in response to Message 5821.

Maintain enough work for an additional
Enforced by version 5.10+ 10 days
Resourceshare 500

http://www.gpugrid.net/show_host_detail.php?hostid=8106

The CPU ist a Opteron 165 with 2.4 GHz each Core.


Under previous incarnations of BOINC it used to add the cache size (10 days in your case) and check the deadline, before downloading the wu. As GPUgrid has 4 day deadlines that means unless you have a smaller number in the cache setting (say 3 days) it won't think its able to finish in time.

I said "previous incarnation" because they changed the whole work-fetch stuff in 6.6.2 so it may not work that way any more.
____________
BOINC blog

Rabinovitch
Avatar
Send message
Joined: 25 Aug 08
Posts: 143
Credit: 64,937,578
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 5938 - Posted: 24 Jan 2009 | 6:45:28 UTC

http://boinc.berkeley.edu/dev/forum_thread.php?id=3542&nowrap=true#22631

I've just installed 6.6.2. Can't get new work at all with this new version:

24.01.2009 12:46:48|lhcathome|Sending scheduler request: To fetch work.
24.01.2009 12:46:48|lhcathome|CPU work request: 0.00 seconds, 0 instances
24.01.2009 12:46:48|lhcathome|CUDA work request: 398769.23 seconds, 1 instances
24.01.2009 12:46:53|lhcathome|Scheduler request completed: got 0 new tasks
24.01.2009 12:46:58|ralph@home|Sending scheduler request: To fetch work.
24.01.2009 12:46:58|ralph@home|CPU work request: 0.00 seconds, 0 instances
24.01.2009 12:46:58|ralph@home|CUDA work request: 518400.00 seconds, 1 instances
24.01.2009 12:47:03|ralph@home|Scheduler request completed: got 0 new tasks
24.01.2009 12:48:13|rosetta@home|Sending scheduler request: To fetch work.
24.01.2009 12:48:13|rosetta@home|CPU work request: 0.00 seconds, 0 instances
24.01.2009 12:48:13|rosetta@home|CUDA work request: 518400.00 seconds, 1 instances
24.01.2009 12:48:18|rosetta@home|Scheduler request completed: got 0 new tasks

How to do?

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5939 - Posted: 24 Jan 2009 | 7:25:48 UTC - in response to Message 5938.
Last modified: 24 Jan 2009 | 7:26:38 UTC

My first thought is to suggest to revert to 6.5.0 which I have had the best luck with ... you should be able to access it thiough the archive link on the DL page.

I have only one instance of 6.6.2 BOINC Manager running and I am not sure that it is requesting for me correctly either. The trouble is that I have at least one additional task in the queue and so it looks like I really need to wait till it is in work or done before I can be sure I am having the problem.

The thing is that the server messages have been changed and I am not at all sure that the change is backward compatible for CUDA projects ... in other words, GPU Grid might not be really set up to respond correctly to 6.6.2 ...

{diit}

If you have no work, you could also try project reset ... but ONLY if you have no work in progress or you will lose that too ...
____________

Rabinovitch
Avatar
Send message
Joined: 25 Aug 08
Posts: 143
Credit: 64,937,578
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 5941 - Posted: 24 Jan 2009 | 7:51:41 UTC

I just rolled back to 6.4.5. Now I am not in good feel to something else. If it doesn't work, I don't need it.

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 5942 - Posted: 24 Jan 2009 | 9:20:54 UTC - in response to Message 5941.

newer version of the client like 6.6.2 are probably not 100% compatible with the server version of GPUGRID. We will do a project update next week. I hope that it will be smooth for users with older clients.

gdf

MarkJ
Volunteer moderator
Volunteer tester
Send message
Joined: 24 Dec 08
Posts: 738
Credit: 200,909,904
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5947 - Posted: 24 Jan 2009 | 10:12:44 UTC - in response to Message 5942.

newer version of the client like 6.6.2 are probably not 100% compatible with the server version of GPUGRID. We will do a project update next week. I hope that it will be smooth for users with older clients.

gdf


I'd probably hold off a bit on that. It would seem the server side changes at seti are causing too much work fetch, even when set to NNW in the client. It does this to both new and old BOINC clients (reports of 5.10.45, 6.2.19 and 6.6.2).
____________
BOINC blog

rapt0r
Send message
Joined: 4 Sep 08
Posts: 16
Credit: 9,366,617
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 5950 - Posted: 24 Jan 2009 | 11:39:46 UTC
Last modified: 24 Jan 2009 | 11:42:09 UTC

Now it's running good. I removed everything via deinstall and manual deletion of directories. Before I tried the 6.5.0, where I had to pause the other projects to getting work for GPU-Grid. The Boinc Beta-Versions are the biggest zone of error, obviously. Since 2 month of manual updating, which is very annoying, the project runs fine on my 2 pc's, now. If I had more pc's or faster systems, it would be quite safe to become insane in this time.

Post to thread

Message boards : Graphics cards (GPUs) : Problems with Boinc 6.6.2

//