Advanced search

Message boards : Number crunching : Question about expired w/u

Author Message
The Jedi Alliance - Range...
Send message
Joined: 9 Oct 10
Posts: 4
Credit: 8,282,509
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwat
Message 26501 - Posted: 28 Jul 2012 | 21:32:27 UTC

I tried searching for an answer to the question I'm about to ask and found nothing.

I have work units running that have already expired. I have work units running that will not finish in time if I can believe the estimated time remaining, which I don't because these w/u never seem to finish as estimated.

My question is will these w/u be used by the scientists once they do finish? This is not a question about BOINC credit, it's a question about usefullness of the results to the scientist.

If the w/u will be used then I will allow them to finish, otherwise I will abort them. For some time my computers, which have not been changed in any way, were turning out 1 or more w/u per day. Recently the w/u have been taking 2 to 10 days each. These w/u are downloaded showing a deadline only 2 days in the future and therefore the probability of completion by the deadline is very low.

Snow Crash
Send message
Joined: 4 Apr 09
Posts: 450
Credit: 539,316,349
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 26503 - Posted: 28 Jul 2012 | 22:45:06 UTC - in response to Message 26501.

Hi ranger,

If they are not going to finish in time I would suggest aborting them as replacements will be sent out before the deadline is reached. This is by design so that the project gets the results it needs in a timely manner (each WU is built from the results of a previous WU so we need to turn them around quickly).

As to why your card may not be returning within 1 day as it had been previously, I can't say as you have not given us any infomration about it or the OS you are running. If you post that info I'm sure someone will be able to help you get things running back to normal again.
____________
Thanks - Steve

The Jedi Alliance - Range...
Send message
Joined: 9 Oct 10
Posts: 4
Credit: 8,282,509
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwat
Message 26507 - Posted: 29 Jul 2012 | 0:41:18 UTC - in response to Message 26503.

Thanks, I'll abort any that will not finish in time. It's a shame to put so many hours effort into these w/u only to abort.

I have two computers identical except for NVIDIA GPU:

AuthenticAMD
AMD Athlon(tm) Dual Core Processor 4450e [Family 15 Model 107 Stepping 2]
(2 processors)

Microsoft Windows Vista
Home Premium x86 Edition, Service Pack 2,

1st computer: NVIDIA GeForce GTX 550 Ti (1023MB) driver: 301.42

2nd computer: NVIDIA GeForce GT 240 (1024MB) driver: 301.42

Both are running 24 hours a day. They are set to run always and I haven't used either one for a week (way too much time at the office lately). The 550 was completing w/u in 3-4 hours not too long ago and the 240 was completing in 8-10 hours. I haven't been around to make any changes on the computers and the only thing I see different is the w/u. No OC. Both running between 98-100% GPU. Temp and fan speed are fine. CPU running about 2-4%, memory about 50% and I don't have any of the original pre-loaded crap on the machines.

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 26510 - Posted: 29 Jul 2012 | 3:33:43 UTC

I have heard reports of the 301.42 driver down clocking prematurely. They have some overheating protection and it triggers too early. Maybe check your card temps to see if that may be the issue. You need to reboot the machine to get it back to normal.

Obvious suggestions are to try and keep card temps down, probably by cranking up the fan speed or maybe try updating to a later driver. Better still if you can improve machine ventilation (extra fans, run with the side off, etc).
____________
BOINC blog

The Jedi Alliance - Range...
Send message
Joined: 9 Oct 10
Posts: 4
Credit: 8,282,509
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwat
Message 26514 - Posted: 29 Jul 2012 | 17:02:41 UTC - in response to Message 26510.

Thanks Mark. I went back to the 295.73 driver and it appears that my 550 card will be completing 2 w/u a day now rather than 3 days a w/u. The 240 card is looking like 3 days a w/u but looks like it will complete before the deadline.

Both cards are running warmer than they did with the 301.42 driver but temp is acceptable. I'll be looking into some alternate cooling for the box.

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 26516 - Posted: 30 Jul 2012 | 7:36:30 UTC - in response to Message 26514.

Thanks Mark. I went back to the 295.73 driver and it appears that my 550 card will be completing 2 w/u a day now rather than 3 days a w/u. The 240 card is looking like 3 days a w/u but looks like it will complete before the deadline.

Both cards are running warmer than they did with the 301.42 driver but temp is acceptable. I'll be looking into some alternate cooling for the box.


I thought the 295.x and 296.x drivers had the monitor sleep bug. Make sure power management doesn't try and power off the screen or you will have lots of failed work. I would stay clear of these driver versions, in fact Einstein blocks work for people with those drivers as they cause so much trouble.
____________
BOINC blog

The Jedi Alliance - Range...
Send message
Joined: 9 Oct 10
Posts: 4
Credit: 8,282,509
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwat
Message 26524 - Posted: 31 Jul 2012 | 13:41:01 UTC - in response to Message 26516.

Thanks Mark. I went back to the 295.73 driver and it appears that my 550 card will be completing 2 w/u a day now rather than 3 days a w/u. The 240 card is looking like 3 days a w/u but looks like it will complete before the deadline.

Both cards are running warmer than they did with the 301.42 driver but temp is acceptable. I'll be looking into some alternate cooling for the box.


I thought the 295.x and 296.x drivers had the monitor sleep bug. Make sure power management doesn't try and power off the screen or you will have lots of failed work. I would stay clear of these driver versions, in fact Einstein blocks work for people with those drivers as they cause so much trouble.


The 295.73 driver is working quite well on both machines. W/U's are completing in slightly under the estimated time and no failed work. I guess I could search for an older driver, I sure don't want to ever go back to the 301. What I find interesting is that my problems started after July 25th and I installed the 301 driver in May. Seti was still running fine with the 301 driver, it was only work from here with an issue (of course I'm not running other projects and don't know how they would have been affected).

I want to thank you for your help. I run this project to help the scientists and I wasn't doing that and now I am again.

Post to thread

Message boards : Number crunching : Question about expired w/u

//