Advanced search

Message boards : Graphics cards (GPUs) : Windows 6.55 CPU Usage

Author Message
Profile The Gas Giant
Avatar
Send message
Joined: 20 Sep 08
Posts: 54
Credit: 607,157
RAC: 0
Level
Gly
Scientific publications
watwatwatwat
Message 5396 - Posted: 8 Jan 2009 | 20:28:45 UTC

Any chance of getting the CPU usage for windows back to something reasonable? Having it run at 80% of a core on my Q9450 - while having it say it only needs 0.03 CPU isn't good.

Live long and BOINC!

Paul.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5405 - Posted: 9 Jan 2009 | 1:31:16 UTC - in response to Message 5396.

See this thread:

http://www.gpugrid.net/forum_thread.php?id=605

Rabinovitch
Avatar
Send message
Joined: 25 Aug 08
Posts: 143
Credit: 64,937,578
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 5414 - Posted: 9 Jan 2009 | 8:11:46 UTC

Which version of BOINC do you use? After I've updated it to 6.5.0 CPU usage became mor or less normal - between 13% and 20%. WinXP x64.

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5419 - Posted: 9 Jan 2009 | 12:45:30 UTC - in response to Message 5414.

Which version of BOINC do you use? After I've updated it to 6.5.0 CPU usage became mor or less normal - between 13% and 20%. WinXP x64.


ON a 4 core system 20% CPU usage is 80% of a core ... Windows reports the percentage on a system basis so that full core usage is 25% on a 4 core and 12% on an 8 CPU system ...

GPU Grid is over using core and the tradeoff is to run it while waiting for the improved version that lowers core usage thus doing work for the project (and earning credit) or to suspend GPU Grid for the moment and have more CPU for other projects ...

My choice for the moment is to "eat" the loss and run the project (as I am doing for FreeHAL which also seems to be using a little more core than I would like) in the interests of progress ...

Every one has to make the choice and there is no single right choice here ...

I mean, what do I do when Nvidia has CUDA running on the mac and GPU Grid does not have an application ready ... do I start running Folding at home? :)

Rabinovitch
Avatar
Send message
Joined: 25 Aug 08
Posts: 143
Credit: 64,937,578
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 5421 - Posted: 9 Jan 2009 | 14:39:22 UTC - in response to Message 5419.



ON a 4 core system 20% CPU usage is 80% of a core ... Windows reports the percentage on a system basis so that full core usage is 25% on a 4 core and 12% on an 8 CPU system ...



You mean that author is processing 4 gpugrid WUs at a time? Hmm...

Rabinovitch
Avatar
Send message
Joined: 25 Aug 08
Posts: 143
Credit: 64,937,578
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 5422 - Posted: 9 Jan 2009 | 14:45:18 UTC

By the way, it could be sweet if gpugrid crunching would not disturb with graphics "freesing". Even on not so slow GPU - on my gtx260 - I don't feel good when I see my desktop and KNOW, that GPU is busy with WU now... And it's without gaming, only during "starfield" screensaver!

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5423 - Posted: 9 Jan 2009 | 16:34:45 UTC - in response to Message 5421.



ON a 4 core system 20% CPU usage is 80% of a core ... Windows reports the percentage on a system basis so that full core usage is 25% on a 4 core and 12% on an 8 CPU system ...



You mean that author is processing 4 gpugrid WUs at a time? Hmm...


NO ...

I mean that if you see a usage number of 20% on a 4 core system, that is 80% of that core's capability. On a 4 core system, 25% is full utilization of a single CPU core. On an 8 core system it is roughly 12% with a couple percent left over that jumps around ...

And, running 4 GPU Grid process is not totally out of the question ... with the two core cards I don't know if both would be placed on one task or that the card can run two tasks at the same time. In theory, if the cards run two processes at the same time, and I bought 3 of them, I would be able to run 6 GPU Grid tasks at the same time ...

Not that I have plans to do so ... for one thing I haven't paid for any of the latest system yet so I should not continue the buying spree ... and with the troubles in the BOINC Manager and the lack of projects using the GPU I am content to have just two cards running at this time ... though if the Mac PRo can start I will certainly be looking at trying to run on it too ...

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5424 - Posted: 9 Jan 2009 | 17:56:55 UTC - in response to Message 5414.

Which version of BOINC do you use? After I've updated it to 6.5.0 CPU usage became mor or less normal - between 13% and 20%. WinXP x64.

That is not at all normal. It should use almost NO CPU time:

Here's the analysis for my 9600 GSO:


Version 6.56 - Total 2 WUs completed OK

WUs worth 3232.06 credits - CPU secs: 334.95 to 338.27 each


Version 6.55 - Total 25 WUs completed OK

WUs worth 3232.06 credits - CPU secs: 34,341.28 to 51,533.36 each

That's over 100 to 1 difference

See this thread for more information:

http://www.gpugrid.net/forum_thread.php?id=605


Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5429 - Posted: 9 Jan 2009 | 21:41:44 UTC - in response to Message 5424.

Which version of BOINC do you use? After I've updated it to 6.5.0 CPU usage became mor or less normal - between 13% and 20%. WinXP x64.

That is not at all normal. It should use almost NO CPU time:


YOu are correct, it should use next to none ...

And one day soon, it should ... :)

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 5439 - Posted: 10 Jan 2009 | 12:38:50 UTC - in response to Message 5424.

That is not at all normal. It should use almost NO CPU time:


If normal is defined as "what we had during the previous months" then it's perfectly normal. If you define normal as "being ideal", then.. yes, 6.55 is not normal ;)

MrS
____________
Scanning for our furry friends since Jan 2002

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5528 - Posted: 12 Jan 2009 | 15:20:07 UTC - in response to Message 5439.
Last modified: 12 Jan 2009 | 15:21:18 UTC

That is not at all normal. It should use almost NO CPU time:


If normal is defined as "what we had during the previous months" then it's perfectly normal. If you define normal as "being ideal", then.. yes, 6.55 is not normal ;)

MrS

I was defining "normal" as the excellent behavior we had with v.6.56. V6.55 wastes a massive amount of CPU time for no reason. It negatively affects other projects that are not getting the CPU cycles they deserve. I proposed a simple intermediary solution here for 32bit machines:

http://www.gpugrid.net/forum_thread.php?id=605

That was on December 30th. So far still no answer. I hate to be a nag but you asked...

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 5551 - Posted: 12 Jan 2009 | 21:51:59 UTC - in response to Message 5528.

Well, I've been crunching here a bit longer than you and witnessed the first clients, which needed 100% of one core and I've ssen the results of the first attempts to reduce CPU usage, which lead to unpleasant performance drops of the GPU (less credit, overall). This is not meant to be an excuse, as we finally do have a much better method, but rather to put things into perspective. And since I dedicated one of my cores for months I don't mind some more days.

I suppose the project team is going straight for the new client. I can't tell you why they didn't implement this 32/64 Bit workaround. Maybe an incompatibility in the WUs due to other changes in the new client? Or maybe the error with 64 Bit XP is a sign of some other serious bug, which is/was still undiscovered and might render 6.56 results worthless, so they choose the "better be safe than sorry" approach? I know that when I spot something unexpected in my results it usually means there's some error in my code.. :p

Anyway, GDF posted just today:
"The delay in doing this [other] change is that we had a very bad period with the server scheduler which drained a lot of human resources."

MrS
____________
Scanning for our furry friends since Jan 2002

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5556 - Posted: 12 Jan 2009 | 22:09:09 UTC - in response to Message 5551.

"The delay in doing this [other] change is that we had a very bad period with the server scheduler which drained a lot of human resources."

MrS


I would guess that they are hydrated again ... in that that would be the opposite of drained ...
____________

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 5558 - Posted: 12 Jan 2009 | 22:16:27 UTC - in response to Message 5556.

Or irrigated?

MrS
____________
Scanning for our furry friends since Jan 2002

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5559 - Posted: 12 Jan 2009 | 23:05:35 UTC - in response to Message 5558.

Or irrigated?

MrS


Ok, does that mean they are back to "swampy"?

Or just that now they are all wet?

Just trying to clarify ...

(Sorry folks ... only sense of humor I have ...)

:)
____________

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5561 - Posted: 13 Jan 2009 | 0:52:04 UTC - in response to Message 5551.

Well, I've been crunching here a bit longer than you and witnessed the first clients, which needed 100% of one core and I've ssen the results of the first attempts to reduce CPU usage, which lead to unpleasant performance drops of the GPU (less credit, overall). This is not meant to be an excuse, as we finally do have a much better method, but rather to put things into perspective. And since I dedicated one of my cores for months I don't mind some more days.

MrS

In the case of v6.56 not only was CPU performance massively better but GPU performance improved a bit too.

Profile [SG]Arsenic
Send message
Joined: 19 Oct 08
Posts: 5
Credit: 2,217,455
RAC: 0
Level
Ala
Scientific publications
watwatwatwat
Message 5564 - Posted: 13 Jan 2009 | 1:31:34 UTC - in response to Message 5561.

If 6.56 only crashes 64-bit XP and 2003 machines, why not make it available via the app-info.xml way? That way, 6.55 would run on ALL machines and 6.56 would be available for people who would like/can run it.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5565 - Posted: 13 Jan 2009 | 2:09:26 UTC - in response to Message 5564.
Last modified: 13 Jan 2009 | 2:10:07 UTC

If 6.56 only crashes 64-bit XP and 2003 machines, why not make it available via the app-info.xml way? That way, 6.55 would run on ALL machines and 6.56 would be available for people who would like/can run it.

Great minds think alike :-) I asked the same question on 12/25, but no answer yet:

http://www.gpugrid.net/forum_thread.php?id=580&nowrap=true#4841

I considered setting one up myself but don't really want to do that without the admin's permission.

Post to thread

Message boards : Graphics cards (GPUs) : Windows 6.55 CPU Usage

//