Advanced search

Message boards : Graphics cards (GPUs) : NVIDIA GTX 295 & driver 197.45 too powerful (?)

Author Message
Profile Dieter Matuschek
Avatar
Send message
Joined: 28 Dec 08
Posts: 58
Credit: 231,884,297
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 17443 - Posted: 29 May 2010 | 20:10:31 UTC

I have been absent to GPUGRID for some time. So I'm not up to date.

Today I installed the newest driver 197.45 on one PC and started crunching WUs. Several times the following behavior occurs. After a few minutes the video output gets lost and I have to shut down the PC. The video card is very, very hot: my impressions are that touching it for a few seconds would probably burn my fingers.

On the other hand, another PC with GTX 295 and old driver 190.38 runs well. I think to have noticed that crunching with 197.45 is much faster than with 190.38.

So, my questions are:
(1) Are these observations correct?
(2) Should I reinstall 190.38 to be able crunching without cooling problems?

Any advice is appreciated. Thanks.

____________

Profile Dieter Matuschek
Avatar
Send message
Joined: 28 Dec 08
Posts: 58
Credit: 231,884,297
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 17444 - Posted: 29 May 2010 | 21:17:45 UTC - in response to Message 17443.

45 minutes ago I've reinstalled 190.38 and restarted BOINC. All seems to work properly again.

Mmmh . . . it's bothering anyway.
How can the new driver (with CUDA 3000) work better than the old one (with CUDA 2030) on a GTX 295?

Can anyone confirm or reject these observations?
____________

ftpd
Send message
Joined: 6 Jun 08
Posts: 152
Credit: 328,250,382
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 17445 - Posted: 29 May 2010 | 23:27:39 UTC - in response to Message 17444.

Dieter,

I am also user of windows-xp-pro + gtx295.
You can also update to nvidia-driver (beta) 257.15, this works OK for all cards.

Good luck,








____________
Ton (ftpd) Netherlands

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 17446 - Posted: 30 May 2010 | 1:09:35 UTC - in response to Message 17444.

How can the new driver (with CUDA 3000) work better than the old one (with CUDA 2030) on a GTX 295?


It cannot work any better.
CUDA3000 is slower than CUDA2030 and only needed for Fermi cards.

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 17450 - Posted: 30 May 2010 | 7:33:55 UTC
Last modified: 30 May 2010 | 7:37:50 UTC

Well i'm running 197.45 on all my cards (2 machines have GTX295's). The 2nd generation cards (ie the single PCB) run a lot cooler than the 1st generation. If you are concerned about the temperature grab a copy of GPU-Z to see how hot they actually are.

I was running 196.21 drivers but recently went to 197.45. Both support cuda 3.0. The 257.xx drivers that ftpd refers to support cuda 3.1. Neither version is of any interest (at this point) to a GTX295 as the GPUgrid apps are still compiled using cuda 2.2 for us. GPUgrid has an app for the GTX470/480 users (ie Fermi) which does use cuda 3.0, but you'd have to have a Fermi based card to get it.
____________
BOINC blog

Profile Dieter Matuschek
Avatar
Send message
Joined: 28 Dec 08
Posts: 58
Credit: 231,884,297
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 17452 - Posted: 30 May 2010 | 8:00:51 UTC

Well, thanks for all the answers.

Tonight the PC has shut down by itself. So IMHO there seems to be a problem with adequate cooling on this PC.
____________

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 17455 - Posted: 30 May 2010 | 10:37:20 UTC

Take a lookt at the GPU temperatures. Maybe take the card out and clean it (blow onto fan and exhaust while fixing the fan with a finger). And you may have to increase the fan speed, if the card isn't at maximum already. There are some tool which can do this. RivaTuner is classic for nVidia, although I'm not too fond of its UI.

MrS
____________
Scanning for our furry friends since Jan 2002

Siegfried Niklas
Avatar
Send message
Joined: 23 Feb 09
Posts: 39
Credit: 144,654,294
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 17457 - Posted: 30 May 2010 | 11:18:16 UTC - in response to Message 17455.

I have always been a big Fan of RivaTuner - but we have to accept that its developer ("Unwinder") did not update RivaTuner for month.

The reason for the update delay is the huge amount of time that "Unwinder" have to
spend for the development of the MSI Afterburner (based on Rivatuner).

Unwinder told us in RivaTuner's Forum that he may be able to continue its development in July.

Until then it may be the best idea to use the MSI Afterburner.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 17459 - Posted: 30 May 2010 | 12:32:53 UTC - in response to Message 17452.

Well, thanks for all the answers.

Tonight the PC has shut down by itself. So IMHO there seems to be a problem with adequate cooling on this PC.

I use a GPU/CPU throttling program named eFMer BOINC TThrottle that's very useful for controlling heat & resource use, interfaces with BoincTasks:

BOINC TThrottle Client Temp Control & Monitor

Profile liveonc
Avatar
Send message
Joined: 1 Jan 10
Posts: 292
Credit: 41,567,650
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwat
Message 17463 - Posted: 30 May 2010 | 15:55:24 UTC - in response to Message 17455.

I used nvclock & ran the fan speed 99%, temps dropped 2-5 degrees. Frustrated, I took out the GPU's & cleaned it of dust (there was much more then I thought), temps dropped 10 degrees! If it only dropped slightly I was about to take off the heat sink & replace the thermal compound with Arctic Silver, but cleaning gave a good enough result.
____________

Profile Dieter Matuschek
Avatar
Send message
Joined: 28 Dec 08
Posts: 58
Credit: 231,884,297
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 17464 - Posted: 30 May 2010 | 16:34:16 UTC
Last modified: 30 May 2010 | 16:34:53 UTC

Dust has indeed been the problem: thanks to Beyond and to ExtraTerrestrialApes.

Limiting to 100°C temperature for the GPU leads to 50% GPU usage on one PC and to far less on the other PC. I took the cards out of both PCs and cleaned them as described.

Now, one of them works well with 92°C at full power but the other card is dead. Its temperature must have been a lot more than the allowed 105°C for some time.

(Okay, I have learned another lesson.)

Bye,
Dieter
____________

Post to thread

Message boards : Graphics cards (GPUs) : NVIDIA GTX 295 & driver 197.45 too powerful (?)

//