GTX 970 stability problem solved...

Moderators: Site Moderators, PandeGroup

GTX 970 stability problem solved...

Postby RipD » Thu Feb 11, 2016 7:32 pm

I have a GTX 970 that has never folded reliably. (I don't think it's card or driver specific issue.) After struggling to it stable I realized that some WUs are pushing the temp up even with core and memory clocks lowered. Also realized that this card does not like to get near 80 degrees. No idea why, but it doesn't.

I set MSI Afterburner to keep the temp at 75. Have had no issues since doing that. Yes, it runs more slowly. I'll take a few less PPD and reliability rather than the constant headache of trying to find the right clock speeds. What's great about this as well is that I can set the fan speed manually, so it'll keep the card at 75 degrees with a loud or quiet fan. The output changes, of course, but it takes out the reliability issues of trying to manage the core and memory clocks.

A very simple solution. I feel silly for not seeing this solution sooner.

(edited for typos)
Last edited by RipD on Fri Feb 12, 2016 4:43 am, edited 3 times in total.
Image
RipD
 
Posts: 99
Joined: Fri Dec 14, 2007 2:34 am

Re: GPU stability problem solved...

Postby wilding2004 » Thu Feb 11, 2016 8:29 pm

I do something similar.

On a pair of 970's, I set the fans manually to 100%, set a 75mhz overclock, and set power to 105%. Then I set temp to 75c and make that priority over power. For some WU's the cards run with the overclock, for others that run hotter, the temp limit kicks in and reduces the overclock and power.

Both cards been running stable 24/7 for a few months now.
wilding2004
 
Posts: 103
Joined: Thu Jan 10, 2008 4:16 pm

Re: GPU stability problem solved...

Postby bruce » Thu Feb 11, 2016 8:41 pm

wilding2004 wrote:I do something similar.

On a pair of 970's, I set the fans manually to 100%, set a 75mhz overclock, and set power to 105%. Then I set temp to 75c and make that priority over power. For some WU's the cards run with the overclock, for others that run hotter, the temp limit kicks in and reduces the overclock and power.

Both cards been running stable 24/7 for a few months now.


That's a pretty interesting way to do it.

Out of curiosity, with whatever case/fans/etc. provides your cooling, how does that compare to what you would see if you used all of the driver's default settings?
bruce
 
Posts: 21276
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: GPU stability problem solved...

Postby wilding2004 » Thu Feb 11, 2016 10:28 pm

bruce wrote:
wilding2004 wrote:I do something similar.

On a pair of 970's, I set the fans manually to 100%, set a 75mhz overclock, and set power to 105%. Then I set temp to 75c and make that priority over power. For some WU's the cards run with the overclock, for others that run hotter, the temp limit kicks in and reduces the overclock and power.

Both cards been running stable 24/7 for a few months now.


That's a pretty interesting way to do it.

Out of curiosity, with whatever case/fans/etc. provides your cooling, how does that compare to what you would see if you used all of the driver's default settings?


With everything at default, the gpu's got well into the high 80's, and the gpu fans never made it above 60%. I guess the default profile is designed to keep the noise down. Also at default, I saw power usage at 110% quite a lot. Would power use be higher at a high temp? or another way to say that might be...are gpus less efficient at higher temp.

As for stabilty. I did get more failed WU's at default, but that could have been down to any number of reasons not related to how I use Afterburner.
wilding2004
 
Posts: 103
Joined: Thu Jan 10, 2008 4:16 pm

Re: GPU stability problem solved...

Postby RipD » Fri Feb 12, 2016 4:41 am

wilding2004 wrote:I do something similar.

On a pair of 970's, I set the fans manually to 100%, set a 75mhz overclock, and set power to 105%. Then I set temp to 75c and make that priority over power. For some WU's the cards run with the overclock, for others that run hotter, the temp limit kicks in and reduces the overclock and power.

Both cards been running stable 24/7 for a few months now.

I may try those settings to see if I get a little better output. I just went back to stock settings after I got the temp issue figured out. If I'm temp constrained the clock settings may not matter much.

Thanks for mentioning setting the temp as the priority - I neglected to do so. (The default on Afterburner is Power as the priority rather than temp. If you sent Temp as the priority it will reduce power to keep the card at the temp you specify).

This is interesting to play with: once you set temp, you can change the fan speed and watch the power on the card go up or down correspondingly. You can essentially change the output of the card by setting the fan speed. It allows you to easily balance output, stability, and noise. (Where noise isn't an issue I don't see any reason to go below 100% on fan speed.)
RipD
 
Posts: 99
Joined: Fri Dec 14, 2007 2:34 am

Re: GPU stability problem solved...

Postby foldy » Fri Feb 12, 2016 6:22 pm

One of my GPUs also did not like temp over 80°C and driver crashed.
I solved it by renew the GPU thermal paste.
The temps still hit 80°C but GPU stay stable.
foldy
 
Posts: 911
Joined: Sat Dec 01, 2012 3:43 pm

Re: GPU stability problem solved...

Postby artoar_11 » Fri Feb 12, 2016 7:19 pm

By NVidia Inspector I reduce the voltage on the GTX 970 Asus Strix, from 1.212 V to 1.187 V. Works stable @ 1430 MHz already three months.
When voltage is 1.162 V also run stable @ 1360 MHz.
The temperature is below 70°С, fans (2 pieces) @1700-1800 rpm. Power Limit @ 120%.
User avatar
artoar_11
 
Posts: 665
Joined: Sun Nov 22, 2009 8:42 pm
Location: Bulgaria/Team #224497/artoar11_ALL_....

Re: GPU stability problem solved...

Postby bruce » Sat Feb 13, 2016 5:41 am

Does anybody know if the same settings are possible in the Linux drivers?
bruce
 
Posts: 21276
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: GPU stability problem solved...

Postby mmonnin » Sat Feb 13, 2016 2:52 pm

Not that I see. Very few options. Not even a fan profile.
mmonnin
 
Posts: 365
Joined: Wed Dec 05, 2007 1:27 am

Re: GPU stability problem solved...

Postby RipD » Sun Feb 28, 2016 1:55 pm

Just FYI, tried some more aggressive settings and it led to intermittent problems. My card GTX 970 has a slight overclock from the factory. In Afterburner I have both core and memory clocks set to -100 with the temp limit set to 75. I've been stable for a week. There are many variables you can play with, including coolers and thermal paste. I've done all of them. For somebody who just wants their card to get stable, I'm finding that running Afterburner with a temp limit is an effective, easy way to get your card to perform reliably.
RipD
 
Posts: 99
Joined: Fri Dec 14, 2007 2:34 am


Return to General GPU client issues

Who is online

Users browsing this forum: No registered users and 3 guests

cron