Problem with GTX470 temps running SETI

Message boards : Number crunching : Problem with GTX470 temps running SETI
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile lupo

Send message
Joined: 29 Aug 10
Posts: 91
Credit: 4,736,407
RAC: 0
United States
Message 1226534 - Posted: 3 May 2012, 5:01:30 UTC

I have been using CUDA and the SETI project for almost a year now. When I have the client running at 100% (all 12 processor cores and the two GTX 470's) the processor never gets above about 115f and the cards would always be around 60c. (One car would always be about 10c higher than the other. The card further back in the system would be at about 50c.)

In the past few weeks though, the video cards are getting out of control with their heat. (I have checked the cards for blocked fans, loose heat-sink, etc. Everything looks fine physically.) When I run the client within 10 minutes the top card will get over 90c with the secondary card getting to about 70c.

I am running the latest nvidia drivers and the latest BOINC client. Here is my question:

Are there changes I can make to the clock frequencies of the card that will drop the temps but not decrease the processing power of the cards. Or, are the clock frequencies affect the performance of BOINC on the cards?

Thanks in advance!
ID: 1226534 · Report as offensive
tbret
Volunteer tester
Avatar

Send message
Joined: 28 May 99
Posts: 3380
Credit: 296,162,071
RAC: 40
United States
Message 1226552 - Posted: 3 May 2012, 5:30:48 UTC - in response to Message 1226534.  

I have been using CUDA and the SETI project for almost a year now. When I have the client running at 100% (all 12 processor cores and the two GTX 470's) the processor never gets above about 115f and the cards would always be around 60c. (One car would always be about 10c higher than the other. The card further back in the system would be at about 50c.)

In the past few weeks though, the video cards are getting out of control with their heat. (I have checked the cards for blocked fans, loose heat-sink, etc. Everything looks fine physically.) When I run the client within 10 minutes the top card will get over 90c with the secondary card getting to about 70c.

I am running the latest nvidia drivers and the latest BOINC client. Here is my question:

Are there changes I can make to the clock frequencies of the card that will drop the temps but not decrease the processing power of the cards. Or, are the clock frequencies affect the performance of BOINC on the cards?

Thanks in advance!


I don't think you'd be horrified by the results of downclocking the cards a little. I also don't think you'd be all that happy with the temperature drop.

Has the ambient temperature in your room risen a few degrees in the past few weeks?
ID: 1226552 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1226566 - Posted: 3 May 2012, 6:04:47 UTC - in response to Message 1226534.  
Last modified: 3 May 2012, 6:05:37 UTC

Ambient temps getting higher? They are here. Maybe add more fans to your case to improve air flow.

Have you taken the cards out and used canned air or a compressor to blow the dust out of the cards. I have had to (kitty fur).

Did the temps increasing coincide with your installation of the latest NV driver? Try going back to the older driver you were using.

Do you have the GPU fans locked at 100% with EVGA Precision or another overclocking tool? Verify with GPUZ that the fans are running at top speed.

Nothing you can do in the way of reducing clock speeds that will reduce the temps substantially without reducing crunching output. They go hand in hand.
But, reducing the GPU clock and RAM speed will affect crunching less than reducing the shader clock, and may reduce temps somewhat.

That's about all I can think of right now to check out.

Best of luck with it!
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1226566 · Report as offensive
Profile lupo

Send message
Joined: 29 Aug 10
Posts: 91
Credit: 4,736,407
RAC: 0
United States
Message 1226568 - Posted: 3 May 2012, 6:07:46 UTC - in response to Message 1226552.  

No, room temps have not changed. So are you saying that the shader, memory and core infrequence will not slow down the processioning of the SETI WU?

Second, winch one of the settings should I change? How much can the cards the be overclocked without affecting the WU processing too much?
ID: 1226568 · Report as offensive
Profile lupo

Send message
Joined: 29 Aug 10
Posts: 91
Credit: 4,736,407
RAC: 0
United States
Message 1226573 - Posted: 3 May 2012, 6:17:00 UTC - in response to Message 1226566.  

I have already checked the cards physically and blown the heat sinks out with canned air. The room temps have not changed.

I use the EVGA tool which automatically sets the fan speeds but I can see in my cad gadget that the fans are at 100%.

Very strange. This is why I was hoping that I cold just down-clock the card to get the temps down. Besides that I'm out of ideas. :(
ID: 1226573 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1226576 - Posted: 3 May 2012, 6:18:05 UTC - in response to Message 1226568.  

No, room temps have not changed. So are you saying that the shader, memory and core infrequence will not slow down the processioning of the SETI WU?

Second, winch one of the settings should I change? How much can the cards the be overclocked without affecting the WU processing too much?

Yes, all of those speeds will affect WU processing. RAM probably the least, core second, and shader speed will have the most effect.

The question is, why did the temps go up in a short period of time.
Something changed a lot. Did you just start to OC the cards?

What about the GPU fan speeds?
I have 1 rig that for some reason wants to keep changing a GPU fan speed to auto instead of 100% every time it reboots, and I have to manually set it back to 100%...that will affect GPU temps a LOT.

"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1226576 · Report as offensive
Kevin Olley

Send message
Joined: 3 Aug 99
Posts: 906
Credit: 261,085,289
RAC: 572
United Kingdom
Message 1226577 - Posted: 3 May 2012, 6:21:54 UTC - in response to Message 1226534.  

I have been using CUDA and the SETI project for almost a year now. When I have the client running at 100% (all 12 processor cores and the two GTX 470's) the processor never gets above about 115f and the cards would always be around 60c. (One car would always be about 10c higher than the other. The card further back in the system would be at about 50c.)

In the past few weeks though, the video cards are getting out of control with their heat. (I have checked the cards for blocked fans, loose heat-sink, etc. Everything looks fine physically.) When I run the client within 10 minutes the top card will get over 90c with the secondary card getting to about 70c.

I am running the latest nvidia drivers and the latest BOINC client. Here is my question:

Are there changes I can make to the clock frequencies of the card that will drop the temps but not decrease the processing power of the cards. Or, are the clock frequencies affect the performance of BOINC on the cards?

Thanks in advance!


Keep the cards clean, even a very fine layer of dust over the heatsink will increase the temps, I remove the outer plastic cover and use a fine brush to clean the fan and heatsinks.

Check your case fans and filters, if the amount of fresh air into the case is reduced this can also cause a temp rise.

The later Nvidia drivers may have also caused a temperature rise, When I first started using GTX470's I could overclock and stay within a safe temp range but as the drivers improved I had to drop back to normal to hold temps down.

I also run TThrottle, this program will slow down the cards if temps try to rise ie: when processing shorties, which do cause temp spikes.

As I write this mine are at 65C, 73C and 78C. They were cleaned a couple of weeks ago and will cleaned again next weekend.

I am running 290.53





Kevin


ID: 1226577 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13722
Credit: 208,696,464
RAC: 304
Australia
Message 1226585 - Posted: 3 May 2012, 6:43:55 UTC - in response to Message 1226573.  

I use the EVGA tool which automatically sets the fan speeds but I can see in my cad gadget that the fans are at 100%.

What RPM are they actually running at?

My GTX560Ti a couple of weeks ago stared to run hotter than usual. A lot hotter than usual. Running GPUz showed that the fans were at 100% speed, but their actual speed was less than 2,000 RPM and the temperature was up to 85°. Usually the speed as a percentage is around 75%, the temperature around 70° & their actual speed around 3,200 RPM.

At time went by, they ended up slowing down to less than 800RPM & the temperaature went up to the high 90°s.
After a lot of swearing & mucking around, i managed to clean out & lubricated the bearings & it's running happily again.
Grant
Darwin NT
ID: 1226585 · Report as offensive
tbret
Volunteer tester
Avatar

Send message
Joined: 28 May 99
Posts: 3380
Credit: 296,162,071
RAC: 40
United States
Message 1226604 - Posted: 3 May 2012, 10:00:37 UTC - in response to Message 1226568.  
Last modified: 3 May 2012, 10:02:16 UTC

No, room temps have not changed. So are you saying that the shader, memory and core infrequence will not slow down the processioning of the SETI WU?

Second, winch one of the settings should I change? How much can the cards the be overclocked without affecting the WU processing too much?


No, what I'm saying is that slowing down the card 10% will drop your RAC 10% but that's not too high of a price to pay to keep the cards from burning-up.

What I'm *also* saying is that I don't think your temps will drop much even if you slow the card by 20%. I could be wrong. All you can do is try it and see, but at least it is a quick and non-destructive experiment.

But I'm guessing what I'm guessing based on my factory overclocked 560ti cards that started throwing errors and had to be downclocked a little to make them stable again. Heat was never an issue with them and downclocking them the amount I did (from overclock to stock, initially) didn't make them substantially cooler (if at all).

If you haven't changed anything, I'm with MSATTLER. Something else has changed. If you were experiencing this with one card only, I'd say the card may be failing in some way. With two doing the same thing, I'm thinking that something else is happening. What else? I haven't a clue.

So starting somewhere, in order to move forward:

Try downclocking them 10% and see what happens.

You might also want to see what happens if you unplug one of the cards from the PSU. That's a real shot in the dark, but so many bad things can happen if a PSU gets flaky that it's worth a try.

BTW - you might want to post a screen shot of what GPU-Z is telling you both "front page" and "Sensors."

I think that would help people who are trying to help you more than anything else.

It's easy. There's a little camera in the top right corner of the GPU-Z screen, click it, free image host it, copy the URL into your message here.
ID: 1226604 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13722
Credit: 208,696,464
RAC: 304
Australia
Message 1226608 - Posted: 3 May 2012, 10:16:32 UTC - in response to Message 1226604.  

If you were experiencing this with one card only, I'd say the card may be failing in some way. With two doing the same thing, I'm thinking that something else is happening.

Or the poor cooling of one card is affecting the cooling of the other card.
Pull out the card that is running the hottest. See if the second card comes back down to it's earlier temperatures.

Grant
Darwin NT
ID: 1226608 · Report as offensive

Message boards : Number crunching : Problem with GTX470 temps running SETI


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.