Overheating GPU question

[Svensk Viking]

My video card has been overheating alot recently. I know that it is overheating, since there are artefacts on screen, driver stopping responding, crashing etc. I have seen in temperature tools that the temps are between 90-105 Celsius
Drivers should not be the issue, since I have tried other drivers, using driver sweeper to get rid of the old remains

Yesterday though, when I played Crysis, something odd happened...I played it for about 3-4 hours, noticing the overheating symptoms...After that time though, the symptoms disappeared

Looking at the temperature tools, the GPU is now between 85-90 Celsius, and Furmark can "only" push it to 103 degrees, whereas it pretty quickly goes down, and the highest it goes to then is 97 degrees(where there is some frequent artefacting)

Is it likely that the GPU simply is downclocking itself, or any other reason?

I can RMA it, but if they don't detect any fault, I have to stand for the costs, which sums up to about 1/4 of what the GPU costs...Not knowing their testing methods make me feel quite unsure too

 

[Grall]

1: Check fan/heatsink for dust contamination. Clean out with pressurized air.

If problems still persist:
2: test card in another PC.

 

[digitalwanderer]

Out of curiosity, what card? Those are pretty extreme temps.

 

[Svensk Viking]

It is a GTS 250, have had it since june last year

Forgot to mention that I have controlled that the whole PC is free of dust, and all the fans are running
Downloaded MSI Afterburner to manually set the speed to 100%, but that only helps temporarily

Don't have another PC to test it in

 

[digitalwanderer]

Downloaded MSI Afterburner to manually set the speed to 100%, but that only helps temporarily

Only helps temporarily? Does that mean it keeps the temps down when it's going 100% or that it only keeps the temps down for a while?

You might have some bad thermal paste action going on.

 

[Grall]

Borrow a friend's PC for a couple hours. Surely you can't be the only guy in town with a pcie 16x slot...

 

[Svensk Viking]

Only helps temporarily? Does that mean it keeps the temps down when it's going 100% or that it only keeps the temps down for a while?

You might have some bad thermal paste action going on.

That it only keeps them down for a while, that they only take a bit longer to go up to the high levels

The thought of looking at the thermal paste has hit me, but I think it breaks the warranty if they can see that I have fiddled with it physically

Borrow a friend's PC for a couple hours. Surely you can't be the only guy in town with a pcie 16x slot...

Well, you are right there...I just don't need to tell them why^^

 

[digitalwanderer]

The thought of looking at the thermal paste has hit me, but I think it breaks the warranty if they can see that I have fiddled with it physically

Yeah, it's a kind of catch-22 that I hate too. If it's still under warranty I'd return it and then if they say there is no problem play hardball with them about its temps being too high. Temps like that will insure a very short card life.

 

[Svensk Viking]

When playing Crysis Warhead DX10 vs DX9 with all other settings the same, it crashes alot faster in DX10...Sometimes, I can't catch a glimpse of artefacts when it crashes in DX10

 

[Malo]

What is your case ambient?

 

[ShaidarHaran]

A GTS 250 should never get that hot unless the cooling system is malfunctioning in some fashion. I'd start by checking for "play" in the heatsink - see if it feels loose at all. It's also possible the fanpath is clogged or the fan itself is no longer functional.

I've had several G92-based cards, I don't think I ever saw GPU temps > 70C even on Furmark, and that was on an 8800 GTS 512 massively overclocked with a voltmod on stock cooling and auto fan speed. I also had an 8800 GT overclocked with a Thermaltake DuOrb and that thing never got hotter than ~50C.

 

[Svensk Viking]

What is your case ambient?

Dunno how hot it is in the case, but I have had table fan blowing into it, and that took the temps down with 5-10 degrees...When I tried that, it was still in the 85-95 area

A GTS 250 should never get that hot unless the cooling system is malfunctioning in some fashion. I'd start by checking for "play" in the heatsink - see if it feels loose at all. It's also possible the fanpath is clogged or the fan itself is no longer functional.

I've had several G92-based cards, I don't think I ever saw GPU temps > 70C even on Furmark, and that was on an 8800 GTS 512 massively overclocked with a voltmod on stock cooling and auto fan speed. I also had an 8800 GT overclocked with a Thermaltake DuOrb and that thing never got hotter than ~50C.

The heatsink doesn't seem lose, at least not noticeably for me to feel
And I have already seen that the fan is working, and setting it to 100% from the beginning makes it take a little longer for it to get too hot

 

[Malo]

Are you using driver version 196.75?

 

[Svensk Viking]

Are you using driver version 196.75?

I had them, but reverted back to 196.21...Used Driver Sweeper too to get rid of them

I know what those drivers did

 

[Silent_Buddha]

Hmmm, is there a BIOS editing tool for Nvidia cards similar to ATI cards? You could check to see if it's inadvertently being over-volted. That would certainly cause high temps.

Regards,
SB

 

[Malo]

Yeah, unless there's something going on with the voltage as mentioned, then it's all pointing towards a bad heatsink mounting.

 

[digitalwanderer]

Yeah, unless there's something going on with the voltage as mentioned, then it's all pointing towards a bad heatsink mounting.

That's my bet too.

 

[pcchen]

Do you have another card sits right next to the video card? Sometimes this may block the airflow. I had a GeForce 8800GT with another card sitting right next to it, and it sometimes hit more than 100 C when running some demanding games (then crashed, of course). After moving the neighboring card to another location, it's never more than 80 C.

 

[Svensk Viking]

Hmmm, is there a BIOS editing tool for Nvidia cards similar to ATI cards? You could check to see if it's inadvertently being over-volted. That would certainly cause high temps.

Regards,
SB

Will see if there is a recommended tool...What can that depend on though? I have kept everything at their stock speeds, not touching anything in the BIOS

Yeah, unless there's something going on with the voltage as mentioned, then it's all pointing towards a bad heatsink mounting.

Well, I guess it is worth taking a look again

Do you have another card sits right next to the video card? Sometimes this may block the airflow. I had a GeForce 8800GT with another card sitting right next to it, and it sometimes hit more than 100 C when running some demanding games (then crashed, of course). After moving the neighboring card to another location, it's never more than 80 C.

No, it is alone...And if it was something blocking the airflow, I think the problems would have been appeared from the beginning

 

[Silent_Buddha]

Will see if there is a recommended tool...What can that depend on though? I have kept everything at their stock speeds, not touching anything in the BIOS.

Who knows? Could have been factory set. The paranoid side could think of a rogue virus that targets Nvidia cards (I'd find this rather hard to believe, but who knows ).

It's just something that might be worth checking if everything else is how it should be. Something is obviously causing unreasonably high temps, so it's worth it to check every possibility no matter how remote.

Regards,
SB