High temperatures may be damaging my card

[Xipeos]

Hello.

I have an x1950pro that runs at very high temperatures with the radeon driver: idles at 69C, and quickly goes up to 80+C when playing games. I haven't though too much of it until now (stupid, I know), as I don't play that much on linux, but the new HB games changed that and I've noticed some big problems: I get white spots on the screen and occasionally a TTM-related Oops. Both likely mean bad VRAM, and the high temperatures might have caused it over time.

My card does not seem to "include the appropriate power state tables in the vbios", as I don't get the power_{method,profile} controls in sysfs (dynclks=1 also does nothing). Furthermore, the fan's duty cycle doesn't change at all, so it might need software control.

dmesg.

Any ideas on how to reduce those temps?

[droidhacker]

My philosophy on component temperatures is that no matter what, even without any kind of power management operational, the physical components used for temperature reduction should adequately be able to remove enough of the heat to keep the component safe over long term use.

If your temperatures are ever so high that it could potentially cause damage to your hardware, you should definitely reinforce the cooling system. Bigger/faster fan, bigger heat sink, etc.

Hello.

I have an x1950pro that runs at very high temperatures with the radeon driver: idles at 69C, and quickly goes up to 80+C when playing games. I haven't though too much of it until now (stupid, I know), as I don't play that much on linux, but the new HB games changed that and I've noticed some big problems: I get white spots on the screen and occasionally a TTM-related Oops. Both likely mean bad VRAM, and the high temperatures might have caused it over time.

My card does not seem to "include the appropriate power state tables in the vbios", as I don't get the power_{method,profile} controls in sysfs (dynclks=1 also does nothing). Furthermore, the fan's duty cycle doesn't change at all, so it might need software control.

dmesg.

Any ideas on how to reduce those temps?

[Xipeos]

My philosophy on component temperatures is that no matter what, even without any kind of power management operational, the physical components used for temperature reduction should adequately be able to remove enough of the heat to keep the component safe over long term use.

If your temperatures are ever so high that it could potentially cause damage to your hardware, you should definitely reinforce the cooling system. Bigger/faster fan, bigger heat sink, etc.

Your idea makes sense, of course, but there's not a lot I can do to the cooler due to its design*, and I can't find an aftermarker one. I hate to say this, but the card works well on windows (proper temps - 40C idle, fan control, etc). The high temperatures I'm getting are caused by something the driver's doing as far as I can tell..

*I guess I could take off the plastic and mount a bigger fan

[danwood76]

I used to use an x1950 pro, the stock cooler that came with it was terrible, it used to crash when being abused in windows and linux (temps were always reported lower in windows but the heatsink was the same temp and the power sonsumption remained around the same).

I ended up buying a zalman flower cooler and cutting up an old P1 cooler for the power supply ICs. The PWM mosfets get very hot and caused all sorts of crashes and temperature issues for me.

I stopped using the card around 6 months ago but never had any issues with heat dissipation with the newer heatsink combination.

The heatsink should be able to cool the card properly for it to run at 100% all the time otherwise it is not designed properly! Silly question but is it free from dust?

[Xipeos]

(temps were always reported lower in windows but the heatsink was the same temp and the power sonsumption remained around the same).

That's a bit disconcerting.. The heatsink doesn't feel that hot to the touch, so maybe the readings are wrong.

I ended up buying a zalman flower cooler and cutting up an old P1 cooler for the power supply ICs. The PWM mosfets get very hot and caused all sorts of crashes and temperature issues for me.

I stopped using the card around 6 months ago but never had any issues with heat dissipation with the newer heatsink combination.

Looked for Zalman heatsinks as well; can't find any around here.

Silly question but is it free from dust?

Yes, I clean it regularly. Also, I think the readings from Catalyst are more likely to be correct than those through radeon+lm63.

Thanks for the suggestions. I'll try to confirm the tempretatures with an external sensor. Since no one else has the same problem, the readings might be wrong and the card might just be dying from old age :-|.

[danwood76]

If the heatsink isn't warm to touch and its reading high then I expect the heatsink is probably not making good contact with the card itself. Even with the best heatsink in the world if its not fitted correctly then it wont work.

Quite often manufacturers put too much thermal compound on the cores, I would try taking the heatsink off the card and refitting it with a small amount of silver based compound and see if the heat issue improves.

lm-sensors should be reading the correct temperatures BTW, the difference is that AMD might adjust this value in their software to allow for the position of the sensor in relation to the core.

[Xipeos]

If the heatsink isn't warm to touch and its reading high then I expect the heatsink is probably not making good contact with the card itself. Even with the best heatsink in the world if its not fitted correctly then it wont work. (...)

Accessing and cleaning the heatsink on this card requires complete disassembly of the cooler, so I've already changed the TIM quite a few times. Think I have enough experience by now to do a good job, but I guess human error is always a possibility - but in that case I must have been doing things wrong the entire time, as I've been getting those temps for as long as I can remember.

lm-sensors should be reading the correct temperatures BTW, the difference is that AMD might adjust this value in their software to allow for the position of the sensor in relation to the core.

Got it.

[crazycheese]

Check Arctic Cooling VGA coolers, esp. Accelero L2 Pro might interest you. Buy cooler with plenty of thermopad heatsinks and make sure to rub the memory and mosfet chips surfaces with pencil eraser, before you put on thermopad. 4770 card gets around 50C max full load when using that cooler in 7v silent mode.

[droidhacker]

Your idea makes sense, of course, but there's not a lot I can do to the cooler due to its design*, and I can't find an aftermarker one. I hate to say this, but the card works well on windows (proper temps - 40C idle, fan control, etc). The high temperatures I'm getting are caused by something the driver's doing as far as I can tell..

*I guess I could take off the plastic and mount a bigger fan

Here you go: http://www.zalman.co.kr/ENG/product/...ad.asp?Idx=283
Or, if you like copper: http://www.newegg.com/Product/Produc...82E16835118037

This one should also work;

http://www.newegg.com/Product/Product.aspx?Item=N82E16835118072

[Xipeos]

Here you go: http://www.zalman.co.kr/ENG/product/...ad.asp?Idx=283
Or, if you like copper: http://www.newegg.com/Product/Produc...82E16835118037

This one should also work;
http://www.newegg.com/Product/Produc...82E16835118072

Can't find a good one *in my country*. Lots of shops list them online, but don't actually have the products in stock. The only ones in stock are either too expensive (ex: prolimantech mk13 - >2/3 the price of a new card ) or not really an improvement (at least not according to the articles I read, ex: Deepcool V4000).

I hooked up the card's fan to the PSU, so at least I'm getting more airflow, but I haven't been able to test it at all cause my mobo died unexpetedly due to unrelated reasons. Think this thread will fall off the first page by the time I get a replacement through RMA.

Thanks everyone for the sugggestions.