I recently decided that it's time to make a serious effort to learn about overclocking. Before I really got started, however, I was brought up short by what seem to be unreasonably high temperatures on my graphics card [link] .
Just so I'd know where I'm starting from, I ran FurMark and noticed that my GPU temps were in the 90-95C range. I checked the reading in CCC, GPU-z and the Asus SmartDoctor (I have an Asus HD5850 reference design card, if you didn't click the link) to confirm the reading, and reproduced the temperature by goofing around in Battlefield 3. I tried to do it with Unigine Heaven, but it only got up to the low 80s in the bench marking tool.
To diagnose/address the problem, I've tried: updating my drivers (worth a try, but unhelpful), checking all my fans (they're all spinning as quickly as I'd expect), dusting the case (I'm pretty OCD about dust, so there wasn't any buildup, but now it's clean enough to eat off of), manually setting the GPU fan to 100% (don't noticeably decrease Furmark/BF3 temps), and taking off the side panel (didn't help, so case airflow doesn't seem to be an issue).
One other disturbing symptom is that when running FurMark (neither in games nor Unigine Heaven), once the temperature has started to level off (it's gotten as high as 95C and rarely dips below 92C, long term) the fan on the GPU becomes erratic, and I can hear it changing between 100% and maybe 50% (wild guess, based on sound) with a period of just a second or two. This happens whether I use Asus SmartDoctor or CCC to control the fan behavior (I even went the to point of uninstalling the Asus tool, thinking that perhaps the two applications were interfering with each other somehow, even when shut off).
If I'm just being paranoid, just let me know (and please give me an opinion on an actual safe max operating temperature) and I'll happily go about my original plan of clocking up this GPU till I have to replace it. However, posts here and around the web seem to indicate that 80C would be a more reasonable temperature to hit in FurMark, with an entirely stock setup, and I'd rather not abuse my GPU if I can avoid it. Options I'm considering include: 1) Sit tight, don't overclock, replace the card as soon as is feasible. 2) OC, risk damaging the card and having to play browser games on my 8800gt until I can get a replacement. 3) Reapply the TIM. 4) Buy and aftermarket GPU cooler. 5) Try to go through the ASUS warranty process.
Each option has downsides, and I'd greatly appreciate any advice. For reference, my full system is:
Windows 7 Professional 64-bit Service Pack 1
Gigabyte Z68X-UD3-B3, BIOS: Award Modular BIOS v6.00PG
Intel(R) Core(TM) i5-2500K CPU @ 3.30GHz (4 CPUs), ~3.6GHz
8GB RAM, G.SKILL Ripjaws Series 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800)
Asus ATI Radeon EAH HD 5850
730W Kingwin PSU
LIAN LI PC-A70F Case
Try replacing the TIM? I found that on my old GTX 260 after blowing it out thoroughly with an air compressor it still ran hot, so I took it apart and put some Arctic Silver Matrix on it and saw my idle and load temps drop back down to normal.
It's worth a shot. You can grab Arctic Silver 5 @ Radio Shack, or if you have a Fry's or Microcenter or even Best Buy you should be able to find something locally and get it done quicker.
rolli59 - To clarify: Both FurMark and Battlefield 3 (a game) bring the temperature up to the hot zone of ~92C. The bit that only FurMark does is the erratic behavior of my GPU fan.
DMI186 - Thanks, I have Kombustor installed as part of my prep to OC, but I didn't think to use it to check it's effect on temps. I'll do that tonight and update.
cliffro - I've never taken apart a GPU before, so I was hesitant to do this. I'll do some video watching today. Will it be a fairly casual operation, similar to changing the TIM on a CPU? Also, I have some TIM left from a Corsair heat sink I purchased and installed on my CPU over the weekend; will that suffice, do you think? I have a Radio Shack and a Best Buy nearby, but neither stocks thermal paste.
Here are some Links to after market type coolers if you have the room. They are not the glove type. One more test. Have you taken of the side and ran a house fan on high to see if the temps come down. If they do then you know for sure that it is the GPU cooler.
I believe so. I wood hate to see you get one and that isn't the problem. After looking at all the different coolers I believe most are universal. They showed 4 screw holes and that would be a better choice then the three I suggested.
It could be just you adding past to your one now. Maybe you should try that first and see if that is the problem. And thanks for trying everything we all suggested.
Whoohoo! Thanks for the advice, and DM186 for the persistence, especially.
Final update: I went ahead and replaced the TIM on my GPU. After removing the shroud and clearing the dust, it became clear that Asus hadn't put a great deal of care into assembling this card. Not only did part of the thermal pads covering the memory unnecessarily obstruct one of the screws holding the shroud to the heatsink, but the thermal paste on the GPU itself was misapplied. TIM had oozed off all four sides of the chip, but there was still a rice grain sized hole right in the center of the chip that had no paste at all. It seemed like the TIM had been applied in some sort of circle originally and an air bubble had gotten trapped in the center. I cleaned it up and applied some Corsair OEM paste I had leftover from last week and reassembled the whole thing.
Result: From hitting the mid 90s under full load with 100% fan speed, my GPU now manages full load with 50% fan speed in the low 70s. It's a wonderful feeling to have the card so quiet.