GPU temp gone up??

rafezetter · Post by **rafezetter** » Fri Sep 05, 2014 11:38 am

First off I've been building pc's over a decade, and doing moderate customizations etc so I'm used to doing equipment alterations, yet this problem is a new one to me, so I hope you guys might be able to have some suggestions.

I recently bought a gtx770, and noticed when running the valley benchmark the temps were around 71deg and holding steady, running standard cooler, standard ootb settings. Although inside tolerance, I want to overclock it so bought some gelid extreme paste (based on this review: http://www.tomshardware.com/reviews/the ... 16-10.html

I used to use the coollaboratory stuff, but my last lot ran out and as they say it IS a PITA to apply.

I was extremely thorough with cleaning the old gunk off, PLUS I gave the copper mating plate a polish, yet after applying the gelid paste the temps went UP, peaking around 81deg.

That's 10deg higher than before. The paste does not mention a burn in time required either.

yes fans were plugged in, yes heatsink array was properly bedded and tightened, yes I was sparing with the paste.

Any ideas?

Edit - I've just re-read that paste review and I annoyingly missed this snippet of text "While value-added pastes may work well with the CPU coolers they were bundled with, they may not behave well under the higher temperatures generated by powerful GPUs, and shouldn't be used with them."

having said that - anyone had any success with replacing the paste on GPU's for aircooled?

Vicotnik · Post by **Vicotnik** » Fri Sep 05, 2014 11:53 am

Redo the whole thing would be my suggestion. Might be the mounting is messed up, an easy thing to to. Check the imprint for any signs of problems.

Also, I really must question the decision to replace the paste with something "extreme". All the pastes out there perform pretty much the same. Replace the cooler if you're looking for better performance.

quest_for_silence · Post by **quest_for_silence** » Fri Sep 05, 2014 12:25 pm

rafezetter wrote:having said that - anyone had any success with replacing the paste on GPU's for aircooled?

Never replaced the stock paste just for the sake of doing that, so I cannot compare the relevant effectiveness, but when I mount aftermarket coolers on GPUs, usually I use the AC MX4 (and up to now I didn't get into troubles with it): maybe you could give it a try?

rafezetter · Post by **rafezetter** » Fri Sep 05, 2014 2:12 pm

I have redone, sadly with the same results.

To be honest this is the first time I've ever swapped out the TIM on a GPU card, but as I already had to order more TIM for the CPU and have a serious surplus (a 20g pot!) I figured the same "the stock paste that comes bundled with a CPU is so/so" could well apply to the GPU paste too.

I didn't think gigabyte would put a premium paste on their GPU's, but after re-doing it twice with the same result, leads me to think they have - a 10deg difference is quite remarkable, I'm amazed the difference is so high.

I guess maybe an email to gigabyte is next to see if I can get some more!

I thought about watercooling after the first result, but this latest PC build is an interim until Oculus Rift is released, then I think I'll build a monster PC to suit.

xan_user · Post by **xan_user** » Fri Sep 05, 2014 3:06 pm

tomshardware wrote:Different rules apply for GPUs; you need to pick the compound you use on your graphics card very carefully

http://www.tomshardware.com/reviews/the ... 16-20.html

and i always love to post this for people swapping TIM for performance reasons;
http://www.hardwaresecrets.com/imagevie ... mage=39371

bonestonne · Post by **bonestonne** » Fri Sep 05, 2014 4:25 pm

SPCR actually got a bunch of samples for a bunch of users a while back for Innovation Cooling Diamond 7. I used it on both CPU and GPU (at the time, 9600GT and Q9600). The results were very much in favor that the Diamond 7 stuff was really good compared to what I had been using.

So good in fact I've gone back and bought more a couple of times. I've put it in my new desktop, my laptops (when I've had to remove the coolers) and even my PS3 when I took it apart to keep it clean and running (fat first gen, known for heat issues). Very glad I did.

If your results are showing that you're not seeing different results even when re-applying, I would venture a guess towards the direction that there's another factor in play here. Has the card flexed significantly, or is it still straight and true? That can make a massive difference by itself.

I haven't done the GTX660Ti in my machine because there was no need, but if I had to, I would use the Diamond 7 on that as well. I've just had consistent good results from it, and that success has led me to just keep using it. I can't imagine the GTX770 is anything wildly special that wouldn't benefit from a better TIM, but I have found that laptops and GPUs are incredibly easy to over-apply. I had to basically re-teach my co-worker how to apply the stuff because he was always using way more than he should have, and he just didn't know better.

I'm not sure what I'd point towards without seeing the machine myself, things like this are generally not cut and try over the web, obviously. From what I've seen, Asus tends to cheap out the worst with stock TIM on their parts and computers, but Gigabyte, I can't say the same about, it seems to look like most other stuff AFAIK. I just know I took apart an Asus laptop earlier, and I had to use a plastic spudger to scrape off the crap because it was so dry and hard after the laptop was on for about an hour. The laptop had a bit of a dust issue, the it was less than 2 years old, so shouldn't have been that bad at all. As the "idiot check" I have to do from time to time, when remounting the cooler, did you miss any thermal pads, screws, or spacers?

rafezetter · Post by **rafezetter** » Fri Sep 05, 2014 6:42 pm

Thanks for the replies, @ Bonestonne - the heatsink was just 4 sprung screws for the main mount and 3 small ones for the ram section. All thermal pads were re-used, however I did notice they were compressed so I wonder if that might be a factor? However I thought the GPU temp was based on the chip - would the temps shown in the benchmark (and realtemp software) be for the whole card? If so then the rams might be getting hotter than they were.

The TIM I was careful to use very little, when I removed the heatsink to re-do it there was very little left on the chip as most of the excess had squeezed out, but coverage was even all over.

I apply it like tile adhesive; spread it very thin and not quite to the edge, then draw thin even lines in it, as you would get with a notched spreader, which removes some of the excess and gives the rest space to level out.

With a well polished but rigorously cleaned heatsink face, what gets left in the mating void is virtually nothing.

I'll look into the Diamond 7, I've read elsewhere it's recommended, but have not seen any testing results compared to other makes, so that's why I went with the Gelid. Can anyone recommend good thermal pads as well?

I've also found these which look promising: http://www.overclockers.co.uk/productli ... catid=2360

I could also re-use when I go up to a 780 later if needed.

xan_user · Post by **xan_user** » Fri Sep 05, 2014 8:07 pm

rafezetter wrote: then draw thin even lines in it, as you would get with a notched spreader,

sounds like a lot of time/handling/chance for foreign particles (maybe even air pockets?)to become involved. what are you "drawing" the lines in it with?

i put on the thinnest possible coat with a brand new razorblade and rotate the heatsink back and fourth a bit as i place it on.

rafezetter · Post by **rafezetter** » Fri Sep 05, 2014 10:00 pm

xan_user wrote:
rafezetter wrote: then draw thin even lines in it, as you would get with a notched spreader,
sounds like a lot of time/handling/chance for foreign particles (maybe even air pockets?)to become involved. what are you "drawing" the lines in it with?

i put on the thinnest possible coat with a brand new razorblade and rotate the heatsink back and fourth a bit as i place it on.

The lines are drawn with the applicator that came with it, and are end to end only so any air has a chance to escape as it's pressed down. Only maybe 6 or so, a few extra seconds at most.

As I said before, when the heatsink was removed the first time, the footprint of the TIM was continuous all over, no gaps or voids anywhere.

rafezetter · Post by **rafezetter** » Fri Sep 05, 2014 10:14 pm

Ok I've just read about this item: http://www.scan.co.uk/products/ek-sf3d- ... ion-cooler

yes you read that right a liquid nitrogen GPU cooler.... designed by a world class overclocker, but what caught my attention was this bit of text:

Additionaly, EK supplies Gelid GC-Extreme premium thermal grease for best results.

which is what I used, so... WTH?

an email to gelid is in order I think...

Vicotnik · Post by **Vicotnik** » Fri Sep 05, 2014 10:21 pm

rafezetter wrote:All thermal pads were re-used, however I did notice they were compressed so I wonder if that might be a factor? However I thought the GPU temp was based on the chip - would the temps shown in the benchmark (and realtemp software) be for the whole card? If so then the rams might be getting hotter than they were.

That's possible I think. GDDR5 gets quite hot and if that heat isn't transfered to the heatsink properly it could heat up the rest of the card. Do you have an alternative to reusing the pads? Maybe use paste for the RAM as well?

rafezetter · Post by **rafezetter** » Fri Sep 05, 2014 10:59 pm

Vicotnik wrote:
rafezetter wrote:All thermal pads were re-used, however I did notice they were compressed so I wonder if that might be a factor? However I thought the GPU temp was based on the chip - would the temps shown in the benchmark (and realtemp software) be for the whole card? If so then the rams might be getting hotter than they were.
That's possible I think. GDDR5 gets quite hot and if that heat isn't transfered to the heatsink properly it could heat up the rest of the card. Do you have an alternative to reusing the pads? Maybe use paste for the RAM as well?

I had read in many places using a thick gunk of paste instead of a pad was a v.bad idea, and a combination of the two, also just as bad.

I've looked around and found that there are thermal pads with a heat transfer of 7 W/mk - and the gelid extreme is rated at just 8.5.

(oddly in america there are thermal pads available by fujipoly with a rating of 17 W/mk!) http://www.frozencpu.com/products/17500 ... 0_WmK.html

maybe I should get a load shipped over and sell them here!

Abula · Post by **Abula** » Sat Sep 06, 2014 4:08 am

I havent replaced the paste on a standard or reference cooled card, or a aftermarket cooled card (factory own cooler), but i have done mulitple changes on gpu heatsinks, lately with accelero III and IV, and i have applied at least a couple of times to each and it has worked fine, although im not sure if it can be better... i mean the cooler itself drops the temps so low that im not too sure if i can do it better, but i do get to the same temps with the same load and ambient temp, so my assumption is that its ok, but i never really done intense testing into different ways. Personally im done with aftermarket gpu coolers, unless the new gens of gpus come with high temps or high noise due to this, then i would stick with whats on the market and not mess with the paste or anything.

bonestonne · Post by **bonestonne** » Sun Sep 07, 2014 12:25 pm

I've re-used thermal pads often, never had a problem with them.

Whenever I apply TIM to whatever it is I'm working with, I just put an appropriate sided blob on the CPU/GPU die, and tighten the cooler down as evenly as possible to ensure the TIM gets spread by the contact. I've never had good experience with spreading the paste prior to installing the cooler.

I also have to say your problem is a bit unique, I've never seen something get hotter after replacing the paste. I would give it a shot without spreading the paste, or look for some other contributing factor to increased heat. Fan unplugged, obstructed, anything like that. I've had plenty of situations where I couldn't remember plugging the fan back in, so I've torn laptops apart to check. I've also had moments when I forgot to plug them in and had to disassemble to take care of that. Sometimes doing something for too long makes you lazy towards certain really basic things.

xan_user · Post by **xan_user** » Sun Sep 07, 2014 1:23 pm

i replaced my tim on a gpu, and what was there from factory was really, really thick, not a pad, but close. the cooler was spring mounted, so i add 3 extra washers on the back to make up for the lost TIM thickness and tightened up the fit.

rafezetter · Post by **rafezetter** » Sun Sep 07, 2014 5:58 pm

Well an update - it's been several days now and no change - the card is still running hot... too hot IMHO for long term reliability. I've contacted Geil to see if they care to comment, but I'm now looking at some fairly hefty aftermarket coolers.

While it's not a top end card, this should still happily handle games a few years old (like Deus Ex human rev) without getting this hot.

GPU temp gone up??

GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??

Re: GPU temp gone up??