Got a bit of an issue with my GTX670's and Custom coolers.

[Mkilbride]

So, I put these on like 8 months ago, at the time, some of the heatsinks fell off, but only on the voltage regulators, which some said was ok...and they did indeed run alright. Temps higher than they should be, but nothing terrible.

However, they've been getting worse lately, reaching 70c, so that my cards downclock, and instead my drivers crash, then recover.

So, I got off my lazy ass, ordered some thermal 3M tape, unscrewed them, made sure all the heatsinks were on firm and it looked great. I swapped out the thermal paste, as it was old, and I read you should always change it when you take off a cooler. (Twin Turbo II here for my 670's, FTW edition)

So, used my Tunique TX Ceramic stuff for the first GPU; that ran out, so I used AS5 for the second GPU.

I applied the thermal pate to the heatsink...using a credit card to smooth it over into a thin, even line. I then put them back together...put them in my PC.

I was greeted with something nice. Both GPU's were idling @ 23c, where previously they would idle @ 30c or so. A marked improvement.

So I opened up Firefox, then saw a spike to 36c suddenly...I ran a small benchmark. My second GPU...went from 23 to 60 in a second so I shut it down. I disabled it. My first GPU took abit longer to heat up, but eventually the temp reached the 60's as well, though in the GPU Benchmark @ 99%, it goes for 51-52c, but in say, the Resident Evil 6 Benchmark, it hits 61c, quite a difference, 10c, both @ 99% load

So I'm trying to figure out what the frack happened here. All the heatsinks are on now. Secure, firm, and properly done.

I'd think it was the thermal paste; but then the strangely lower idle temps confuse me here. I'd think it was the tension holding the heatsink to the GPU, perhaps not tight off and having a little lift off...but it's as tight as it can go.

I ran out of Thermal Paste, so I ordered some http://www.amazon.com/gp/product/B0056QHJ8E/ref=olp_product_details?ie=UTF8&me=&seller=

Because I could get it by Saturday with prime for only 3.99$

Looking at videos...been a long time since I applied thermal paste, it seems the riced sized option would be best? Did I somehow apply TO much thermal paste? But this confuses me.

My First GPU, seems to be doing better than before. Overall, more stable temps. However, it gets hotter, faster, overall. My second GPU is completely unusable until I can take it out and examine it again.

Figured I'd ask here. Kinda annoyed. I took two hours to do something I'd been putting off for nearly half a year, feeling very confident when I looked down at them, thinking "This'll solve my issues"

To getting into Windows and seeing a 7c drop in idle temps...then to see my second GPU nigh unusable and my first getting hotter faster.

 

[Josh Weston]

Too much thermal paste can indeed cause more heat. Basically, you want a thin even layer. This is used to bridge the gab between the metal cap protecting the chip and the heatsink's base. The more thermal paste used, the more energy it takes to work through the paste and into the heatsink - the thing which is actually doing the work.

You may also find that putting the thermal paste on the chip and using the pressure of the heatsink to spread it out is a better option than spreading it out yourself. Spreading it yourself can cause air bubbles to form in the paste which then insulates the chip - not good.

61*C when running a GPU intensive benchmark is extremely good though. Try running Unigine: Heaven a few times and see what kind of temperatures you get. At 61*C you shouldn't be seeing any throttling or crashes whatsoever.

 

[Mkilbride]

It's still a little high for a non-oc'ed 670 with a custom cooler. Granted; it is hitting 1300MHZ just by Kepler boost. I also understand what you said, I know that.

I said I spread a thin, even layer. Paper thin. I plan to do the chip next time with a chocolate chip sized grain.

And that's 61c, sure, but got to consider this - that is with one GPU.

In SLI, you can throw 10 extra degrees onto that. THe extra power being drawn through the SLI Brdige causes a severe temp spike.

Before my second GPU, my first 670, the one that is hitting 61c, used to hit 49-52c @ 99% in a game like BF3.

Thing is, my two x16 slots are close to each other, my first GPU BARELY has any breathing room, an inch at best, not even that. Terrible air flow even with a box fan next to it.

I -can- solve this by putting my second GPU in a 4x slot...but yes, the performance loss...which is crap, cause my mobo supports 8 x 8 x 8, but reading INTO IT, it only supports that in Tri-SLI...if it's Dual GPU, it's 8 x 8 x 4, the last PCI-E for some reason becomes only x4, and I never found a way to change this. =/ x4 is a significant performance drop for me, when I had it in that slot, I lost about 7-8 FPS.

Still, the fact they were running earlier today, broken thugh they may be, and I could enjoy them, to now this...is beyond frustrating.



OH:

Another interesting tid bit? My second GPU, completely disabled, rising in temp when my first GPU is being used, even though my second GTX670 is disabled in NVCP and is seeing 0% usage. Around 40-45c @ 0

 

[alienware]

Hmm apparently these coolers (like most of Arctic's crap) can bend your PCB causing the cooler to make poor contact with the GPU.

http://www.hardwaresecrets.com/article/Arctic-Accelero-Twin-Turbo-II-VGA-Cooler-Review/1603/9

I would copy and paste the last bit but I can't, but in short it says that over time the cooler can bend the PCB.

TBH the issue you describe (it being good for a while then getting progressively worse) could be that it has bent your PCB.

Arctic have always been incredibly flaky tbh. Their coolers are a royal PITA to fit and their heatsinks always fall off. Their latest way of trying to keep them on is to use thermal epoxy (so glue, basically) which you can't get off once you use it.

Which of course does nothing for a warranty apart from to completely void it.

I don't understand why people don't just buy a custom cooled solution in the first place

 

[Mkilbride]

So I re-applied some new thermal paste, a dab, screwed on, but it idles @ like 39c and quickly heats up, even when not being used.

I don't get it, looks identical to the next one.

I trried more, no effect. I tried tightening it more, or less. It's weird because, it worked before I did all this.

I let my second GPU run and it hit 79c! within 15 seconds.

Gah, I don't get it, I've tried little thermal paste, and alot. My first GPU, which runs amazingly, I used alot. My second GPU, I used alot the first time. High temps. Second time I used a rice sized grain. High temps.

Third time I used about as much as my first GPU. As seen above. It appears to be some kind of issue with it not making contact, I'd guess? But it worked before and I screwed it exactly the same.



EDIT: Found a semi lose screw on the plate that attaches to the PCB; re-tightened it and temps are fine again. Crazy idle temps of like 22c. I'm worried I put to much paste now, but I'm kinda iffy on taking it all apart now that it works. Also thinkng of using my new paste on the first GPU, but eh.

Got some burn in for the second card; but it runs alot cooler than the first. Obviously...airflow.

Not happy with this situation.

My first GPU; is obviously hotter again now that it has reduced airflow, instead of 40c 99% load, it's up to 55-56c. 15c difference because the lack of air flow. God damned; that's unacceptable.

 

[Charliew87]

So I re-applied some new thermal paste, a dab, screwed on, but it idles @ like 39c and quickly heats up, even when not being used.

I don't get it, looks identical to the next one.

I trried more, no effect. I tried tightening it more, or less. It's weird because, it worked before I did all this.

I let my second GPU run and it hit 79c! within 15 seconds.

Gah, I don't get it, I've tried little thermal paste, and alot. My first GPU, which runs amazingly, I used alot. My second GPU, I used alot the first time. High temps. Second time I used a rice sized grain. High temps.

Third time I used about as much as my first GPU. As seen above. It appears to be some kind of issue with it not making contact, I'd guess? But it worked before and I screwed it exactly the same.



EDIT: Found a semi lose screw on the plate that attaches to the PCB; re-tightened it and temps are fine again. Crazy idle temps of like 22c. I'm worried I put to much paste now, but I'm kinda iffy on taking it all apart now that it works. Also thinkng of using my new paste on the first GPU, but eh.

Got some burn in for the second card; but it runs alot cooler than the first. Obviously...airflow.

Not happy with this situation.

My first GPU; is obviously hotter again now that it has reduced airflow, instead of 40c 99% load, it's up to 55-56c. 15c difference because the lack of air flow. God damned; that's unacceptable.

Everything under the keppler downclock at 70-71c is more than acceptable. These cards dont break before well over 100c. 55c load temps are incredible.

What the hell kind of fanprofile are you running? At 40% my FTW card runs at 30ish C idle. My max fan % is 60c and then Ill hit 75c.

The 13mhz downclock really doesnt bother me since the thing upping the temps, the overclock, is far beyond 13mhz.


Downclocking 50mhz so your card doesnt downclock itself 13mhz isnt optimal IMO.

 

[S_I_N]

seeing as you have re-applied paste several times leads me to think your issue may be with the thermal tape you replace on the parts that fell off. I'd get new better thermal adhesive and redo the entire card. Also as AlienAX said check for card warpage. Specially around the core.

 

[Mkilbride]

Everything under the keppler downclock at 70-71c is more than acceptable. These cards dont break before well over 100c. 55c load temps are incredible.

What the hell kind of fanprofile are you running? At 40% my FTW card runs at 30ish C idle. My max fan % is 60c and then Ill hit 75c.

The 13mhz downclock really doesnt bother me since the thing upping the temps, the overclock, is far beyond 13mhz.


Downclocking 50mhz so your card doesnt downclock itself 13mhz isnt optimal IMO.

As I said, I bought a 40$ cooler for the cards, I run them @ max fan profile and they're pretty much silent, designed to be that way. I ran the Valley Heaven Benchmark for 10 minutes and hit 63c max on my first GPU with no air flow and about 57-58c max on my second.

 

[alienware]

Everything under the keppler downclock at 70-71c is more than acceptable. These cards dont break before well over 100c. 55c load temps are incredible.

What the hell kind of fanprofile are you running? At 40% my FTW card runs at 30ish C idle. My max fan % is 60c and then Ill hit 75c.

The 13mhz downclock really doesnt bother me since the thing upping the temps, the overclock, is far beyond 13mhz.


Downclocking 50mhz so your card doesnt downclock itself 13mhz isnt optimal IMO.

Kepler is a lot more prone to heat problems than Fermi, though. Firstly it's drastically reduced in a nm sense so making it run at 106c for a shut down would be far more devastating than doing so on Fermi.

Not only that but the die size is so much smaller on the 104 that you're not even spreading that heat over a bigger surface.

I run a pair of Jetstream cards and I've never seen the hotter of the two go over 60c. It may have even been lower tbh. What I do know is that on a stock fan profile in the Nvidia drivers my fans never spin up, nor do they need to.

 

[Charliew87]

Kepler is a lot more prone to heat problems than Fermi, though. Firstly it's drastically reduced in a nm sense so making it run at 106c for a shut down would be far more devastating than doing so on Fermi.

Not only that but the die size is so much smaller on the 104 that you're not even spreading that heat over a bigger surface.

I run a pair of Jetstream cards and I've never seen the hotter of the two go over 60c. It may have even been lower tbh. What I do know is that on a stock fan profile in the Nvidia drivers my fans never spin up, nor do they need to.

Then youre one of the lucky few. The GTX670 FTW fan is a good one and depending on stock keppler boosts the temps seem to be 65-90c with stock profiles.

When running 3d I actually hit 100c on my card, hence the reapplied thermal paste etc. Now that Ive bought a backplate and reapplied thermalpaste I run my fans silent and hit 75c tops. What happens is that my card downclocks from 1280-1267mhz, which isnt really noticable. Im thinking of removing my 140mm side-fan and going SLI tho, and this thread is scaring the living **** out of me.


Anywhos, Im just trying to say that Ive never actually read a review that puts one of these cards at under 60c during load. Thirdparty coolers in all their glory but its just so beautiful with the stock cooling

 

[alienware]

Then youre one of the lucky few. The GTX670 FTW fan is a good one and depending on stock keppler boosts the temps seem to be 65-90c with stock profiles.

When running 3d I actually hit 100c on my card, hence the reapplied thermal paste etc. Now that Ive bought a backplate and reapplied thermalpaste I run my fans silent and hit 75c tops. What happens is that my card downclocks from 1280-1267mhz, which isnt really noticable. Im thinking of removing my 140mm side-fan and going SLI tho, and this thread is scaring the living **** out of me.


Anywhos, Im just trying to say that Ive never actually read a review that puts one of these cards at under 60c during load. Thirdparty coolers in all their glory but its just so beautiful with the stock cooling

That card is basically a cut down 680. The PCB cooler and etc are all from the 680. I really wanted to go with those as the power connectors are on the ends but I got such a good deal on mine I couldn't refuse tbh.

Whilst my coolers are regular 670 length they are triple slot and have a lot of pipes and fins. Mind you, it's helped by the fact that there's a 120mm fan sitting right in front of them and a duct that compeltely seals them in with that fan blowing on them.

I also live in a pretty cold house (non insulated 1930s farm house) so that's the most likely cause :lol: