EVGA GTX660 Temperature issue

[LKnight]

Hello, I've been a reader of this forum anonymously for almost six years, since I came only for knowledge and updates on technology (you guys rock).

Now, I have a issue that is simple, but maybe a lack of experience on the subject might have hindered me on solving this one.

My friend disassembled his GTX 660 for TIM replacement, the temperatures before were in the high 70's to mid 80's, as he replaced the stock TIM with standard white TIM, the temperature skyrockets to high 90's at the first glance of load (it rises in a VERY, VERY short timespan as if there is little to no contact with the heatsink).

Obviously I thought he must have applied TIM in an incorrect way, got the card myself, applied the standard TIM to no effect: skyrockets just as well.

I had a tube of VERY old TIM that came with my A50, applied what was left (just a drop) and the temperature got under control @ high 80's, 87 to be specific, which was way higher than it was before he even touched the card, but at least was under control at 15 minutes of furmark.

Then I thought to myself: It must be the viscosity of the dried out TIM. Got myself an tube of MX-4, applied it and got worse, a little better than the standard TIM, but gets to high 90's just as well.

Tried Akasa 455 (known to be bad) and it was bad as supposed, tried a local brand for silver TIM, got it to 100 degrees in 4s of furmark.

Note that ALL TIM's that I have are in good standing (except for the dried out A50 stock TIM), the standard TIM has been applied in everything I own and temps are great, the MX4 is applied to my OC'd X3 720 and is fine (splendid actually, got a 16 degree drop from the A50 dried TIM).

Don't know what to do, tried every paste I have (and acquired) in every spread method I know and heard about, contacted EVGA RMA and they told me to do what I already did (apply high viscosity TIM).

Sorry for the long post, but I tried to include everything that could possibly help you to help me.

Thank you for your time.

 

[Ketchup]

I reapplied TIM to my Gigabyte card, and I don't remember the temps, but they seemed to be reasonable. I'll try to check this weekend.

What speed is the GPU running?

 

[Ketchup]

I looked at my temps I recorded a few months ago, and I was hitting about 60 in 3DMark. Which model card is it? The speed it ships with will have a factor on the temps you are seeing, not just the thermal material.

 

[LKnight]

http://www.evga.com/articles/00700/images/hero/660_2GB_sm.png

This is the model, I believe it is the reference model.

 

[denis280]

How is the fan on the card?? and the thermal paste.just apply the size of a pea in the middle.

 

[LKnight]

The fan is working properly: quiet while on idle and turns into a jet (actually there is a very hot stream of air on the card's exhaust when on full load) when things get hot.
I applied the thermal paste in every way, the small dot in the middle is the standard and the first one I tried, no change from small dot or spread, line, big dot...

 

[Ketchup]

It sounds like the temps your friend was getting to begin with average for that model, so once you get it back to that you should be fine.

Did he say that there was a problem with it?

 

[LKnight]

Nope, all he wanted was to replace the stock TIM, as it was a little bit dried out.

I do want to get to previous temps, but doesn't matter what TIM I use, it seems like it doesn't make contact with the heatspreader.

The problem is I have never faced a problem where doesn't matter what TIM I use (other than a dried out TIM, which actually held temperature in place), how much I tighten a screw or anything else: stuff doesn't matter, temperature keeps out of control.

Still looking for a silver lining.

 

[Ketchup]

I wonder if you are going too tight. If you go too far, it could be actually pushing the material off to the sides.

I also put a tiny drop on the heatsink and spread it just to cover any imperfections. I doubt the stock heatsink is perfect in this regard.

 

[Virgorising]

I wonder if you are going too tight. If you go too far, it could be actually pushing the material off to the sides.

I also put a tiny drop on the heatsink and spread it just to cover any imperfections. I doubt the stock heatsink is perfect in this regard.

I think this is very smart, and so possible. Not like wood joinery where U tighten the clamps fully, but enuff glue remains to make the bond sound.

 

[corkyg]

Hmmm! The purpose of the paste is not to create a bond - it is to fill in the microscopic pores present in all metal surfaces. The thinner the better in order that the two surfaces can transfer heat more easily. There should never be residue left on the outside edges. Sounds to me like too much paste not spread thin enough. The edge of a credit card works well for this chore.

If replacing old paste, the old dried up stuff should be thoroughly removed so as to reach bare metal. Anything left on will detract from heat transfer and result in higher temps.

 

[Virgorising]

Hmmm! The purpose of the paste is not to create a bond - it is to fill in the microscopic pores present in all metal surfaces. The thinner the better in order that the two surfaces can transfer heat more easily. There should never be residue left on the outside edges. Sounds to me like too much paste not spread thin enough. The edge of a credit card works well for this chore.

Who ever said the purpose of thermal paste was to create a bond? Not I, not Ketchup79. What he envisioned and shared as a possibility, I think, is one.

In wood joinery, the choice of method, the goal....is always maximizing the surfaces for GLUE. You choose yr joinery strategy to accomplish this, be it biscuits, mitering, daddo, rabbet, dovetailing. All of them are meant to increase the surface for the glue. Then you clamp, let the glue dry, and that creates a bond.

I said using thermal paste is unlike that.

It may well be possible that tightening the heatsink too much could defeat the purpose of using thermal paste to more perfectly conduct, given the imperfect realities of the respective surfaces in question. Someone could, in fact, squeeze too much of it out, yes?

Good thermal paste is clearly not supposed to dry. If and when it does, then it becomes useless.

 

[Drummerdude]

So has the problem been solved?

 

[LKnight]

I tried everything that was suggested before it was suggested, I guess I could have been clearer on my posts.
It remains as if there is little to none contact with the cooler, the heating pattern is very similar to computers I've repaired that had loose coolers on their processors, the difference being that it isn't.
Anyway, I've got a little bit more on my plate than just this card, I will get to it later.

Still grateful for the help, though.

 

[Fardringle]

Make sure the heat sink is actually getting good contact with the GPU. If you take it apart to try again, clean off all of the existing thermal paste then reattach the heat sink without any new paste at all. Look sideways under the heat sink across the top of the GPU to see if you can see any light at all between the GPU and the RAM chips and the heat sink. If so, it is not getting reattached properly and you need to check the mounting screws to make sure they are aligned properly and to make sure there isn't anything stuck on the card or the heat sink getting in the way.

Of course, once you get the mounting worked out, then put some thermal paste on the GPU and reassemble it properly.