New GPU, should I replace thermal paste immediately?

[lexluthermiester]

If you break the seal on the screw that holds the bracket in place it voids your warranty.

Not anymore, at least not in the USA. "Warranty Void" stickers and policies were deemed unlawful a couple years ago.

 

[Wolvyreen]

Not anymore, at least not in the USA. "Warranty Void" stickers and policies were deemed unlawful a couple years ago.

UGH! I want to move to the US!

 

[RealNeil]

Thanks All I'll leave it as is for now.

There you go! Smart move.

I think that GPU makers are paying close attention to TIM application these days, making the possibility of substandard application statistically moot.
I ~could~ happen, but it's unlikely to,.....especially with premium cards.
Consider that they've gone through considerable lengths to design their products to work properly under normal conditions. (and slightly overclocked)
This relies on the whole (heatsink, fans, and interface to the GPU core and memory/power handling substrate) system design being done right. TIM application is a critical part of the whole.
They can fudge it a little with low-performance GPU's, but even they're built to a designed spec that they adhere to.

If there are temp issues with a new card, it's best ~not ~ to crack it open, and better to return (or maybe RMA) it.
I sold a perfectly good GTX-1070 to someone and it quit working. He sent it back to me and I RMA'd the card since it was under warranty.
Gigabyte refused to fix it because he had repasted the card.

 

[freeagent]

I scraped off the stuff they put on my 970 and 980 and replaced it with my own.. Maybe a 2-3c difference.. maybe.. probably closer to one or two. When the load is off and temps drop is where I see the biggest change, I think.. and idle temps are lower for sure. my 970 and 980 are sitting at 25 doing nothing in quiet systems. If it were me, I would just leave it for a couple of years.

 

[64K]

I took apart a GTX 560 TI that I didn't have any need for one time out of curiosity and it looked like EVGA had put way too much TIM on the GPU. It looked like a sloppy mess to me but it probably wasn't conductive TIM.

 

[EarthDog]

I took apart a GTX 560 TI that I didn't have any need for one time out of curiosity and it looked like EVGA had put way too much TIM on the GPU. It looked like a sloppy mess to me but it probably wasn't conductive TIM.

If it was conductive, it likely wouldn't have worked spilling out all over the caps surrounding the die.

This is what I have seen in general are liberal applications of TIM. The other situation I noticed is that some were hard as a rock. Those, when taken apart, need reTIM'd, the others, I clean it off, put some paste back in the middle and put the HS back on.

 

[blued]

Only card I did it to was GTX 470, hot as hell (90c) load. Replaced TIM, achieved 2c improvement. Original TIM sloppy but generously applied. Result hardly worth it.
And to anyone thinking of doing it, DO NOT assume its done the same way as CPU, with a thin layer of paste. It needs a lot more, especially if you have heat pipes.
Manufacturing application may be sloppy, but generally it is effective. You will not benefit much from re-application.

 

[TheoneandonlyMrK]

Well, your temperatures would indicate if it is substandard or improperly applied. Does it happen? Yes. But unless you see evidence it is inadequate, just leave it be for a couple years.

While I understand the point and principle, I think given what the OP has said, He will OC and is after what gains can be had, I personally would definitely consider a re-tim , it is not needed or necessary,
but putting conductonaught or some other very high-end tim on can get you between 5-10 degrees(10 for me) and therein gain impressive sustained bost gains.

Given the Boost/auto OC of both GPU manufacturers, more cooling is always better.

this would optimize his performance and enjoyment For the next few years.

needed no, but beneficial, possibly.


just consider can you do it safely, i see no risk since i am very experienced but you may be different, liquid metal Tim's are better but are very risky and tense to use for the first few times since more care is required but as other's told me putting silicon tim all over the bits you don't want liquid metal on(caps and stuff around chip but on die) works astonishingly well so with the right technique its easy, my main rigs main bits have liquid metal on.

 

[dgianstefani]

As has been mentioned before - workstation cards can see double digit temperature improvements just from a re-TIM.

Gaming cards won't see as much of an improvement - but contrary to what some people have said in this thread - many manufacturers allow repasting of GPUs without voiding warranty.

I would avoid conductive TIM such as conductonaut on GPU - no issues with CPU but GPU unless done right/perfectly can lead to issues - personal experience. Also with GPU the die size is large enough that the improved thermal conductivity of liquid metal isn't as important as on a small CPU die.

If you're going to do this, take extra care when disassembling and reassembling your cooler - often there are supplementary RGB cables, or in the case of the Founders edition 2060 - copious amounts of glue.

Use a high quality paste like Kryonaut and you should see a minor temperature decrease, but more importantly a more noticeable sound pressure dBA decrease as the fans won't have to work as hard.

TLDR it's worth it if it doesn't void the warranty of your particular card and the card itself is simple to disassemble/reassemble.

 

[EarthDog]

As has been mentioned before - workstation cards can see double digit temperature improvements just from a re-TIM.

Gaming cards won't see as much of an improvement - but contrary to what some people have said in this thread - many manufacturers allow repasting of GPUs without voiding warranty.

I would avoid conductive TIM such as conductonaut on GPU - no issues with CPU but GPU unless done right/perfectly can lead to issues - personal experience. Also with GPU the die size is large enough that the improved thermal conductivity of liquid metal isn't as important as on a small CPU die.

If you're going to do this, take extra care when disassembling and reassembling your cooler - often there are supplementary RGB cables, or in the case of the Founders edition 2060 - copious amounts of glue.

Use a high quality paste like Kryonaut and you should see a minor temperature decrease, but more importantly a more noticeable sound pressure dBA decrease as the fans won't have to work as hard.

TLDR it's worth it if it doesn't void the warranty of your particular card and the card itself is simple to disassemble/reassemble.

I agree with most of this. However, a couple of C isn't going to change the fan speeds unless it is sitting on the edge already... which can also be accomplished by adjusting the fan curve up a couple of C.

Why would workstation cards see different improvements when the TIM and application methods are the same?

 

[dgianstefani]

It's not the couple of degrees that changes the fan speeds. Change how you think about it.

It's actually decreasing the temperatures by more than a couple of degrees at the same fan speeds so you'll see a couple of degrees lower temperatures and lower fan speeds for a quieter card.

Workstation cards see more improvements because their coolers aren't as overbuilt as on gaming cards, so the increased TIM performance matters more. It's why repasting your laptop with Liquid metal as I have done before several times will lead to much greater improvements in temperatures than doing the same thing on a desktop with a good cooler.

 

[Hardcore Games]

I repaired a Gigabyte GTX 750 which was thermally throttling. I used MX-4 and the card now is stable and available for use.
The card has 2GB GDDR5 which gives it more memory bandwidth which at least does not bottleneck the ASIC.