GPU thermal paste replacement
jojo -
Hello,
I recently purchased a second-hand RTX 2070 Super (MSI Ventus OC model), and I am satisfied with its performance. However, I noticed that the card gets quite hot, reaching a maximum of 87°C during a 30-minute stress test and around 82°C while gaming on very demanding games, and of course, the fans get loud and noisy.
Being quite particular, I would like to find a solution to lower the temperature by a few degrees and, at the same time, reduce the noise levels.
While doing some research on forums, quite a few people recommend replacing the thermal paste. Before I dive into this, I would like to get the community's opinion on whether it's worth it and if there will really be a significant gain.
Thank you in advance for your advice and opinions.
6 réponses
I'm back with updates for those who may be interested. So, after replacing the thermal paste, the improvement isn't extraordinary; however, I found that the old paste was quite dry, resulting in a -3°C change.
I then decided to use another method to try to limit the heat and noise from my card, undervolting, and the result was very effective, -11°C with the same performance and lower consumption.
To summarize:
- Before thermal paste replacement: temperature 87°C, fans 92%, max power 215W
- After thermal paste replacement: temperature 84°C, fans 92%, max power 215W
- After thermal paste replacement and undervolting: temperature 73°C, fans 85%, max power 171W
Hello,
The graphics card is mounted in a Silverstone Grandia GD09 case with 3 120mm side fans for intake and 2 80mm fans at the back for exhaust, plus a ventilation grille on the top just above the graphics card.
Hello, I think that with thermal paste, you should have gained more temperature improvement than that.
I bought a used GTX 1070 that, over time, heated up to 85° on any games.
I disassembled it, the original paste was completely dry, I changed the thermal paste and used
Arctic Silver MX-4 paste and I dropped to between 63/68° depending on the games of course.
Afterward, your card is more powerful than mine, so maybe that affects the temperatures.
Hello
The gain from changing the thermal paste is only proportional to the loss due to aging. Only if there was a big problem would you see a big gain.
There remains a very small difference if your paste is more effective than the original one.
The risk of taking it apart depends on the skill of the person; I would try it on mine, maybe not on a friend's, I would rather keep the friends.
Overclocking or undervolting done by someone who knows the subject and gets a good unit, it's almost normal that it pays off. The manufacturer has to set a configuration that suits all the cards made, there's bound to be a margin.
Not doing the optimization results in a lower cost price; doing it means selling the product more expensively for the manufacturer. And they can't provide valid values for all cards, otherwise they would have to set the value of the least performing one, which ends up being more or less the same for them.
Basically, they are doing industrial production in series, while you are doing custom handcrafted, numbered pieces.
Good result, you've gained 14° 45W and silence; that's just huge.
Hello,
Changing the thermal paste is like a trend for those who know everything.
For a CPU, we do this approximately every 10 years (unless you change the CPU), otherwise it's exaggerated.
Also, it may not be a paste but a silicone (Permanent) for the GPU.
So, in principle, you never have to replace it.
Hello fabul
I largely agree with you, 8 to 10 years but on the condition that a good quality thermal paste has been used, otherwise it can degrade before that.
For graphics cards, there is a risk in disassembling their cooling systems; it's not like a simple CPU cooler, it's really something to be done as a last resort.
Hello Flo88,
Indeed, I once broke a screw without even forcing it, goodbye to the Asus 9800 GTX+ DK :(
Replaced by a Zotak GTX 580
But by blowing too close and without holding the blades with a big vacuum cleaner (Shop Vac), I broke some blades, goodbye to the GTX 580 :(
I finally have an EVGA GTX 780 SC with a blower fan instead of blade fans, less likely to break, and I’m not taking it apart.
I add: for undervolting settings on a MSI RTX 2070 Super Ventus oc: 1920MHz@900mv
I push my analysis of undervolting a little further, here are the results from Time Spy on 3DMark, on the left the undervolted GPU and on the right the stock GPU. We can see that the score of the UV GPU is slightly higher than that of the stock setting GPU, even though the maximum clock frequency is a bit lower on the stock version at 1920MHz vs 1965MHz, which is due to the fact that the UV GPU consumes less, thus heats less, allowing the boost to stabilize at a higher frequency over a longer period (1852MHz vs 1816MHz).
It seems that GPU manufacturers do not optimize the output of their chips to the maximum, perhaps to save time on the production line.
In conclusion, undervolting your GPU is simply optimizing it for an optimal performance/power consumption ratio.