Liquid metal TIM for 4090?

[magellan]

I tried liquid metal TIM for my 1080ti, it reduced temps enough for me to squeeze 26Mhz. more out of it (to 2164Mhz.) and, more importantly, stopped all throttling. Would it be worth it to use liquid metal on a 4090? I'd figure w/the greater power draw it might make more diff. w/the 4090 and maybe prevent throttling. If no one's tried it and no one recommends it though I'm not going to be the first.

 

[EarthDog]

Depends on the card (be it 4090 or 4060) and it's heatsink/how it's cooling. My 4090 has an AIO so temps never get above 63C. Dropping a couple of C isn't worth it in this case.

As far as worth it, that's up to the user. But it's the same risk involved as on your 1080 Ti. I bet it will drop temps several C, but not sure if it will stop throttling due to temps. The biggest problem isn't with the TIM, but getting the heat out of the 'tiny' core.

 

[magellan]

ED, what about the Honeywell PTM7950 phase change pad discussed in another thread? I've read it's almost as good as liquid metal TIM and the amazon reviews for it on the 7900 XTX are all positive.

 

[Flamethrower1972]

How did you get that high of an OC on a 1080 TI GPU? I got 300Mhz on the memory but the gpu 30Mhz more anything after that crashed Maby I need to do something with the voltage.

I tried liquid metal TIM for my 1080ti, it reduced temps enough for me to squeeze 26Mhz. more out of it (to 2164Mhz.) and, more importantly, stopped all throttling. Would it be worth it to use liquid metal on a 4090? I'd figure w/the greater power draw it might make more diff. w/the 4090 and maybe prevent throttling. If no one's tried it and no one recommends it though I'm not going to be the first.

 

[EarthDog]

ED, what about the Honeywell PTM7950 phase change pad discussed in another thread? I've read it's almost as good as liquid metal TIM and the amazon reviews for it on the 7900 XTX are all positive.

It's less risk, not being metal, but otherwise still the same with air where you're almost always fighting thermal throttling. Better TIM helps, but ultimately, getting that much energy out of a tiny die proves to be difficult on air.

 

[Flamethrower1972]

I have heard that using liquid metal is dangerous to apply and that you have to do it just right.

 

[magellan]

Liquid metal is a PITA to install, the conformal coating and the fact the liquid metal is difficult to spread are the worst aspects, but for my 1080ti flashed w/the Asus XOC VBIOS it was the last piece in the puzzle to get it to 2164Mhz. I also had the 1080ti mounted to two massive, shrouded 127mm fans (one 38mm thick, the other 50mm thick) and the entire backplate was covered in thermal tape. I think the Hwbot record for any 1080ti is 2190Mhz.

 

[Kenrou]

Yeah, it takes a little bit of patience to get it just right, but it's usually worth it

 

[yoadknux]

When I was thinking what solutions are available for my TUF 4090, liquid metal was never an option. It's just too much risk for such little gain.
As for PTM7950, it's not meant to be "better all around", I think Kryonaut Extreme resulted in slightly lower temperatures upon re-paste, it just degrades within a short period (weeks). PTM7950 starts off slightly worse but supposedly doesn't degrade.
If you want something that is close to liquid metal but less messy to apply, there's Thermal Grizzly Kryosheet

 

[magellan]

@yoadknux
Would the Thermal Grizzly Kryosheet have more thermal conductance than the PTM7950? It wouldn't have any degradation or pump-out issues would it?

Although my Gigabyte 4090 Gaming OC is running warmer than when FOB, the highest I've ever seen for my hot spot temp is 86.1°C, while the GPU was at 73°C and the ambient room temp was 70°F.

I'm reluctant to do much w/the Gigabyte Gaming OC 4090 until my warranty expires, unfortunately that's not until 2026

 

[yoadknux]

@yoadknux
Would the Thermal Grizzly Kryosheet have more thermal conductance than the PTM7950? It wouldn't have any degradation or pump-out issues would it?

Although my Gigabyte 4090 Gaming OC is running warmer than when FOB, the highest I've ever seen for my hot spot temp is 86.1°C, while the GPU was at 73°C and the ambient room temp was 70°F.

I'm reluctant to do much w/the Gigabyte Gaming OC 4090 until my warranty expires, unfortunately that's not until 2026

Both Kryosheet and PTM7950 are relatively new products so there are not a lot of reviews on them (compared to thermal pastes, which people have been using for decades). Kryosheet in theory should be the best overall solution in terms of temperature and longetivity, but I haven't seen many people using them.
Regarding GPU temps, the first thing to check is how much difference in temps you get when running the same benchmark with side panel on vs side panel off. There are many small things you can do that will lower the temperature by one degree here or there and they add up. For example vertical mounting really helps with some GPUs. Removing PSU shroud and bottom intake can help with pass-through GPUs.
13c delta between Hotspot and core is normal. Look to optimize airflow and open the card only as a last resort.

 

[EarthDog]

86.1°C, while the GPU was at 73°C and the ambient room temp was 70°F.

It's like you were speaking "PC" English (Celsius), then went to French (Fahrenheit).

*kicks stones walking away having to figure out what 70F is in C....

 

[magellan]

The only accurate room thermometer I have is graduated in Fahrenheit. Would the motherboard monitor temp. in HWiNFO64 be a better way to calculate delta T temperatures? At least it's in celsius, but I guess it really isn't reporting the internal temp. of the air inside the case either is it?

 

[EarthDog]

When I was actively into that type of data gathering, I put a temp sensor in front of the front intake fan so I know the temperature of the air that's being used to cool the machine......it reports in C.

I don't believe using the motherboard sensor that's god knows where is a good barometer, no.

 

[yoadknux]

The only accurate room thermometer I have is graduated in Fahrenheit. Would the motherboard monitor temp. in HWiNFO64 be a better way to calculate delta T temperatures? At least it's in celsius, but I guess it really isn't reporting the internal temp. of the air inside the case either is it?

well you could monitor it as you see fit but you have to understand the moment you wrote that message I had to Google "70F to C"