Discussion What thermal pastes are you using?
- Thread starter findingmyfeet
- Start date
CPU: Mainly Arctic MX-2 cause I decided it was a good idea to get a 65g tube in 2015. I might get through 10% of it by 2022.
GPU: Thermalright TF8
Can't remember why the MX-2, the TF8 kept coming up as recommended for GPUs. Non-conductive was a must (iMessy)
Related: does anyone use thermal pads? Used Thermalright 12.8W/mK recently on a 3200g build; thermals are OK, cost comparison to paste is not.
GPU: Thermalright TF8
Can't remember why the MX-2, the TF8 kept coming up as recommended for GPUs. Non-conductive was a must (iMessy)
Related: does anyone use thermal pads? Used Thermalright 12.8W/mK recently on a 3200g build; thermals are OK, cost comparison to paste is not.
I tend to use whichever high quality paste I can get ahold of a good amount of for an acceptable price. I abhor having a million tiny paste syringes, so after GC-Extreme supply dried up where I used to live I got a 10g tube of Kryonaut, and then a 10g tube of NT-H2 once I had used that up. I've been happy with each of these, though the Kryonaut is a bit of a pain to spread due to its stickyness and tendency to pick itself up as you're trying to spread it. GC-Extreme was probably the easiest to use out of those three, and the cheapest, so if I could get a huge syringe of that I would. Truth be told, thermal differences between pastes mostly aren't very noticeable in real life unless they are truly garbage tier stuff.
Do you mean conventional thermal pads (i.e. the thick, sticky, fragile type)? Or those CPU-oriented graphite sheet "pads"? I could see the latter being useful for anything where ease of maintenance or long-term untouched operation is desirable, but I don't have any builds where that has fit the bill. Normal thermal pads ... nah, I would never put that on a CPU. Expensive, fragile, often come in strips far too narrow for a CPU, and thermal transfer will be terrible compared to paste, both due to the lower conductivity and the much thicker interface. It would likely work on a low power CPU, but I don't see any reaosn to go that route, ever.
Thanks - I did mean both pastesMX-2 is also non conductive, don't worry
I used Thermalright Odyssey 1.5mm, so the former. It was left over from replacing the thermal pads on my GPU (thanks NVIDIA for cheaping out).
At 1.5mm this pad is objectively bad for CPU use; the fact that it was viable at all has made me start looking to get something more performant. Use cases would be quick-swap type testing, and systems I expect to make changes to in 6-18 months (low to medium TDPs).
Agreed cost is extortionate compared to paste; don't expect thermal transfer to match thermal paste but don't agree it's terrible, and not sure I agree with the other parts.
Gamers Nexus did some videos (watched them after your post, to improve my understanding). The numbers were better than I expected:
The chart on the left shows the Grizzly Hydronaut paste/IC Diamond pad delta; so e.g. at 95W you're looking at a ~1.2°C diff, at 150W you're looking at 2.21°C for a decently flat cooler and 2.88°C for an unlevel cooler.
Chart on the right is delta over ambient, because there wasn't one between the pads v paste. The absolute value doesn't really matter, more the difference between the methods; e.g. at 150W the (yellow) IC Diamond is worst, with the (red) Grizzly Carbonaut pad 1.02°C cooler, and the (blue) Grizzly Hydronaut paste 2.21°C cooler than the IC Diamond pad (the first chart is nicer at it just shows the relevant 2.21°C at 150W, but does not include the numbers for Grizzly's Carbonaut)
Chart on the right is delta over ambient, because there wasn't one between the pads v paste. The absolute value doesn't really matter, more the difference between the methods; e.g. at 150W the (yellow) IC Diamond is worst, with the (red) Grizzly Carbonaut pad 1.02°C cooler, and the (blue) Grizzly Hydronaut paste 2.21°C cooler than the IC Diamond pad (the first chart is nicer at it just shows the relevant 2.21°C at 150W, but does not include the numbers for Grizzly's Carbonaut)
The GPU block I bought came with a couple strips of (very, very thin) Kingbali pad for the VRMs.
Dunno if those pads are fine or too poor quality ?
For the GPU & memory chips proper, well I have some rather fresh MX-4 around...
Dunno if it's alright, but I might give TF8 a first try.
This GPU will see some workload, I hope for a reliably long-lasting solution, like ~3 years of peace would be nice.
TL;DR is it maybe better to put TF8 on everything on the GPU and forget about VRM pads ?
Dunno if those pads are fine or too poor quality ?
For the GPU & memory chips proper, well I have some rather fresh MX-4 around...
Dunno if it's alright, but I might give TF8 a first try.
This GPU will see some workload, I hope for a reliably long-lasting solution, like ~3 years of peace would be nice.
TL;DR is it maybe better to put TF8 on everything on the GPU and forget about VRM pads ?
Pads on low thermal loads like VRMs and VRAM are absolutely fine. Often they are also entirely necessary, as coolers are engineered to contact the GPU die first and best, with the tolerances for everything else being much looser. Anything else, like the cooler hitting the VRAM before it properly contacted the die, could lead to catastrophic results after all. So pads are used to fill the gaps inevitably left between these secondary/tertiary heat sources and the cold plate, as paste simply can't build up very thick in a stable way. And engineering a cooler with paste-thin tolerances for all available components is essentially impossible due to manufacturing variance.
If the included pads are thin, as long as they are making proper contact (remove the cooler after mounting and check for dents/marks), that is ideal. Thicker pads = more thermal resistance, so thinner pads are always better. That's also a large part of why paste performs better than pads - you can never squeeze a pad as thin as you can a paste. Ideal thermal paste thickness in situ is in the range of a few micron, after all. The less interface material there's room for between the heat source and cooler, the better.
VRMs also produce so little heat as to make the quality of the thermal pads pretty meaningless (unless they are deteriorating, far too thick, or you are pushing the VRM very hard). I've used the inlcuded thermal pads on my EK water block since I installed it in 2016, including removing and reapplying them during a complete block teardown and cleaning, and they work perfectly still. IIRC I first installed that using GC-Extreme on the die+HBM (which needs paste and is the same height as the die, being on the same package), and that was still working perfectly when I tore it down in late 2020 - I just wanted to check and clean the internals of the waterblock - if I could have done that and re-used the paste I would have.
While building my ITX early this year I found the Phobya NanoGrease Extreme which has 16W/m which was quite higher than most of the other pastes I have seen/used so far.
I only used it on the CPU so far and I had no issues. But I do not really run anything really demanding or stress test my computer.
I only used it on the CPU so far and I had no issues. But I do not really run anything really demanding or stress test my computer.
W/mK figures can't really be compared between manufacturers, as there is no accepted standard for measuring these values. As such, real-world testing is the only trustworthy way of comparing pastes. That being said, the vast majority of thermal pastes perform within a very narrow range, so differences aren't huge. Anyrhing that seems to perform decently and is easy to spread and clean up is good enough IMO.
Ah, I assumed/hoped that this was standardized. Not very helpful then. So I guess as long as nothing catches fire its "good".
Pretty much, yesAh, I assumed/hoped that this was standardized. Not very helpful then. So I guess as long as nothing catches fire its "good".
Similar threads
Discussion
Looking for a new case
Discussion
CPU getting too hot! help
