i totally agree with MX4 but not with MX5
i have GC Extreme, NT H2, Kryonaut and Kryonaut Extreme.
and even on a GPU Arctic MX5 beats Kryonaut Extreme (3090 and 6900XT tested so far)
Very interesting results!
I tested Arctic MX-5 on both my R9 290X and my MXM laptop GTX 1070 (230W TDP mod), and in each test, MX-5 was 1-2C worse than Kryonaut Extreme.
I also did a quick test on a 9900k IHS and MX-5 was about 2C behind Kryonaut Extreme.
Oddly enough I got best results with both pastes by doing full manual spread on the GPU cores. I found Kryonaut Extreme and Thermalright TFX to be within 1C of each other, with TFX winning out.
Currently have TFX on my R9 290X and Kryonaut Extreme on my MXM GTX 1070, TFX on my 3090 FE and TFX on my 11900k ES chip.
I've only used Themalright TFX on my 3090 FE (Kryonaut extreme was awful with the non compressible Odyssey pads), and after switching thermal pads first then spread method after, I saw two different results, each better than the other with TFX:
1) switching from Thermalright Odyssey 1.5mm pads to Gelid Extreme 1.5mm pads on core side: stopped Core to Core hotspot delta from degrading from 14C (Odyssey) to 11.5C (Gelid Extreme), both starting originally at 10.5C. This was with the heatsink pressure spread method. The low mounting pressure by Thermalright Odyssey pads caused "Pump Out" of the TFX with time, giving the slow rise in temps and increasing delta. This slow rise to 14C took three weeks. When I took apart the card and looked at the die, a few small edges of the die didn't have thermal paste on it at all (the die had full coverage on fresh application, so this is indeed pump out due to expand/contraction + low mounting pressure).
Switching to squisher (much better for this card) Gelid Extreme 1.5mm pads almost completely eliminated this but still required more work:
When I saw the delta rise from 10.5C to 11.5C after 2 weeks (it was 10.2C to 11C in one week) with the Gelid Extremes; --with the Thermalright Odyssey pads it was 12C by this time-- I dismantled the card and did a Fujifilm Ultra Low Prescale test, which confirmed there was excellent mounting pressure (still not uniform but that's an Ampere issue--see Igor's Lab article about this! Oddly enough there seemed to be more pressure around the edges of the die than the middle, which is not what I was expecting!) with the Gelid Extreme pads, unlike with the Odyssey pads, which had atrocious mounting pressure, even though the extra pad compression was only 0.2mm. That 0.2mm was massive. But I still saw a small increase, just a lot smaller than the TR pads.
Here is a picture of the pad compression between the two pads when I was preparing for the full TFX spread test, this was right before I cleaned the heatsink off to do the Ultra Low Prescale pressure paper test.
Gelid Extreme 1.5mm (ignore the partially melted 2mm pads on the backplate):
Odyssey compression: (discarded pads, this was before I put the Gelid pads on for the 2 week run).
A world of difference. You can even read the chip markings on the Gelids. Thats how well they compress.
Since pressure was fine, but it was already shown that the contact area on Ampere isn't fully uniform, I tried to do a full spread manually of TFX instead of the heatsink pressure (with large X pattern) spread, to see if I could get the slowly degrading deltas under more control.
This wastes a lot more paste, since TFX is so hard to spread and most of it sticks to the spatula, so you have to get a lot on. I used the pattern and method shown here.
Seanwee's post of where he spread TFX by yeeting it:
Thank you for visiting the TechnologyGuide network. Unfortunately, these forums are no longer active. We extend a heartfelt thank you to the entire community for their steadfast support—it is really you, our readers, that drove
forum.notebookreview.com
Result 10 days later: 10C Delta (400 FPS uncapped, Fortnite main menu @ 530W (this is a shunt modded card), up to 73C/83C where I stopped it), identical to fresh application. Highest I saw during these runs was 10.3C after a high temp ambient run (probably due to VRAM getting up to 96C at 550W), lowest was 9.9C (last night). Note that at 400W TDP instead of 530W, for some reason it's 11.2C, (something like 63C / 74.2C), I have no idea why. The delta is also 11C if I cap the FPS to 165, instead of running it uncapped, go figure.
So at least for this TFX paste with direct die, full spread is important.
(X pattern with heatsink pressure method was perfect on my 10900k ES however, even beating Kryonaut (original) by 2C after 1 week. Trying to spread TFX on an IHS is an exercise in complete futility)
Whats also noteworthy is when the card goes from full load to 210 mhz idle and cools down, it immediately drops to 10C Delta and remains there all the way to steady state. On the "pressure" TFX spread + Gelid pad before, this idle was 11C, and on the Odyssey pad idle cooldown test, it was 12C, so these are perfect results.
I did not take a picture of the fujifilm ultra low prescale paper. I threw it in the bin after testing.
I'm not sure if this 10C minimum hotspot delta is an 'accurate' reading and I'm not even sure if that this is a hardwired bios limit since I couldn't get it to decrease below 10C no matter what I did, no matter how cool the card got (even at 22C idle, it was 22C/32C exactly). Another user with a different card maker, using a custom loop claimed he got 8C, so this may possibly be set on different Vbioses. I have seen no 3090 FE users with less than 10C. Perhaps I'll ask on the Nvidia sub. On Pascal, this is 11-12C (11C Mobile gtx1070, may be 12C on desktop 1080?), hard offset and never changes. Thermspy 3.3.0 shows that at 11C on my MXM 1070. Oddly enough, Thermspy 3.3.0 which doesn't fully support the 3090, shows a non-changing "fixed" 8C Delta, regardless of what the delta is in HWinfo64 /GPU-Z.
