The Official Thermal Interface Material thread

[Veseleil]

I think AMD IHS is soft compared to Intel.

True. Figured when I did that 3930K lapping.

 

[freeagent]

True. Figured when I did that 3930K lapping.

I figured that when my TIM pitted the IHS on my 5900X but not my 3770K

 

[ThrashZone]

I think AMD IHS is soft compared to Intel.

Hi,
Intel chips are never flat they are always concave which is really stupid and why a lot of people end up delidding the chip to get good contact
Some intel cpu water blocks are designed for the concave intel surface but I doubt any amd/ intel compatible air/ water cooler would be because on amd the concave wouldn't do very well.

 

[freeagent]

Hi,
Intel chips are never flat they are always concave which is really stupid and why a lot of people end up delidding the chip to get good contact
Some intel cpu water blocks are designed for the concave intel surface but I doubt any amd/ intel compatible air/ water cooler would be because on amd the concave wouldn't do very well.

I have some old TR coolers, the LGMRT and TS140P and they do really well with Intel because of their convex coldplate, but they do poorly with AM4, specifically 7nm, 12nm or whatever Zen 2 was produced in was fine, excellent even..

 

[Count von Schwalbe]

they do poorly with ... 7nm, ... whatever Zen 2 was produced in was fine

Zen 2 was 7nm... Did you mean Zen+?

 

[freeagent]

Zen 2 was 7nm... Did you mean Zen+?

I thought cpuz said 12nm on my 36XT..

Whoops now I’m a story teller

Sorry

 

[Pictus]

My Akasa Venom Medusa was very convex and the Thermalright Frost Commander a little, I lapped both to better fit the Ryzen 5600X.

Intel Core i9 vs. AMD Ryzen 9: How convex or concave are the current heatspreaders really and what are the consequences? | Labs | igor´sLAB

We already know it from many (also mine) articles about cooling, thermal paste and heatspreader: really plan is really nothing. And while the quality of the radiator floors has really improved over…

www.igorslab.de

 

[Borc]

I have a ~9 year old Dell laptop, which probably has been using the same thermal interface material since whenever it left the factory. Few days ago I've bought Genesis Silicon 851 as it was quite cheap & supposedly offered almost same thermal conductivity as the best non-liquid metal TG thermal paste, which even though seemed somewhat fishy, I still went with. After applying the whole tube (0,5g) to the two dies in my laptop & then checking thermals, I am honestly not sure if it helped... kind of feel like the temperature at the hot end of the spectrum is worse than it was before as I think it heats up easier & a bit more than it did before but also manages to lower the temperature quicker, however the idle thermals have been improved by at least 5*C at the minimum. The stock thermal paste my laptop has been using might've not even been a paste, but rather something akin to thermal glue or similiar compounds, as it was not crumbly or cracked when I took off the cooler & was stuck to it as well as the CPU die quite well despite its potential age

I tested Genesis Silicon 851 some time ago on my Dell laptop. It's not the worst paste but there is a gap to the best ones.

Genesis Silicon 851
87/91/87/88

Kooling Monster KOLD-01
85/87/86/86

Noctua NT-H2
86/90/87/87

 

[Soul_]

So I realized why some people weren’t having too much success with PTM 7950 SP, which is a paste (some websites are incorrectly showing SP as a pad, as well as a paste. But as per Honeywell white paper, SP is a paste and non SP is a pad). The paste has drying time of 5 minutes at 100C or 24 hours at 23C. And it seems that people running into temp issues weren’t letting the paste dry for 24 hours before clamping and squeezing it out.

As for the PTM 7950 non SP, which is a pad, as per Honeywell white paper, no drying time is required.

So I just ordered PTM 7950 (pad), no drying time, and less chance of error.

Thank you everyone above for your help and insight. I will keep you in loop on how it goes with it.

Here's my followup on the statement I had made. Apologies for the delay, I was away for past 4 weeks.

Once I came back, I switched my KOLD-01 application on 3080ti FTW3 Ultra to PTM-7950. A few key items of interest, once I took the GPU apart:

• KOLD-01 did pump out.
• There were portions of surface, which were completely bare, on both GPU die and its corelated point on the cold plate of the heat sink.
• But there were no signs of "drying out" anywhere.

As for PTM-7950:

• The application was easy-ish, I did have to cool it in freezer first to ensure I could peel the plastic off without distorting the TIM layer.
• Temps are similar to brand new application of KOLD-01, from all my tests.
  • 74C-75C during extended Kombustor run at 450Watts.
  • 71C-72C during Superposition 1080p extreme run, but the scores were marginally better, hitting 13.5k, vs 13.4k on my previous runs at the same OC.

Now what I wait for is, to see if PTM-7950 is resilient to pump-out and runs well for the lifetime of the GPU. I will post back in a few weeks to keep you updated.

I hope this helps.

 

[Borc]

I tried PTM7950 on my i7-1165G7 today, in my first test it's roughly on par with the best pastes I have like KOLD-01 or AeroCool Fuzion or Thermalright TFX. So there is no out of the box magic from what I can see when compared to the best pastes, of course longevity it could be better than the traditional paste. Also it's better than like 95% of the pastes out of the box from day one, because most pastes are really bad for laptop CPUs including many of the popular ones like Phobya NanoGrease Extreme, Arctic MX-4 or Gelid GC-Extreme (fresh application!). Compared to these even from the first day PTM7950 is 5-7 degrees better. Combined with some pump out after 1+ month gap could increase to like 10 degrees. I think that most people on reddit compare PTM7950 with their old not so fresh paste which explains the huge gains.

 

[freeagent]

I went on super thin with SYY, I think I was using too much before

Never saw pump out though, that's crazy. Must be some thin greasy TIM?

So far I am seeing about 7-10c difference between SYY and AS5/Ceramique at the top end, maybe a bit more depending on the wattage

 

[gaseosx]

PTM7950 requires several high temp cycles to melt and prime up. You should have consistent temperatures with a day or two of use.

 

[Frick]

This might be off topic, but I just repasted my RTX 3060ti with some decade old MX-4 and temps went from mid 90's C to low 70's C with full load, with fans at 60%. Should have done it earlier, but I didn't really notice it until a few months ago and I did the easy thing and just lowered the power limit. The card was bought as a refurbished item (with full warranty) and I assume the people refurbishing it didn't notice the excessive amounts of paste. I didn't take any pics, but it was a lot. Feels good to not have a computer that doesn't get that hot smell.

 

[Borc]

Did anyone try out Keratherm KP12/KP99? I wonder if this is a good choice. They have a very low thermal impedance (which is more important than the thermal conductivity). KP12 slightly lower than KP99 but in a very low pressure environment KP99 is better than KP12 according to them: https://www.kerafol.com/_wpframe_cu...nch_KP97_98_99_12_10-19___145030-10012020.pdf

Low pressure is a use case for laptop cooling.

@Frick, you could try out MX-6 or some other top paste.

MX-4 1° higher than MX-6 with a 12% higher fan speed.

 

[Frick]

@Frick, you could try out MX-6 or some other top paste.

MX-4 1° higher than MX-6 with a 12% higher fan speed.

Well, the 11g tube is only like 50% empty, so I should be in the market by 2033.

 

[lexluthermiester]

This might be off topic, but I just repasted my RTX 3060ti with some decade old MX-4 and temps went from mid 90's C to low 70's C with full load, with fans at 60%.

How is that off topic?

 

[Veseleil]

decade old MX-4

I have a 8g tube of MX-4 maybe 5 years old, it started to separate oil, but it's still perfectly fine. I guess I'm not buying >4g tubes in the future.

 

[Sithaer]

This might be off topic, but I just repasted my RTX 3060ti with some decade old MX-4 and temps went from mid 90's C to low 70's C with full load, with fans at 60%. Should have done it earlier, but I didn't really notice it until a few months ago and I did the easy thing and just lowered the power limit. The card was bought as a refurbished item (with full warranty) and I assume the people refurbishing it didn't notice the excessive amounts of paste. I didn't take any pics, but it was a lot. Feels good to not have a computer that doesn't get that hot smell.

I've also repasted my 3060 Ti a few months ago with a years old GD 900 that my brother bought from Aliexpress a good while ago.
Supposedly its the original first edition thats close-ish in performance to MX-4.

In my case my card had a completely dried out paste, there was nothing left but a thin layer of dust like cover.
Its a second hand card but otherwise its in a good condition and it still had the screw sticker on it, also came with a warranty from the seller which ofc I've voided but eh the temps were bothering me. 'it was fine when I bought it, I guess that factory paste was on its last legs'

So far its holding up, not really noticing worse temps than after repasting.
My card dropped like ~20 C on the hotspot and ~10 C on the core down to barely 60+ C while gaming and that also lowered the fan speed by quite a lot. 'around 400-500 rpm lower so now its working as it should'
I'm tempted to give MX-6 a try but it seems kinda pointless since its still working just fine with good temps and I doubt that it would be much different. 'I'm also FPS capping my games and undervolt the card'

 

[Borc]

I finished my Honeywell 7950 phase change pad test on my laptop with an i7-1165G7. I'm using my Kooling Monster KOLD-01 for this comparison because it's basically best working for my laptop. Sure AeroCool Fuzion and Thermalright TFX might or should offer the same but TFX is hard to spread and I don't have much Fuzion paste left.


Dell i7-1165G7 @2.8 Ghz
Honeywell PTM7950= 84.75°C (ambient temperature 20.6°C, fan speed= 2547rpm)
Kooling Monster KOLD-01= 84.25°C (ambient temperature 20.6°C, fan speed= 2557rpm)

I tried two applications with KOLD-01 and this is the better one. My first try with a bigger layer wasn't as good, the thinner layer worked better. So it depends a bit on the application and how much paste I use. I think I nailed my second attempt with the thinner layer.

1 day of curing for both and 6x30 minutes OCCT CPU load cycles. 0.5°C difference with a 0.4% higher fan speed which might affect the temps by 0.2-0.5. PTM7950 is roughly as good as the best traditional paste after a fresh application. There is no magic against the best traditional paste after a fresh application, sustained performance however PTM7950 should be the winner I believe.

I should add that many pastes are struggling in a laptop, PTM7950 is winning against most pastes even on a fresh application. For example with Arctic MX-4 and same test procedure the temps are at 93°C. Big difference and combined with some pump out the difference can be even bigger. I think this explains some of the huge differences reported from laptop users on reddit.

 

[Veseleil]

Repasting and repadding my 6900XT OC Formula from MX-4 and stock pads to MX-6 and Arctic TP-3 thermal pads yielded 10-15 °C all around temperature drop (hotspot, mem, core etc.).
Also, I did stack pads to get needed thickness. So 1+1mm to get 2mm, 1,5+1,5mm to get 3mm.

 

[Soul_]

Here's my followup on the statement I had made. Apologies for the delay, I was away for past 4 weeks.

Once I came back, I switched my KOLD-01 application on 3080ti FTW3 Ultra to PTM-7950. A few key items of interest, once I took the GPU apart:

• KOLD-01 did pump out.
• There were portions of surface, which were completely bare, on both GPU die and its corelated point on the cold plate of the heat sink.
• But there were no signs of "drying out" anywhere.

As for PTM-7950:

• The application was easy-ish, I did have to cool it in freezer first to ensure I could peel the plastic off without distorting the TIM layer.
• Temps are similar to brand new application of KOLD-01, from all my tests.
  • 74C-75C during extended Kombustor run at 450Watts.
  • 71C-72C during Superposition 1080p extreme run, but the scores were marginally better, hitting 13.5k, vs 13.4k on my previous runs at the same OC.

Now what I wait for is, to see if PTM-7950 is resilient to pump-out and runs well for the lifetime of the GPU. I will post back in a few weeks to keep you updated.

I hope this helps.

1 month followup. Temps with PTM-7950 are rock solid. In addition to previous tests, I ran power-normalized temperature tests at 80% fan speed:

• At 400 Watts, temps stablise at 69C to 70C.
• At 420 Watts, temps stablise at 71C to 72C.
• At 440 Watts, temps stablise at 73C to 74C.

This a 1C improvement over stock TIM from EVGA, and in-line with fresh application of Kryonaut or KOLD-01.

Overall I would say, absolutely better than any high end TIMs (for baredie application), as there is no pump-out effect noticed in the 1 month period (of dozens of gaming and benchmarking sessions), given that typical TIMs show noticeable pump-out degradation within 2-3 weeks of gaming (with temps easily degrading by 3-4C).

Would I recommend this over stock TIM from EVGA, no. I don't think 1C is note worthy. Would I recommend it for repasting (if stock TIM has degraded), absolutely yes.

I hope this helps.

 

[delshay]

I have a new experimental way of applying thermal paste for Vega cards or any other chip that's molded. This experiment is new & I have only done this once, so I need to do some more test to see if I can push any further.

Anyway here hows the thermal paste is applied. Instead of covering the complete die including the molded part, what I have being running for the past few weeks is, I applied paste just to the die itself avoiding any part of the paste coming into contact with any molded part. I applied the paste with a tiny flat-head screwdriver so I did nod not cross the boarder of any die.

The result as follows, HBM & GPU die now more or less now have the same temperature. It does drift a bit depending on GPU load, but in normal use when playing games it's the same temperature. Hot-spot could do with some improvement, it's about 10C higher than the core.

This is 100% experimental & I can explain the idea behind this.

 

[maxfly]

Forgot to update the Arctic mx6 application on my i7 8086k rig. Spread and coverage is still fine on both the ihs and heatsink. Paste was nice and gooey after a short heat up period. Very little squeezed out during initial application. Could've reused it tbh. Temps hadn't changed prior to swapping the CPU out.

Mx6 showed the same results on rtx2070. Coverage and spread were just as it should be between the die and hsf. Again, temps hadn't changed at all.
Reason for pulling? Swapping the CPU and GPU into different rigs and wcing the rtx2070. Will continue to use mx6 for both. Temps were maybe a degree off of pk3 mounts but mx6 is cheap as chips!

 

[lexluthermiester]

1 month followup. Temps with PTM-7950 are rock solid. In addition to previous tests, I ran power-normalized temperature tests at 80% fan speed:

• At 400 Watts, temps stablise at 69C to 70C.
• At 420 Watts, temps stablise at 71C to 72C.
• At 440 Watts, temps stablise at 73C to 74C.

This a 1C improvement over stock TIM from EVGA, and in-line with fresh application of Kryonaut or KOLD-01.

Overall I would say, absolutely better than any high end TIMs (for baredie application), as there is no pump-out effect noticed in the 1 month period (of dozens of gaming and benchmarking sessions), given that typical TIMs show noticeable pump-out degradation within 2-3 weeks of gaming (with temps easily degrading by 3-4C).

Would I recommend this over stock TIM from EVGA, no. I don't think 1C is note worthy. Would I recommend it for repasting (if stock TIM has degraded), absolutely yes.

I hope this helps.

Those are very reasonable performance numbers. Seems like good stuff!

 

[Soul_]

Those are very reasonable performance numbers. Seems like good stuff!

Agreed. I was a bit skeptical going PTM route, but pleasantly surprised. After repasting with off the shelf TIMs last year, mostly I was regretting it, due to pump-out every 3 weeks or so. Now, I think I should be good till nVidia 5xxx or AMD 8xxx comes out.

Cheers!!