The Official Thermal Interface Material thread

[JWNoctis]

But no processor. The end resulted data is great for the measurement of those two blocks of copper. Neither is a common cpu cooler, and neither is a cpu. If it where a cpu, it would need to be nickel plated copper.

Is this testing actually accurate for the intended uses? At least what we do have is constant pressure. So that's a good thing.

So this thermal paste does X wm/k an hour between those 2 specific pieces of copper. That's what we get from that lab testing.

The question is, would this study be close enough to realistic usage in a PC??

Contrarily, this is probably as accurate as it gets: Control every other aspect and test the TIMs alone, without confounding factors such as socket size, IHS thickness, hotspot location, TIM spread, mounting pressure, heatsink type & make, ambient temperature, what the reviewer had for breakfast, and all that.

It is not real life because it is not meant to be, real life is too complicated to get a fair and repeatable result. If a make of TIM works well here, it would probably work well in real life, misapplication/bad socket thermals/bad mount/bad cooler notwithstanding.

Igor's just did a test showing TFX had thermal conductivity similar to PTM7950.

And the article just came up. He also more or less said it would also double as a grinding compound, and has in fact damaged the testing equipment, due to the forces involved to get it down to the tested thicknesses.

Is that actually the reason it worked well? By making its own scores into the IHS/silicon and the heatink surface, to enhance thermal contact between them and the TIM particles?

Definitely acquired taste, and probably not okay for bare die. Now I know...

 

[Borc]

25 µm results are useless from TFX yes, unrealistic pressure. We can see it on the odd curve between 25 to 50 µm. Minimum bondline of just 35 µm (50 µm with a more realistic pressure) can be a problem for some. For me the biggest issue however was the fast degradation.

 

[Nordic]

But no processor. The end resulted data is great for the measurement of those two blocks of copper. Neither is a common cpu cooler, and neither is a cpu. If it where a cpu, it would need to be nickel plated copper.

Is this testing actually accurate for the intended uses? At least what we do have is constant pressure. So that's a good thing.

So this thermal paste does X wm/k an hour between those 2 specific pieces of copper. That's what we get from that lab testing.

The question is, would this study be close enough to realistic usage in a PC??

Igor's is great because they test the actual thermal conductivity of a product is. Several pastes tested have not performed up to their marketing. Igor does not show what real world performance would be like. For example, Igor found the kryosheet to have mediocre conductivity but real world testing shows to perform just shy of liquid metal. Similarly, a lot of people in this thread were turned off from the Helios pads from Igor's test when real world reviews showed it performing just shy of liquid metal.

And the article just came up. He also more or less said it would also double as a grinding compound, and has in fact damaged the testing equipment, due to the forces involved to get it down to the tested thicknesses.

Is that actually the reason it worked well? By making its own scores into the IHS/silicon and the heatink surface, to enhance thermal contact between them and the TIM particles?

Definitely acquired taste, and probably not okay for bare die. Now I know...

How many of us can actually get enough pressure to get the paste down to 25um. When putting PTM7950 on my CPU, it starts at 0.25mm or 250um. How much does it compress before heating? I was able to get the nuts of my water block a half rotation tighter once it was warmed up. I have no way of measuring the pressure my block produces.

Toms Hardware's megatest showed Toplamp TL-A40 to be the best standard thermal paste. Toplamp TL-A40 is stated to have the same thermal conductivity as TFX. Tom's hardware said it was easy to use. Maybe a better choice than TFX.

 

[FreedomEclipse]

a lot of people in this thread were turned off from the Helios pads from Igor's test when real world reviews showed it performing just shy of liquid metal.

Ive slept on it and im still putting it on my 7800X3D later. I will double layer it and see what happens.

 

[Shrek]

25 µm results are useless from TFX yes, unrealistic pressure. We can see it on the odd curve between 25 to 50 µm. Minimum bondline of just 35 µm (50 µm with a more realistic pressure) can be a problem for some. For me the biggest issue however was the fast degradation.

Was pleased that my estimate of 1/20 mm (50µm) was not so far out.

 

[Nordic]

Was pleased that my estimate of 1/20 mm (50µm) was not so far out.

Google does the conversion for me. 50um is 0.05mm or 1/20mm.

 

[KLiKzg]

But no processor. The end resulted data is great for the measurement of those two blocks of copper. Neither is a common cpu cooler, and neither is a cpu. If it where a cpu, it would need to be nickel plated copper.

Is this testing actually accurate for the intended uses? At least what we do have is constant pressure. So that's a good thing.

So this thermal paste does X wm/k an hour between those 2 specific pieces of copper. That's what we get from that lab testing.

The question is, would this study be close enough to realistic usage in a PC??

So? He has:

• a source & target for heat transfer
• always the same amount of heat space to measure
• can measure the heat transfer between source & target
• has got 3 input sensors & 3 output sensors
• has always the same amount of pressure applied & his results are repeatable
• can measure the heat transfer from one unit to the other

Now lets look CPU:

• you have a source, but target is not mounted always the same way
• mounting target (cooler) is not in one axis, but always slips in X & Y axis, not only Z
• it has only sensor(s) on heated source (CPU), not on cooler - so we actually do not know what is the transferred heat to target (cooler)
• it has a variable amount of pressure applied, as the linkage of CPU mouting is not very accurate & the same...nor it is the linkage of target (cooler) while it bends & tightens into motherboard...as well as the mounting screws are not mounted with always the same torque.
• can't measure the heat transfer from one unit to the other.

Considering materials, Ni plated Cu is worse for heat transfer then Cu.
Considering it is not CPU, that is OK. As the each CPU has a cooling plate of given size, so the heat transfer can be calculated! The results are in different measurement units, but you can't expect that John, Wendy, Sam & you have same results for same CPU on same motherboard with same cooler (because of above mentioned deviations).


This test is more accurate then any mounting of CPU on motherboard with cooler! So this is true science.


What this test does is give you comparison from X W/mK (please use correct upper letters & correct divider) written on the box & measurement on the test device. So you, me & everyone else on this site & in the World; as a customer, will not be screwed up from some company. That is why this test in a lab is needed.


What can you expect in realistic usage o PC? Well, it depends from system to system, what kind of readings will you have on your CPUs.
But what you for sure can expect is:

1. Tested paste which gives results as written on box, can expect results when applied on CPU. Compared to some paste / pad which has a statement for example 11W/mK, but shows only 7W/mK.
2. With pastes & pads which have better results proven in the lab, same results can be expected in the real world. Which means, you will have few °C lower temperature on your system (if you use proven paste or pad, which has better (higher) results measured on a test then your current one).

And that is all you need...as you do not need to have 60°C on CPU while putting it on 100% stress...as if you need to have that kind of cooling, then air-cooler is not a way to go. But a water cooling, which can do those kind of results!

 

[ShrimpBrime]

So? He has:

• a source & target for heat transfer
• always the same amount of heat space to measure
• can measure the heat transfer between source & target
• has got 3 input sensors & 3 output sensors
• has always the same amount of pressure applied & his results are repeatable
• can measure the heat transfer from one unit to the other

Now lets look CPU:

• you have a source, but target is not mounted always the same way
• mounting target (cooler) is not in one axis, but always slips in X & Y axis, not only Z
• it has only sensor(s) on heated source (CPU), not on cooler - so we actually do not know what is the transferred heat to target (cooler)
• it has a variable amount of pressure applied, as the linkage of CPU mouting is not very accurate & the same...nor it is the linkage of target (cooler) while it bends & tightens into motherboard...as well as the mounting screws are not mounted with always the same torque.
• can't measure the heat transfer from one unit to the other.

Considering materials, Ni plated Cu is worse for heat transfer then Cu.
Considering it is not CPU, that is OK. As the each CPU has a cooling plate of given size, so the heat transfer can be calculated! The results are in different measurement units, but you can't expect that John, Wendy, Sam & you have same results for same CPU on same motherboard with same cooler (because of above mentioned deviations).


This test is more accurate then any mounting of CPU on motherboard with cooler! So this is true science.


What this test does is give you comparison from X W/mK (please use correct upper letters & correct divider) written on the box & measurement on the test device. So you, me & everyone else on this site & in the World; as a customer, will not be screwed up from some company. That is why this test in a lab is needed.


What can you expect in realistic usage o PC? Well, it depends from system to system, what kind of readings will you have on your CPUs.
But what you for sure can expect is:

1. Tested paste which gives results as written on box, can expect results when applied on CPU. Compared to some paste / pad which has a statement for example 11W/mK, but shows only 7W/mK.
2. With pastes & pads which have better results proven in the lab, same results can be expected in the real world. Which means, you will have few °C lower temperature on your system (if you use proven paste or pad, which has better (higher) results measured on a test then your current one).

And that is all you need...as you do not need to have 60°C on CPU while putting it on 100% stress...as if you need to have that kind of cooling, then air-cooler is not a way to go. But a water cooling, which can do those kind of results!

You guys are awesome because these are exactly the answers we need.

In a nut shell measuring W/mk at any size should determine thermal conductivity between two points and should net similar results even with different surface area sizes.

Thank yous for the explainations!! @Nordic @JWNoctis also!

 

[Borc]

Toms Hardware's megatest showed Toplamp TL-A40 to be the best standard thermal paste. Toplamp TL-A40 is stated to have the same thermal conductivity as TFX. Tom's hardware said it was easy to use. Maybe a better choice than TFX.

The thermal conductivity numbers from these two pastes are a meaningless marketing number, forget about it. Really good pastes are in the range of 4-6 W/mk. I tested Toplamp on my laptop last year and yes it is easy to use but the performance was mediocre, some results below with some other best pastes or pads.


Toplamp TL-A40= 87.25°C (ambient 20.6)
Dow Corning TC-5888= 82.25°C (ambient 20.8)
Koolingmonster Kold-01= 82.0°C (ambient 20.8)
Shin-Etsu X-23-8117= 83.75°C (ambient 20.7)
Thermal Grizzly Kryonaut= 85.75°C (ambient 20.8)
AMeCh SGT-4= 85.75°C (ambient 21.2)
Gelid Heatphase Ultra= 81.75°C (ambient 20.8)
PTM7950= 81.75°C (ambient 20.9)

 

[R-T-B]

So, quick question, is a 30mm*30mm ptm7950 pad adequate for a am5 cpu?

Might be ordering some ptm7950 pads just to see whats what.

 

[Nordic]

So, quick question, is a 30mm*30mm ptm7950 pad adequate for a am5 cpu?

Might be ordering some ptm7950 pads just to see whats what.

You need 38x38mm

 

[R-T-B]

You need 38x38mm

Even if not covering beyond the center? I imagine those capacitor ridges don't need much or any coverage.

I ordered 4 30mm pads. Guess I'll find out... lol. If it doesn't work they can make good gpu pads.

 

[Nordic]

Even if not covering beyond the center? I imagine those capacitor ridges don't need much or any coverage.

I ordered 4 30mm pads. Guess I'll find out... lol. If it doesn't work they can make good gpu pads.

I am pulling my information from what thermal grizzly recommends for the kryosheet. I was wrong anyway. AM5 needs 33x33mm, so 30x30mm will probably be fine.

The thermal conductivity numbers from these two pastes are a meaningless marketing number, forget about it. Really good pastes are in the range of 4-6 W/mk. I tested Toplamp on my laptop last year and yes it is easy to use but the performance was mediocre

Igor's is great because they test the actual thermal conductivity of a product is. Several pastes tested have not performed up to their marketing.

I understand that the conductivity shown on most pastes is for marketing rather than performance. TFX has the highest conductivity out of all pastes Igor has tested and trades blows with PTM7950 depending on the pressure. Both are within 4-6 W/mk. For those of us chasing that last 1% of thermal performance, or at least for me, TFX looks attractive despite the drawbacks that Igor mentioned. TFX is ridiculously thick, difficult to apply, and might damage your cooling equipment.

Toms Hardware's megatest showed Toplamp TL-A40 to be the best standard thermal paste. Their results showed it to be a top performer, not far off from PTM7950 or liquid metal. Toplamp TL-A40 is stated to have the same thermal conductivity as TFX, so maybe it actually has similar thermal conductivity. Tom's hardware said it was easy to use. For those of us chasing that last 1% of thermal performance, Toplamp might be an attractive choice. I hope Igor includes it in a test sooner than later.

Your one anecdotal experience isn't enough to convince me that toplamp is a mediocre paste. Tom's Hardware showed just the opposite. At the very least, I have two datapoints that conflict. At least for now, I am more willing to believe Tom's hardware but your experience gives me some doubts. This IS EXACTLY the situation where Igor's reviews shine @ShrimpBrime; Igor informs us which pastes are top performing so we don't rely on marketing alone. Igor can tell us if Toplamp has exceptional thermal conductivity or not.

So, quick question, is a 30mm*30mm ptm7950 pad adequate for a am5 cpu?

Might be ordering some ptm7950 pads just to see whats what.

I have PTM7950 installed on my cpu right now. I have done at least 10 heating and cooling cycles. I am not sure on the exact count but it is a lot. When I was comparing Kryonaught extreme to Kryosheet, I only ran the stress test for 5 minutes because the thermal results were extremely consistent from 30 seconds in. I did not see a reason to go longer if the numbers wouldn't meaningfully change. In my testing so far, PTM7950 performed 1c worse than the kryosheet after 5 minutes of stress testing. Ambient temperature was the same so I am as confident in these results as I can be.

I did have an odd behavior. After almost exactly 20 minutes of stress testing, the temperature rapidly dropped 5c giving me a new personal best at 73c. I am not entirely certain why this is happening. I wish I had tested Kryonaught Extreme and Kryosheet for a 30 minute run before hand to know if this behavior is unique to PTM7950.

I have read PTM7950 isn't good for CPU cooling because the IHS might not heat evenly, allowing the PTM7950 to melt evenly and provide the best cooling. People say PTM7950 should only be used for direct die cooling. I have seen no data to support this. Tom's Hardware showed PTM7950 performing just shy of liquid metal which is an actual data point for comparison. My best guess to what is happening is after 20 minutes of stress testing my IHS and waterblock have become heat saturated enough that the PTM7950 fully melts and provides exceptional heat transfer. @R-T-B would you please confirm this when you apply PTM7950. Please use the following OCCT with the following settings and see if you have a rapid drop in temperature after 5 minutes.

OCCT 13.1.4 CPU
Stability Test Mode: Extre
Load Type: Stea
Start at Cycle:
Instruction Set: S
Thread Settings: Auto

 

[KLiKzg]

Even if not covering beyond the center? I imagine those capacitor ridges don't need much or any coverage.

I ordered 4 30mm pads. Guess I'll find out... lol. If it doesn't work they can make good gpu pads.

It will not work, order correct ones 37,5x37,5mm: https://www.igorslab.de/en/amds-socket-am5-mounting-technical-data-and-the-fitting-boxed-coolers/

 

[Solaris17]

fine.

Nice they included the measurements for 4677. But I would still need to test it. This CPU is not square, and Intel as well as other reviewers say the IHS should be totally covered. Which I believe more. SPR may only be 50x50 under the IHS but the IHS on these big chips helps drastically with thermal stability; they are thick and quite heavy. I’d be MUCH more comfy with like 70x50

 

[R-T-B]

I am pulling my information from what thermal grizzly recommends for the kryosheet. I was wrong anyway. AM5 needs 33x33mm, so 30x30mm will probably be fine.

I am hoping so but like I said if they look too small they'll make great gpu pads.

 

[lexluthermiester]

So, quick question, is a 30mm*30mm ptm7950 pad adequate for a am5 cpu?

Might be ordering some ptm7950 pads just to see whats what.

No, but a 40mmx40mm would be fine. 45mmx45mm would be more optimal as there would be wiggle room.

 

[R-T-B]

No, but a 40mmx40mm would be fine. 45mmx45mm would be more optimal as there would be wiggle room.

Considering I was using a 33mm * 33mm Kryosheet until recently I know I can get away with at least that little. I guess the question is if the additional loss of yet another 3mm will be problematic re where the actual IHS "hotspots" are.

We'll find out lol. If it does not work I'll order a 40mm*40mm one next, and use the smaller ones on various gpu dies.

 

[lexluthermiester]

Considering I was using a 33mm * 33mm Kryosheet until recently I know I can get away with at least that little. I guess the question is if the additional loss of yet another 3mm will be problematic re where the actual IHS "hotspots" are.

Fair enough. With pad/sheet type TIMs it's important but not critical to get complete coverage of IHS. However, if you're careful(and many of us know you are) a slightly smaller pad/sheet can work well.

and use the smaller ones on various gpu dies.

That would be an ideal use-case-scenario for those.

 

[freeagent]

That Heilos stuff isnt too bad. I have had it on for a week now I think..

Fans are on a curve, and the card can actually breathe now.

 

[Nordic]

That Heilos stuff isnt too bad. I have had it on for a week now I think..

Fans are on a curve, and the card can actually breathe now.

Turns out 95% performance of ptm7950 is pretty good. Who knew.

You don't even have to worry about sourcing real stuff. Such convinience.

 

[jayjr1105]

Even if not covering beyond the center? I imagine those capacitor ridges don't need much or any coverage.

I ordered 4 30mm pads. Guess I'll find out... lol. If it doesn't work they can make good gpu pads.

If you do just try a 30x30, you should run it for a few days and see how it spreads. You can cut and use multiple pieces on the IHS as the PTM will melt and blend itself. It's quite forgiving.

 

[Borc]

Igor added results from a newly purchased Kold-01 after restock (unknown if it differs to the older Kold-01). This paste is crazy, it makes me wonder if TC-5960 can beat this.

Industriepaste vs. AliExpress und Mainstream: DOWSIL TC-5550, Gelid GC-4 und Snowman MTGS im Wärmeleitpasten-Test der großen Tuben | Seite 3 | igor´sLAB

Heute habe ich die nächsten drei Pasten von Eurer Wunschliste im Programm (vielen lieben Dank an die Spender aus der Community!) und war mit dem Ergebnis schon eher zufrieden (aber nicht ganz).

www.igorslab.de

 

[R-T-B]

Joining in the fun I guess:

Yes my order was updated to 40mm * 40mm pads. Ordered from MODDIY, a known legit supplier of Honeywell stuff. Also got some halfway decent Fujipoly pads because I was running low.

BTW, if the storage conditions for this stuff seem surprisingly strict, that's because that's for lab grade measurement work (or so I was told). Aparently it should have no issue lasting to the expiration date at standard common room temps. If you believe MODDIY support (I trust em anyways).

Already applied one set to my 7900XTX. Had to trim it down a bit of course lol. Tired now but it goes on my AM4 and AM5 CPUs tomorrow (5800X and 9950X respectively).

 

[Ruru]

Got a 10g syringe of Noctua NT-H2 bundled with components I bought recently. Nothing to complain and what I've read, Noctua's paste is more than fine.