IC Graphite for hotspots

[JohnnyDiggs]

HI team,

I just picked up a Threadripper and X399 for a steal last weekend, and I'm waiting for the monoblock to get here from EK before I get it all setup. I've watched about a million videos about threadripper thermal paste application, and I've also seen the rumors about the X3900 hotspot issues and making sure your heatpipes are aligned. This makes me think that the threadripper would also have hotspots that would need to be managed correctly.

 

I'm also a fan of the IC Graphite pads, since I don't get that lingering doubt of "did I use enough paste" that only a 12 pack will get rid of. Unfortunately, I'm newly broke (I bought a threadripper and ek monoblock) and can't afford a 12 pack anymore.

 

In the video review they discuss for a second how the pad is 10 times more conductive sideways than through. That suggests that it would be great a spreading out hotspots in hotspot prone chips. Has anyone done any testing of the anti-hotspot properties of the graphite? Are there any obvious reasons why this is stupid and I'm stupid and I should delete my browser?

 

I've attached the material datasheet that shows the conductivity: " Thermal conductivity : X-Y direction 400W/m∙K, Z direction (28W/m∙K) "

 

AYA0000C50-1129407.pdf

Edited September 7, 2019 by JohnnyDiggs
SPELLING

[JohnnyDiggs]

Also, the datasheet says it works best under 600 kPa of pressure. How much pressure does a heatsink apply?

[Guest]

I used it on my 2950X for a while. I needed to use two of the big ones, so it wasn't any cheaper than paste. I lost a little bit of benchmarking top-end, but no impact on day to day.

I switched back to paste because I was having issues with Enermax coolers and it was actually more frustration to line up the two pads than cleaning/reapplying paste.

 

[JohnnyDiggs]

So it didn't help your thermals? Interesting, if a bit discouraging.

[For Science!]

7 minutes ago, JohnnyDiggs said:

 

I think a typical key misunderstanding in the community is that thermal compound/pads are "improving" the thermal conductivity between the CPU IHS (or die) and the metal heat sink. This is categorically untrue as metal-to-metal contact has far higher thermal conductivity than any thermal compound can dream to provide. If the two surfaces were completely flat at a atomistic level, there would be no need for thermal compound.

 

How the thermal compound improves is by displacing insulating pockets of air between two non-flat surfaces and therefore improving (by eliminating) thermal transfer between CPU IHS-->Air-->Heat sink. Now, the problem with pads is that it completely covers the contact area and does not "squeeze out" and so while you remove some of the air pockets, you also completely remove the metal-to-metal contact you would have had. Therefore you are completely in the mercy of the Z-directional thermal conductivity of the pad (which isn't that great) and overall likely to lose mass heat transfer capability when compared to having some degree of metal-to-metal contact.

 

Furthermore, specifically about spreading chiplet heats from Ryzen/Threadripper also fails to acknowledge that, yes, the chiplets are small dies with focussed heat; however the IHS sitting on top of it is soldered, and also a solid chunk of nickel plated copper. And so the heat from these chiplets will travel via the solder layer to the IHS, and within the IHS can spread in the X, Y, Z direction with 389 W/m K  being a solid chunk of copper, so by the time you are thinking of spreading the heat further with a pad, the spreading has already been done, and you are simply adding another layer of insulation.

 

@bignaz made some fairly bullish claims in the thread below but also never followed up on the results. So my best guess is that the results were disappointing. On the other hand @porina did come back with results and as I had originally predicted, the results are worse than using standard paste.