Best way to apply thermal compound!

Soldato
Joined
20 Oct 2002
Posts
11,293
Location
Derby
The problem comes with that method that a) the newer i7 and threadripper both have significant amount of cores and bigger/more dies so some methods will be better for different cpus. b) different thermal compounds have different properties like viscosity. For instance using Hydronaut is fine with a 6-8mm method on newer cpus but the same with Kryonaut doesn't work as well as it doesn't really spread very well so needs to be spreaded.
 
Soldato
Joined
1 Jul 2011
Posts
8,641
The problem comes with that method that a) the newer i7 and threadripper both have significant amount of cores and bigger/more dies so some methods will be better for different cpus. b) different thermal compounds have different properties like viscosity. For instance using Hydronaut is fine with a 6-8mm method on newer cpus but the same with Kryonaut doesn't work as well as it doesn't really spread very well so needs to be spreaded.
Post #12 shows several CPU die under IHS and says AMD dies are sometimes in corners. I know Threadripper is much bigger, but so is it's IHS. Could you post images of CPU die size of AMD and other Intel CPUs?
 
Soldato
Joined
20 Oct 2002
Posts
11,293
Location
Derby
Post #12 shows several CPU die under IHS and says AMD dies are sometimes in corners. I know Threadripper is much bigger, but so is it's IHS. Could you post images of CPU die size of AMD and other Intel CPUs?

Not got any pics (They can be found easily on google) But the die on the Skylake-X cpu is about the same size as 2 x sandybridge cores sidebyside.
 
Soldato
Joined
1 Jul 2011
Posts
8,641
Not got any pics (They can be found easily on google) But the die on the Skylake-X cpu is about the same size as 2 x sandybridge cores sidebyside.
A little bigger, but not anything like Threadripper.

Skylake-SP Die Sizes
Arrangement Dimensions(mm) Die Area
LCC 3x4 (10-core) 14.3 x 22.4 . 322 sq mm
HCC 4x5 (18-core) 21.6 x 22.4 . 484 sq mm
XCC 5x6 (28-core) 21.6 x 32.3 . 698 sq mm

Bloomfield die is about 263 sq mm
Sandy Bridge E dies is about 435 sq mm
Haswell SSE dies are about 661 sq mm
Nehalem EX die is about 684 sq mm

Threadripper dies are about 191 mm sq each with all 4 (2x dummies) setting in a 1519 sq mm

The above data for Googling shows new Intel similar to older Intel. I didn't look at what the IHS sizes were, but assume in the 40-43 x 40-43mm range.

Each Threadripper die is almost as big as smaller Intel dies .. and each die is spaced about 11mm from it's neighbors. But Threadripper PCB is more than twice the size of any Intel.
 
Associate
Joined
4 Oct 2017
Posts
1,221
I tend to just put a dot on the middle of the heatspreader. The size of the dog depends on the cpu and size of the die.

Too much thermal compound can actually be a bad thing. It’s not a sandwich you’re spreading.
 
Soldato
Joined
17 Dec 2004
Posts
8,700
I have always spreaded mine thinly...I use AS5 and put a small blob in the center of cpu. then spread it so it completely covers the cpu with my finger in a food bag. Also doing it this way, no paste squeezes out of the sides when securing the the heatsink to cpu,
 
Man of Honour
Joined
13 Oct 2006
Posts
91,158
Gamer's Nexus found that some pastes like AS5 could be as bad as 10C from optimal when hand spread due to the potential for air pockets, etc. in my experience its been more like 2-3C but still.

I did a whole load of tests back in the day when I was building systems more regularly and spreading a lot of pastes resulted in a "stippling" like effect and less efficient results so I only spread if the application notes say to.

If I don't know the die layout then I've always gone for a small blob in the middle - for some CPUs a thin line can work a little better especially older LGA775, etc.

EDIT: They seem to have tweaked AS5 along with changing the application notes at some point though to make it more tolerant of hand spreading I guess because a lot of people still do - originally the application notes strongly suggested not spreading it by hand.
 
Soldato
Joined
1 Jul 2011
Posts
8,641
The key is to not have a layer of TIM between CPU IHS and cooler base. While a layer of TIM is better than no TIM because TIM tranfers heat many times better than air, direct metal to metal contact transfers heat many times better than TIM. Obviously the IHS is not really a heat spreader at all, but a cap over CPU die/s that reaches out near the edges of CPU PCB to supportcooler bases without their bending the PCB and damaging CPU and connections to motherboard. A dob that spreads more than a couple mm beyond the CPU die contact area on IHS is not transferring heat .. but bigger dobs do spread farther and thereby increase the chance of TM layer between surfaces .. which lowers heat transfer to cooler and thereby increases CPU temps.
 
Associate
Joined
23 Jul 2012
Posts
149
Always have gone for the pea-sized application. But there are plenty of articles and videos to show it doesn't really matter how you apply it. Only thing I know/seen for sure is too much is actually worse.
 
Soldato
Joined
17 Dec 2004
Posts
8,700
I use the pea-sized and spread it technique, because your filling in the microscopic holes on the cpu, so you might aswel fill all the holes instead of just filling in a few in the middle of the cpu, because its its not just the middle of the cpu that gets hot, because heat spreads. Plus it makes a much tidier job then letting the heatsink spread it out for you. Thats why people dont recommend Arctic Silver 5, because they use the quick and easy way to apply it then drips on other components and shorts them out.
 
Soldato
Joined
1 Jul 2011
Posts
8,641
Key is enough to spread into a print that covers CPU die under IHS and not so much it's oozing out (and there is a decent amount of pressure to thin it out) the difference is only a couple three degrees. This is assuming TIM is not liquid metal or so thick it only spreads enough to look like a miniature hamburger on IHS mini-grill. ;)
 
Back
Top Bottom