Thermal Pad shim for CPU/GPU

Has anyone thought to use a couple thin lines of thermal pad along the edges of the core, so the sides of the core can dissipate into the heatsink as well? I've been searching for about an hour and can't find anything, so maybe this hasn't been tried yet. I was thinking of trying this with the vram as well. Thoughts?

 

I think it's a good idea.

If you try it, let us know the results (if the temps change at all).

 

The "core" you see is not the actual core of the processor. It's a heat spreader that's soldered onto the chip itself. Thermal pads are usually placed directly over the physical cores to try to transfer the most heat. Same with the notebook heatsinks - the pipes are typically placed in the block directly over the actual core, not across the entire heat spreader. Still, you'll likely see at least marginal improvement.

 

It should not make a difference.

The actual "top" or "cap" of the CPU that you see is the heat spreader. And the material used to make that is so thermally conductive (because that's what it was designed to do - conduct heat), that adding more physical contact with the heat sink won't really do anything.

So, you can go for it... a thermal pad isn't electrically conductive, so you won't short anything out if you were to bridge some contacts that are on the surface of the CPU. But you probably won't notice any difference in temps.

 

I think you should try that but with some good conductive thermal pads. Theoretically they shouldn't do any harm but if they improve heat transfer then that would be a very interesting fine. I support such experiments .

 

Are you thinking about making a piece of copper, that is flat on top, but has an indent on the bottom for the entire CPU/GPU heatsink?

 

It does not do anything. I've done that experiment before using .5mm and 1mm thick thermal pads. (Unfortunately 1mm is too thick and it didn't work) The heat still dissipates through the core (mirror area) of the CPU. I've also substitute with copper, measured 3.5x3.5cm and cut a 1.2cm wide hole in the center to fit the core, shimmed, applied ICD7 in and -around- the core to fill the gaps completely and have it placed on the CPU. The temperature output is the same and does not yield any difference as before.

Unfortunately I didn't document the whole process, as I knew before hand that it wouldn't do any improvement. But this was what it was when I tested it.

 

This isn't going to do anything. The weak point of the cooling system isn't the transfer of heat from the core to the heatsink, it is far and away the heatsink itself. On top of that, any heat transfer from the cpu packaging is going to be hugely inefficient compared to the core, as none of that is designed to transfer heat. Added to those two things, the desire to use thick, very inefficient thermal pad material means that there is going to be no measurable difference. Of course it would still be neat to try to see if there is maybe .5 degree difference, but I would really doubt it is anything measurable.

There is no heat spreader on Intel's mobile CPU's. What you see is the actual die. Yes, the 45nm nehalem core is huge.

 

I think you risk the heat sink not sitting squarely on the die more than anything, specially with thicker pads. As far as any difference in temps, you will not see any, the most heat is dissipated through the surface with the most area, which is the top of the die not the sides.

 

I'm thinking I am going to try it. Think about the fuss some of us put up over a corner on the die missing paste. Surface area is surface area, so I don't see ho it could hurt, especially with video memory.

When I get my 940 next month, I'll order some different size pads and test it out.

 

Don't You guys forget, that when You apply the thermal paste, not all of it (all of us usually apply a bit too much, as it is generally better to have too much rather than too little) stays between the core (or heatspreader) and the surface of the heatsink. Some amount of the thermal compound may (and usually will) be squeezed out from between the core (or heatspreader) and the heatsink. This little amount would (IMHO) do exactly the same job as the stripes of thermal pads around the core, mentioned earlier by the OP. If there's something wrong with my logic, please enlighten me, as 12 beers so far might have blurred my vision tonight ;-).

 

If your paste squeezes out, you've used too much. Thermal paste is not a glue. It's a microscopic gap filler.

 

The problem is that thermal paste that thick transfers heat very, very poorly, and there is no way you should ever have enough squeeze out to make contact between the packaging and heatsink. Even if you had thermal past bridging the gap across the whole package, it would do even less than thermal pads and not be stable as it would ooze out.