GPU - Thermal pads instead of paste? Anyone tried it?

[Stu @ MSD]

Hi Guys,
Long time no chat... I used to be a regular but got tired of the frustrations of Tri-Crossfire and crap AMD driver and game support to the point I literally gave up on it for a few years, but I am back.

I have built a new rig thats centered around (what I thought was) a bargain used GPU with waterblock installed.

The GPU:
EVGA RTX 2080 XC Ultra Gaming
https://www.evga.com/products/produ...ily=GeForce+20+Series+Family&chipset=RTX+2080

Cooler: (Installed by 1st owner)
EVGA Hydro Copper
https://www.evga.com/products/product.aspx?pn=400-HC-1189-B1

I built the rig up and whilst under load my 8700K CPU looked fine at 50c max during benching, but the GPU was averaging late 70's. I checked for air pockets etc but knew it had to come apart.


So, I stripped it down last night and over numerous assembly and dissasembly runs i determined that there is a significant air gap between GPU and the Block.

To clarify, I can take a strip of A4 paper, folded three times and slide it in there with it fully assembled and tightened down.

I thought maybe the ram pads were excessively thick for some reason but when I tried it without any ram pads it only just makes very light contact once bolted up if you use excessive paste to bridge the gap.

I have powered it up like this and temps are down to 60 ish now, but I am literally just bridging a gap with TIM and have no ram contact now bar the pads on the backplate side of them, which is far from ideal, and of course I know learn that this 2080 has no ram temp outputs for me to monitor!

So I got to thinking, If I buy new pads for the ram etc, can I just use say, a 1mm pad on the GPU instead of thermal paste? Not something i have ever tried, but cant see why it wouldnt work quite well.

 

[Demon]

As you eluded to, the instructions for the cooler show no thermal pads used on the RAM (based on the instructions: https://www.evga.com/support/manuals/files/400-HC-1189-B1.pdf) So any fitted would be jacking the heatsink off the GPU.

But if you say even with that there is still an appreciable gap, I'd check the bracket being used is correct as the instructions state there are 2 brackets, one for the 2080ti, one for the 2080/2070, not sure if it's hole spacing or GPU height that it might affect but worth checking.
In addition they state iten 11 are washers to go under the bracket, that might also space things off incorrectly..
I've seen people fit far too thick pads and actually bend the bracket mounting points a bit, just check the bracket is straight, it's an easy one to fix if it has.

I'd say that using a thermal pad on CPU or GPU is really not ideal to the point I'd contact EVGA customer services to chat it through with them and ensure it's the right cooler and to check which bits you need etc, they probably can point out any obvious issues.

 

[Stu @ MSD]

Thanks for taking the time to reply.

Sorry, I perhaps should have elaborated better. The pads come Pre-installed on the cooler. Just requiring the user to pull the stickers off, So in theory, cannot be wrong.

But with them in place I have a big air gap.
The CPU support bracket is the correct one and is lower than the die itself so can’t interfere.

All standoffs are in contact.
I just have this crazy gap.

My theory was to install a thermal pad on the die. With just far too much paste bridging an air gap it was in the 70s and not throttling so my thinking is that maybe with a pad, although not ideal it must still be better than an air gap and maybe salvage the situation.

Will also try and open an EVGA ticket.

This is the first and last time I buy a used GPU.

 

[D1craig]

An air gap between the block and the pcb is entirely normal. Their only needs to be no gap on parts that need cooling ie. Vrm, vram, gpu etc.

 

[MikeTimbers]

Use thinner pads. Try without pads.

 

[Stu @ MSD]

Sorry for the radio silence, been busy with work and my bloody backs gone. Not a good week...
As an update in case anyone else comes across this problem.


i took some tools home, a (micrometer and some feeler gauges) and found that ALL the installed pads were too big.


Hard to say what size size they were uncompressed, but compressed they were still approx 1mm on ram and 1.7mm on the VRM.


Seemingly the VRM pads were still holding the block high while assembled on the first test on day 1, making me believe the block was machined wrong. Stupidly, I only removed the ram pads as they surround the GPU.


With all the pads stripped off, I had GPU contact, so I measured up as best I could from the edges and had about 0.75mm gap on the ram chips and about 1.5mm on the VRms, so I installed some 1mm on the ram and 1.5 on the VRM and hoped for the best.


I cant get a feeler in there with it assembled with MX4 and temps are back down to sensible levels although my loop only has one radiator at the moment as I am tired of draining it for this damn card issue.


I think I am out of the woods.