Just finished stress testing the GPU after repasting and repadding it.
First of all, for the next fella who tries this a searchable string:
Reference RX-480 uses all 1mm thermal pads.
RedDevil RX-480 is essentially “reference” with regards to thermal pads.
Pre-repasting/repadding this thing was idling in Windows at 45C and WoW on high settings was pushing it to 85-90+ and into shutdown (shutting down about 4 times a day).
Post-repasting/repadding I was seeing startup temps of 30C and after shutting down the stress test the temps returned to 35C (I’m sure there’s some residual heat in there that affected that).
The paste itself was a brick - hard as a rock - also it only covered about a third of the main chip - sloppy installation to begin which probably didn’t help much (if it matters - it might not).
The pad (a large single one) over the power control chips was intact and not oozing oil so I left it. The others ripped when I removed the cooling block.
I cleaned everything with 99% isopropyl alcohol and a combination of cotton wool, swabs, and blue shop towels and cut and installed new replacement pads.
There was no discernible dust, dirt, molted bug husks, or anything else in the fins, fan blades, or anywhere else except for a little surface dust on the inside of the interface plate which I cleaned off. I do a monthly blow-out with an air-compressor and a fine-point nozzle and that seems to be working well.
I repasted with the Thermaltake paste that the local shop carried (they had paste, they didn’t have pads). Don’t recall the particular variety if Thermaltake has multiples, but I’ve used this before and never had a problem.
I may have overapplied it a bit, but it’s non-conductive and I’d rather have too much than too little (and it wouldn’t have been MUCH too much).
When I was done with the front of the PCB, I looked at the back of it and measured the distance between the top of the back-plate standoffs. 2mm almost exactly (probably really about 1.8 but my calipers might have been off by that much) and decided to try stacking two layers of 1mm pad behind the main chip and the backplate to try to alleviate the “the backplate is causing more problems that it’s solving” issue I’ve read about with this card.
No clue if it’s effective or not, but it didn’t break it and it seems a reasonable solution (any cooling there is better than just dumping that heat into an air pocket between the PCB and the backplate).
Benchmarking and the Stress Test:
No measurable change in actual card performance, other than it’s not crashing.
I reset the fans to the defaults (I’d had them essentially capping speed at anything over 75C before in an attempt to get a bit more time before it hit its thermal limit)
The absolute highest sustained temps I could get pushing the stress test as far as I could for half an hour was 62C with very short, less than 5 second, spikes to 63C.
I had two different heat monitors running and they agreed with one another well enough that I shut one of them off (just too much clutter to watch all of it).
For now, I’m gonna leave it alone. “Fix it until it’s broke again” ain’t in my repertoire as far as risking my GF’s computer goes. I still kinda like her.
Thanks all!