Question Sapphire Pulse RX5700XT Be - repaste issue

Aug 20, 2020
38
1
35
0
I bought a Sapphire Pulse RX5700XT Be about three months ago. It's slightly different to the normal Pulse in that it doesn't have the dual BIOS switch and it has one less heat pipe.

I replaced the thermal paste about a month ago just out of curiosity and with that and an undervolt, I managed to get it to a max of 66 deg edge, 76 junction locked at 60FPS at 1440p for hours of gaming which I was pretty pleased with.

Because I've been enjoying the tweaking, I decided next to try to replace the thermal pads. In doing so, I've done another repaste (with the same tube of almost new MX4) and the junction temp shot up to 105 almost instantly in a Heaven benchmark. I redid it again and it was even worse. Turned out I'd missed a couple of corners on the die with the paste. Redid it a third time with more paste and now it's slightly better but the difference between edge and junction is at least 30 degrees at all times e.g. 70 and 100. I have noticed a couple of degrees lower on the memory temps though, from 66 to 64 with the new pads, but then I've not been able to run it for longer than five minutes!

Gutted. Any suggestions of getting the temperatures back to where they were?
 
Dec 18, 2020
4
0
10
0
As far as I know, when it comes to those thermal pads, they have to be pretty on point with manufacturer standards. A little too thick, and your GPU won't touch your heat sink, too thin, the pads won't disperse heat correctly. Best of luck to you mate.
 
Aug 20, 2020
38
1
35
0
As far as I know, when it comes to those thermal pads, they have to be pretty on point with manufacturer standards. A little too thick, and your GPU won't touch your heat sink, too thin, the pads won't disperse heat correctly. Best of luck to you mate.
Thanks! I need it!! I can tell from opening it up for the fourth time that the pads are making good contact with the memory and other bits and they seem to be doing their job.

Problem is the gpu die/paste. Going to post a photo of my latest effort that is better but still no good compared to before.
 
Aug 20, 2020
38
1
35
0
Take the cooler off and repaste it again.

Sometimes when you repaste it and reassemble the mounting isn't quite right, and remounting it solves this issue.
I'm about to do this a fourth time. Could it be anything to do with the heat sink clamp around the gpu on the back of the PCB? Understandably it would lose tension after multiple replacements - if I make it tighter would this be better for temperatures? More contact with the copper plate?
 
Dec 18, 2020
4
0
10
0
You really should make sure that your new pads are not preventing your heatsink from making contact with your GPU before trying to brute force contact.
 
Aug 20, 2020
38
1
35
0
You really should make sure that your new pads are not preventing your heatsink from making contact with your GPU before trying to brute force contact.
I'm fairly sure they aren't. They are 1.5mm pads and replacing what looks like some 2mm and some 1.5mm pads. They all have an indent on them.

It's only the little silver die surface that needs to contact the heatplate via the paste, right?
 
Aug 20, 2020
38
1
35
0
Just the die, not the silver surrounding it.

You really should make sure they are the exact thickness and not just guess at the size they are replacing. If they are too think they might not contact it properly either.
Thank you, I do appreciate that but would that change of contact explain why just the junction temp has gone up, rather than the average?
 
Yeah, I would also suspect it might be an issue with the thickness of the thermal pads. If the pads are too thick (or maybe more dense, so they don't pack down as much), there might not be adequate pressure between the heat sink and the graphics processor. The thermal compound might still spread, but it may be too thick of a layer to provide an optimal transfer of heat.

And as NightHawkRMX suggested, it could be uneven mounting pressure, possibly from replacing two thicknesses of thermal pads with a single thickness.
 
Aug 20, 2020
38
1
35
0
Thank you. I've tried one more different method of paste application - a thin layer spread across with a credit card - and have remounted everything as is, not too tightly. Will see how it is tomorrow and if no luck will get some 2mm thickness pads. Cheers for all your help.
 
Aug 20, 2020
38
1
35
0
Pleased to say the next morning that this has worked!

Only thing I did differently was the thermal paste application but this time I made sure that each corner of the die was covered from the top surface to the pcb. Now getting 59 average, 71 junction on a stress test.
 

ASK THE COMMUNITY