GPU ram/chip heat pads?

[jimbouk]

Hello,

My 'trusty' GPU (RX 580) has started borking out playing games; black screen (Discord chat was still working away), default Microsoft drivers after a hard power off/on. I've just done a disassemble of the card and will give everything a good clean and new thermal paste to see if I can squeeze another 6 months out of it (stolen my wife's R9 285 - fortunately not playing anything too modern!).

However there are 8 ram chips and 4 others (power?) which use pads rather than direct contact / thermal paste. Has anyone replaced these with something reputable for a reasonable amount of money? Or should I just save my pennies and get a new card?

Cheers!

[Domestic_Ginger]

You can replacements off ebay for a few quid, you need to get the correct thickness. Which card do you have? GPUz or HWinfo will read sensor info to see if heat is the problem. you can log it or flick back and forward to check max values.

You can undervolt in catalyst. The 580 is a juiced up 480, I think they are about 1.25V on the core. It 'should' run at about 1.15V. I would knock 0.1V off all the states except the lowest 2 and then stress test.

EDIT: if going ebay route, check location as a lot of the Chinese sellers state the items are in the UK when they are not.

[jimbouk]

Not sure how my calipers would work with a soft pad Looks about 1mm... The model is Sapphire Pulse Radeon RX 580 8GB GDDR5.

Cool yeah - I should get some logging on and crash it again! Will have to put it together again first. Guess I'll just try with the same pads and see.

It was "Used - Acceptable" over 2 years ago so it's probably had it's time. Still not enjoying the current prices though, don't really want to go over £200.

[jimbouk]

Well it's back together with the original pads and be thermal paste. No issues running the standard desktop. Not tried a game/load test but at least I didn't brick it even after the running the heatsink and fans/shroud under the tap - don't try this at home kids!

Another gaming session planned on Monday but away on the weekend so hopefully will have a quick chance to test before then.

[cptwhite_uk]

I've seen 580 8Gb models going locally for £100 should you need to replace, it's not too expensive anymore. Although refurbished RTX 2060 are available for less than £200 online now, so I'd probably be tempted to upgrade...

[jimbouk]

Yeah, the games we currently play are far from new so no issues running 1440p ultrawide. I hate to think what would happen in a new game though! I was taking a look at the tomshardware graphics card hierarchy to try and work out where everything was placed at the moment. And it is an odd state for PC components for second hand to be such a big market, mining had really messed things around.

[jimbouk]

3D Mark 11 is currently on loop maxing out my GPU. 100% load and the only temperature I can see in CPUID HWMonitor is 70°C max. Memory load isn't that high in case it's a ram chip on the way out (or cooking) but so far so good and hoping this will see me through for the foreseeable. I'll update if it all goes wrong again on No More Room In Hell later this evening!

[[GSV]Trig]

Try GPU-Z, will tell you a lot more about the card, well, as far as reporting what the sensors are reporting, so depends on what sensors the card has, core, ram, hotspot, power draw etc.

[jimbouk]

All stable

Not sure why I ended up with CPUID HWMonitor - I think I was just doing a before after when I got a new CPU cooler many moons ago. Cheers for the tip though, downloaded for next time.

Now to sort out the other half's old AIO to save having to use my laptop as a server...