Graphics Cards

[bittersweettreat]

Hi

I have a problem with my AMD RX 5700 XT

It started when I did a bit of a benchmarking using FurMark 3D test. At first, I had around 240FPS at 1280 x 720 resolution. I soon noticed the FPS starts to drop significantly. I then monitor it on the AMD Driver software and saw my junction temperature was around 110C while my GPU temp was around 65-70C. My GPU clock speed was limited to 1.2Ghz and I couldn't make it go higher than that. Further more, games like even Minecraft but with custom shaders started to crash. I already had this GPU since 3 months ago, and it's starting to fail as time goes

[morelock]

Hard to say without a bit more information. Is it a reference card, or a custom? How is your case cooling?
Have you tried undervolting it? Start with the stock undervolt profile in Radeon Software (under the performance tab) and see if that helps.

[bittersweettreat]

Reference, actually. It's a MSI MECH OC RX 5700 XT. I did some Undervolting, lowering volts to as low as 1050mv. It made it run a bit cooler but makes errors often. I had it somewhere around 1100mv, making my card run cooler while it won't affect the max boost clock too much and not causing some graphic error on games and such, but then the AMD Driver software starts to crash when I run AAA Games for around 15 Minutes more less. It would force any game I play to close, then resetting my Undervolting settings to back where it started at first

[morelock]

In gaming do you have a frame cap set? My 5700 would crash randomly when uncapped in some games. Worth trying setting the cap a few frames below monitor refresh.

There are some video's on youtube of some of the reference cards having some poor application of thermal paste and also you can see if the "washer mod" might apply to your card. How is your memory temp? Some of the MSI cards had heating issues on the memory.

[bittersweettreat]

Memory temp on idle is around 40C. On gaming it's stable around 70C. And no, I don't cap my FPS. 

[elstaci]

First the RX 5700 series GPU card are engineered to run normally with temperatures up to 110C. The GPU card has several Thermal sensors though out the GPU card.  If any of the Thermal Sensors reaches 110C it will automatically throttle the GPU card to keep it at or below 110C.

Also you must remember that Furmark Stress test puts an excessive amount of Stress on a GPU card. Something you normally would not see under normal heavy loads for the GPU Card.

[bittersweettreat]

I'm aware that the RX 5700 XT getting as hot as 110C is how it's designed to be. But getting that high while the clock speed is far below the Boost clock is quite absurd. It doesn't only get as high as 110C while being at relatively low Clock speeds, but also on AAA games, it thermal throttles almost instantly. 

[elstaci]

I would open a MSI Support ticket and see it the GPU card needs to be RMAed.

[fyrel]

I had the same issues with my XFX 5700xt junction temperature maxing out.

It seems the fans profile wasn't running correctly.

You could try setting up a software fan profile to see if it helps you.

[bittersweettreat · Accepted solution]

[Update]
My problem has been Fixed.

I open up my GPU to see if there's any Thermal conductor issue and noticed my Thermal paste wasn't applied well. So I replaced it with a dot of Thermal paste from Deep Cool, which is the Z5 (Not a very high performance Thermal paste but it's the only Thermal paste I have).

Ran some benchmarks and temps were quite stable. my hot spot temp (or Junction temperature) was going no more than 100C at 2Ghz. FPS stays stable over 60FPS on most games I played. And my fans which previously will run often over 3000RPM now stays around 2500RPM. Unless overclocked to 2.15Ghz @ 1.2mV. Core clock often stays at 2.1Ghz while gaming AAA title games, sometimes going down to 1.6Ghz if the game's environment isn't complex. Overall, it has been a lot better now. I had also re-installed my AMD driver to see if my last version might be full of bugs that causes it to often crash randomly when gaming. And it did fixed it.

So my conclusion is, if GPU runs hot quickly, check its thermal conductor. Try to replace it when you can and see if it fixes

Then update your Drivers to the latest version of it when available.

I thank to all those who had given me their solutions or explanations. I appreciate to all those that had tried helping me on previous replies

[prospector]

I'd like to bring this thread up again as I am having the same issue.  Problem is I don't know where to look in my GPU to see if I need to put more thermal paste or not.... 

I have built my own computers before, just fyi  Can someone assist me here please?