[SOLVED] New Build - RX 580 Failing During Stress Test

[_Makky_]

I made this build for a friend of mine and for some reason the GPU would fail withing 5-10 minutes of gaming or 1-2 minutes of FurMark stress testing.

Specs:

Part	Model
CPU	AMD - Ryzen 5 3600
Motherboard	Asus - PRIME B450M-A Micro ATX
Memory	Corsair - Vengeance LPX 8GB DDR4 3000MHz C16 (CMK8GX4M1D3000C16)
Storage	Samsung - 970 Evo Plus 250 GB M.2-2280
Storage	Western Digital - Caviar Blue 1 TB 3.5" 7200 RPM
Video Card	Sapphire - Radeon RX 580 8 GB Pulse
Power Supply	Cooler Master - MWE 550 W 80+
OS	Window 10 64 bit

I have been trying many many things from trying many old and new drivers for 4 days and then finally getting GPU replaced via my dealer.

This happened even after getting a new replacement from my dealer. The GPU won't fail if I manually set my GPU fan speed to 100%.
The GPU fails on stock settings at temp of approx 72 C, however, won't fail at 100% fan speed as the GPU maintains a temp of 60-61 C.

I have seen GPUs handling temp above 90 C and not failing. Is this RX 580 behavior normal or am I doing something wrong or have I got another faulty GPU?
I really need some guidance here, please

 

[jay32267]

I'm leaning on that there might be a problem with the GPU.
I don't think the GPU should fail on "auto".
....and I don't think it should fail at 72C.

 

[_Makky_]

I'm leaning on that there might be a problem with the GPU.
I don't think the GPU should fail on "auto".
....and I don't think it should fail at 72C.

I know right, I have never seen any GPUs ever fail at stock setting what so ever.

Any ideas on How do I be absolutely sure that it's a GPU issue so I could be confident while talking to my dealer coz he has been dumping bad GPUs on me.

 

[jay32267]

"Any ideas on How do I be absolutely sure that it's a GPU issue "
I think that the evidence you have is pretty good as is.
Maybe take a video of it failing (if that's possible) to show the dealer.

 

[King_V]

It MIGHT be the power supply. I don't know a whole lot about it, but doing a little digging, on this page, which, like here, is a forum, someone posted the following:

MWE 550W is made by HEC, and it uses a double forward, group regulated topology - the same outdated principle as the Seasonic S12II, which doesn't respond well to load from modern PCs. The MWE line is tricky, because only the 600W and 650W models use DC-DC converters, while 550W and everything below uses group regulation.

This might suggest maybe the PSU is struggling at the extreme end. Now, you do mention that temperatures seem to set it off, and with the fans at 100%, it behaves at max load.

That said, I'm given to understand that as things get hotter, power draw will increase. This also makes me wonder if it's due to the PSU, because 72 C is not dangerous territory at all for GPUs. Plus, while nobody is perfect, Sapphire is generally well-regarded in terms of their GPUs and coolers.

Do you have a known, high quality PSU that you can swap in, and then repeat the same tests?

 

[_Makky_]

It MIGHT be the power supply. I don't know a whole lot about it, but doing a little digging, on this page, which, like here, is a forum, someone posted the following:



This might suggest maybe the PSU is struggling at the extreme end. Now, you do mention that temperatures seem to set it off, and with the fans at 100%, it behaves at max load.

That said, I'm given to understand that as things get hotter, power draw will increase. This also makes me wonder if it's due to the PSU, because 72 C is not dangerous territory at all for GPUs. Plus, while nobody is perfect, Sapphire is generally well-regarded in terms of their GPUs and coolers.

Do you have a known, high quality PSU that you can swap in, and then repeat the same tests?

I did think of it in a similar way. Then I stress tested CPU and GPU simultaneously with 100% fan capacity for GPU.
This setup did not trigger a fail.

Another setup I tried was:
Plug GPU in another build, however, powered the GPU via above-mentioned MWE 550 PSU << This setup triggered a fail.
This is another test that makes me lean towards a faulty GPU.

I do not have a confirmed good quality PSU > 550W. I will be taking this system to a repair store 18 hours from now.
This will indeed bring more clarity.

Will keep this post updated.

 

[TJ Hooker]

Define "fail". Driver crash? Blue screen? PC abruptly shuts off?

 

[_Makky_]

Define "fail". Driver crash? Blue screen? PC abruptly shuts off?

GPU stops giving out any information to the screen (blank black screen). << At this state I can press Alt + F4 to stop FurMark Stress test and the screen would be back in maybe 10 seconds after closing the stress test.

If I do not blindly close the stress test windows. The computer would freeze (numlock/capslock stop responding) after some more second and then restart after maybe some 15-20 minutes.

 

[TJ Hooker]

Hmm, as said above ruling out the PSU is probably a good next step. Did you do a clean install of Windows for this build? Have you tried uninstalling graphics drivers (using DDU) and then reinstalling?

 

[tennis2]

I'd be getting an RMA ticket started.