Why are my GPU's dying?

I got an RX 5700 Red Devil two weeks ago, flashed it with an XT BIOS and undervolted it slightly and it was totally stable (TimeSpy for ~20 minutes) and cool (~70°C).Applied a Windows update from the slow ring on Monday that was required for 20H1 as I've been trying to reset Windows due to another issue, and it was causing the driver to hang and recover a few times on the login screen. That evening the PC went to sleep then it never displayed anything again, regardless of CMOS reset or trying another PCI-E slot. The VGA light on the motherboard was lit. Tried it in another PC, no display either.

I was kind of content the flash may have killed it (although the card was cool and I wasn't playing a game when this happened). I setup an RMA regardless, they didn't have any in stock so they said they'd refund me.With the lockdown in effect here, I didn't want to wait bored senseless and unable to do coursework so I bought another one from a different supplier. This flickering on logon still happened, but I reinstalled Windows and it was fine. I booted into the BIOS, disabled CSM support to enable safe boot and then I have no display again. This was only 12 hours after I put the card in. Put in my sisters GTX 1660 Ti, it POSTs, revert the setting, I swap the 5700 back in. Nothing...?

How can this happen twice? Can a Windows update from the insider rings actually kill hardware? Or is this a reported problem with Navi? I have read random black screens, but not failing POST after sleep or a reboot.Called CS from one of the suppliers and he said that he's aware of problems. I even said "yea, I know the drivers have been bad, but I know nothing about cards not POSTing" and he said that he's heard of customers having said problem.

Do you guys think this is a Navi issue? Or am I just supremely unlucky?

- - - Updated - - -

Am I experiencing the first computer problem in history that no one can answer? I've posted this to the amdhelp subreddit as well and no one has responded.

bobkn said:

I doubt that your problem justifies teh Trumpina superlative.

I'm sorry, the problem just exceeds my knowledge and it doesn't seem like anyone can make sense of it.

bobkn said:

I have had exactly two video cards expire on me in the past 25 years. One was an excessively overclocked TNT2. (Heard of that one?) The other was a GTX 680. (Well, I also had a Voodoo 1 card with a bad PW board trace.) I prefer nVidia cards, but I've used a few AMD/ATI cards, including a 7970.

My very first GPU was a reference HD 6950, the fan sounded like a lawn mower. The RMA replacement was a card that, upon cold boot, displayed artifacts on screen and couldn't play certain games or the drivers would hang. My first nvidia GPU, an EVGA GTX 670, fried itself with no overclock after like 1.75 years. The RMA replacement for that, a GTX 770 SuperClocked, wouldn't display anything in my computer, but would in another.

An R9 Fury and RX Vega 64 are the only two cards that worked until I sold/replaced them. So I have had amazing luck with hardware.

bobkn said:

I have never seen a card damaged by drivers or Windows updates.

It makes no sense to me either, I'm just going by this post on Tom's Hardware. The fans on the card do spin, it also doesn't display anything on the second vBIOS.

bobkn said:

Maybe it's the Universe, telling you to stick with nVidia. (Even though the AMD cards appear to offer a better price/performance.)
As for Navi, what's that? (Some gamer thing, I suppose.)

I wouldn't mind getting an RTX 2070 Super, but it's £150 more expensive for similar performance. If it means having a working computer it obviously doesn't matter, but I'm trying to make sense of the problem to make sure it's not other hardware causing this to happen or something.

Navi is the architecture of the RX 5000 series GPU's.

Supra said:

(snip)

It makes no sense to me either, I'm just going by this post on Tom's Hardware. (snip)

(snip)

Navi is the architecture of the RX 5000 series GPU's.

A software update that messes with a graphic's card's BIOS? I can't swear that it's impossible, but it would be very bad news. I hope that it's false speculation by someone whose level of knowledge is no better than mine.

Sorry that I didn't recognize Navi. I'm more into physicists and mathematicians (nVidia series).

I just tried again with the second vBIOS (there's a switch on the card), without any display cables plugged in or with just one. This is now also with two separate VGA power cables instead of one that splits, just in case.

Here's a streamable of what happens.

For reference



I can't really test this component in another system as I'd have to go to my sisters, and the UK is in lockdown with a certain thing happening. I'm sure it's dead, but obviously don't know why and don't know if I should risk getting another one.

TV2 said:

The problem with a lot of computer hardware failures is they rarely just blip out and go dead, but instead linger with odd behavior and strange glitches.

The problem is, the card ran fine for 2 weeks. Then I updated Windows, and the driver started hanging at the login screen. Could be related or could not. Then it went to sleep and didn't display anything.

Got another card from a different supplier and it also stopped displaying anything. I reckon I wouldn't be in this situation if I didn't install that insider update (one required for 20H1).

Supra said:

The problem is, the card ran fine for 2 weeks. Then I updated Windows, and the driver started hanging at the login screen. Could be related or could not. Then it went to sleep and didn't display anything.

Got another card from a different supplier and it also stopped displaying anything. I reckon I wouldn't be in this situation if I didn't install that insider update (one required for 20H1).

Can you thoroughly examine the card ? Most of the time dead hardware have signs that may reveal the cause .

If you can notice yellowish color on the board around a certain part it may reveal its in particular the part that fried , so if its surrounding a capacitor then probably the problem might be your board or a failing PSU , if it surrounds the GPU seating then the problem might well be heat , where we can suspect either the cooling on these are weaker than processing heat and maybe this is a result of bios failing / glitching that would slow down the cooling system along since it controls it . if no visible defect then it gets harder but you can blame the graphics bios .

Also use an air blower to clean the slot , there might be metal debris stuck some where inside that may lead to a short circuiting .

Why didn't this happen with Nvidia may be related to different tolerance of capacitors and usage of printed circuits on the board differs so the bursts of deadly electric pulses are not going to a highly sensitive part .

So bottom line Check card - Blow slot - Replace PSU so you are not trapped in this loop forever .

But then you repeat , but this only happened after 20H1 , well yes but this is on you still , its not a public build so it comes with no warranty what so ever (even public builds are not warrantied but at least Microsoft fears public rage if they have glitches like these) .

So all you can do is just eliminate other causes in hand and hope for the best .

Also remember , your system felt unstable which lead you to such update , so i highly recommend you do viral scans using Malwarebytes and any other reputable antivirus you like .

nIGHTmAYOR said:

Can you thoroughly examine the card ? Most of the time dead hardware have signs that may reveal the cause .
Also use an air blower to clean the slot , there might be metal debris stuck some where inside that may lead to a short circuiting .

This wouldn't explain why it also didn't work in the second PCI-e slot. I blew it out with o2 hurricane anyway.

nIGHTmAYOR said:

So bottom line Check card - Blow slot - Replace PSU so you are not trapped in this loop forever .

The PSU is an eVGA advanced RMA :/ I thought that was a problem before, which I'll go into in the next quote.

nIGHTmAYOR said:

Also remember , your system felt unstable which lead you to such update , so i highly recommend you do viral scans using Malwarebytes and any other reputable antivirus you like .

The problem that I was having before this was random restarting. My previous Vega 64 with my 650 watt power supply exacerbated it, but it still happened even with the 5700. I have narrowed it down to either the RAM or motherboard. I swapped RAM DIMMs around trying to get the PC to boot when the first card died, and when I got the second card, it said that one of them (socket 0) failed memory training. Weirdly, if I swap them back around, it passes and Windows recognises 16GB again. Doesn't make sense because I will do Karhu memtest for like 26,000% and it'll have 0 errors, but the PC is obviously unstable because it restarts playing games or doing Premiere/Photoshop etc.

Have an ongoing thread about it here.