Never ending black screens and sudden restart woes on my PC.

[emothxughts]

PC specs at time of writing:

Processor	AMD Ryzen 5 3600
Motherboard	ASUS PRIME B450M-K II (AM4) @ American Megatrends Inc. 3802 (04/28/2022) BIOS
Cooling	Deepcool GAMMAXX 400 V2 64.5 CFM CPU Cooler
Memory	PNY Electronics 8192 MB (DDR4-2933 with XMP/DOCP) P/N: 8GBF1X08QFHH38-135-K (x2)
Video Card(s)	GIGABYTE RX 5600 XT Windforce OC 6GB, 2 fans version, Sapphire Pulse RX 580 4GB, @ Adrenalin 22.11.2 WHQL
Storage	HP SSD EX900 500GB, PNY CS900 960GB
Display(s)	LG 24MP400
Power Supply	Cooler Master MWE Bronze V2 650 W 80+ Bronze Certified ATX Power Supply (230V non fullrange)

I previously made two separate threads about my computer's graphics hardware acting up, 1 and 2. Long story short, two of my graphics cards, RX 580 and RX 5600 XT are not working right.

My RX 580 keeps black screening (display output cuts off) with PC, keyboard and mouse lights still on and the sound of games still on, and this forces me to force restart my PC by long pressing the power button. This occurs randomly in games and sometimes even outside games. I have sent this card for repairs, the repair shop said my HDMI ports were broken and they fixed it. This black screen problem happened again 2 weeks after I got it back.

Next, my RX5600XT card makes my display blank black for a few seconds, then my PC restarts. Upon restarts, I get this in Event Viewer, where Processor APIC ID is different every time:

A fatal hardware error has occurred.

Reported by component: Processor Core
Error Source: Machine Check Exception
Error Type: Cache Hierarchy Error
Processor APIC ID: 10

The details view of this entry contains further information.

At first, I thought it was my CPU that's at fault, so I RMA'd it. It happened again with the replacement CPU. So I thought the GPU was at fault, I never had this problem before I got the RX 5600 XT. So I then RMA'd the GPU. 2 months later, I got it back today, installed it again, managed to play Genshin Impact and modded Skyrim for 1 hour each with no problem. Then when I played Destiny 2 for 15 minutes, my PC restarted again, exactly like it did before I RMA'd my CPU and GPU, as if it did absolutely nothing.

I am really at a loss here, I am unable to enjoy playing games on this PC anymore, CPU & GPU RMAs and repairs did nothing to solve my problems. Now I'm beginning to think the other components of my PC are at fault here, like maybe the motherboard, RAM or even the PSU, which is why I turn to you, the experts here at TPU. It is also why I decided to make this thread, because the problems with my RX 580 and RX 5600 XT are related to some other parts of my build somehow. Help would be welcome!

 

[Shrek]

Have you tested the RAM?

Do you have another power supply to try?

Update to BIOS 3810

 

[Sithaer]

If all else fails I would try another PSU cause I did have a Cooler Master MWE 650W Gold PSU a few years ago that developed a weird issue when it out of sudden turned off/restarted my PC and then it worked fine for the next week or so. 'spent a good ~2 weeks troubleshooting that damn issue trying everything I could..'
Tried a different PSU and the issue went away.

Tho my model was the first edition 'MPY lineup' not the V2 but still worth a check if possible.

 

[emothxughts]

Have you tested the RAM? Do you have another power supply to try?

Update to BIOS 3810

1. 1 have tested RAM in the past with memtest86+ and one in-Windows test that I forgot the name, but one of its configs was named Absolut. Didn't throw any error. That said, what tests would you recommend?
2. My only other PSU is a 450W 80+ white PSU from AVF, not even on the tier list. Might as well say I don't have another PSU.
3. Just updated to 3810 from 3802 now. What is the exact difference, besides "increase compatibility"? What does that even mean?

 

[Kodehawa]

Sounds like a PSU issue to me. Failing PSUs will do this.

Another option as someone else said is RAM, but I'm betting more on a PSU issue, as the PSU failing to deliver the appropiate voltage and amperage under load (or at idle even, I remember Vega cards blackscreening on idle if the PSU couldn't handle them) will cause the hardware to trigger a safety restart, as the hardware can't operate correctly.

 

[Shrek]

Are you running both video cards at the same time?

 

[emothxughts]

If all else fails I would try another PSU cause I did have a Cooler Master MWE 650W Gold PSU a few years ago that developed a weird issue when it out of sudden turned off/restarted my PC and then it worked fine for the next week or so. 'spent a good ~2 weeks troubleshooting that damn issue trying everything I could..'
Tried a different PSU and the issue went away.

Tho my model was the first edition 'MPY lineup' not the V2 but still worth a check if possible.

Speaking of PSU, do you think a faulty PSU can cause my GPUs to go from x16 to x8 lanes and then back to x16, as is the case with my RX 580? It does that without me reseating the card or even touching it.

 

[Assimilator]

Considering you've had the CPU replaced, and you say you've verified the RAM passes MemTest, that leaves either the board or the PSU. I'd strongly suggest using your backup PSU just as a smoke test, if the problems persist it's probably the board, if not it's the 650W PSU.

 

[emothxughts]

Are you running both video cards at the same time?

No. Couldn't even do that if I wanted to (I don't), since my mobo has only 1 GPU slot.

 

[Kodehawa]

1. 1 have tested RAM in the past with memtest86+ and one in-Windows test that I forgot the name, but one of its configs was named Absolut. Didn't throw any error. That said, what tests would you recommend?
2. My only other PSU is a 450W 80+ white PSU from AVF, not even on the tier list. Might as well say I don't have another PSU.
3. Just updated to 3810 from 3802 now. What is the exact difference, besides "increase compatibility"? What does that even mean?

The other PSU is likely good enough for a test. The combination of hardware you're running isn't gonna put any strain on it.

 

[emothxughts]

I'm tempted to think my PSU is the problem (it's tier C in the tier list, my setup seems to require something in at least tier B). Before I sink my money into yet another PSU, I need to know first, can Cache Hierarchy Errors be caused by PSUs?

 

[Kodehawa]

I'm tempted to think my PSU is the problem (it's tier C in the tier list, my setup seems to require something in at least tier B). Before I sink my money into yet another PSU, I need to know first, can Cache Hierarchy Errors be caused by PSUs?

Hardware won't operate correctly with a faulty PSU since its gonna be receiving all sorts of weird voltage or extreme amounts of ripple or amperage variations its not made to handle, leading to all sorts of weirdness (like this one). You already ruled out everything but PSU and board, so you have those two to test.

Use your spare PSU. It's gonna be alright for testing, assuming you know it works.

 

[Sithaer]

Speaking of PSU, do you think a faulty PSU can cause my GPUs to go from x16 to x8 lanes and then back to x16, as is the case with my RX 580? It does that without me reseating the card or even touching it.

That I cannot say since I'm not sure about that.
I agree that your secondary PSU should be enough for a test, I also only had a cheapo budget Thermaltake 80+ 450W spare PSU around and it was enough for me to narrow the issue down to my 650W PSU.
At the time I was using a R5 1600x+RX 570 so your spare PSU should be okay too for a test run.

 

[emothxughts]

Hardware won't operate correctly with a faulty PSU since its gonna be receiving all sorts of weird voltage or extreme amounts of ripple or amperage variations its not made to handle, leading to all sorts of weirdness

Does a bad PSU also make a graphics card go from x16 lanes to x8 lanes and back too, like what happened to my RX 580? Also, my 3rd (emergency) card, the RX 550 which does not use PCIe cables, also goes from its initial x8 lanes (it really only has x8 lanes) to x4 lanes and then sometimes back.

 

[Kodehawa]

Does a bad PSU also make a graphics card go from x16 lanes to x8 lanes and back too, like what happened to my RX 580? Also, my 3rd (emergency) card, the RX 550 which does not use PCIe cables, also goes from its initial x8 lanes (it really only has x8 lanes) to x4 lanes and then sometimes back.

I cannot say, but you have the means to test with your spare PSU. It *could* also be a bad board, but you gotta rule out what you can test with now first.

Also, could just be Link State Power Management doing its job if its not under load.

 

[Norton]

Does a bad PSU also make a graphics card go from x16 lanes to x8 lanes and back too, like what happened to my RX 580? Also, my 3rd (emergency) card, the RX 550 which does not use PCIe cables, also goes from its initial x8 lanes (it really only has x8 lanes) to x4 lanes and then sometimes back.

That sounds more like a motherboard or chipset driver issue- I would try a newer/older chipset driver first if I was having a similar issue

 

[emothxughts]

About the CPU, for the past month I've been running Stable Diffusion on my R5 3600 CPU (because I don't have an NVIDIA GPU), the thing hammers my CPU for hours on end and raises my CPU temps up to 77C, yet it never goes WHEA 18. It only WHEA 18's when it is paired with the RX5600XT and is playing a game, which is less intensive than Stable Diffusion generation. So whatever is causing this, it's most probably not the CPU itself...I think?

 

[Kodehawa]

About the CPU, for the past month I've been running Stable Diffusion on my R5 3600 CPU (because I don't have an NVIDIA GPU), the thing hammers my CPU for hours on end and raises my CPU temps up to 77C, yet it never goes WHEA 18. It only WHEA 18's when it is paired with the RX5600XT and is playing a game, which is less intensive than Stable Diffusion generation. So whatever is causing this, it's most probably not the CPU itself...I think?

I was under the impression you ruled the CPU and GPU out on your other posts.

There's a few options really:
- Chipset drivers are making it go mayhem (maybe? as stated above it's a possibility, could also test it on a Linux LiveCD)
- Board is toast.
- PSU is toast.
- PSU is toast *and* toasted some components along the way.

The fourth one would be the worst one, but also the most unlikely -you still have a system that turns on- (but could happen, for ex. making the VRMs or MOSFETs on your GPU only act up under load)

 

[emothxughts]

What I've done so far is update my BIOS to the newest one, version 3810. It's already bedtime and I'm too mentally exhausted to play a game right now, if it means it will only blackscreen and restart my PC yet again. I'll try playing a game after sleep, see if this problem doesn't occur again. (I'm using the RX5600XT right now.) In the meantime, feel free to reply to and discuss this thread.

 

[TheinsanegamerN]

Does a bad PSU also make a graphics card go from x16 lanes to x8 lanes and back too, like what happened to my RX 580? Also, my 3rd (emergency) card, the RX 550 which does not use PCIe cables, also goes from its initial x8 lanes (it really only has x8 lanes) to x4 lanes and then sometimes back.

Like was said before, if your psu is sending out weird voltage, it can cause all sorts of issues. Change out for the spare and see if the behaviour continues

If it continues then the motherboard could be the problem. This can be frustrating, I chased a similar issue for three years and eventually found that it was an arcing trace on my vega 64 that was making the entire system go into a panic. Found that out by total mistake

 

[Norton]

A test through Linux like @Kodehawa
mentioned might be a helpful step.

Check your power plan/settings also, chipset driver and Windows updates sometimes change settings and mess with stuff. I run into this from time to time when running my crunching rigs, which run full load 24/7

 

[emothxughts]

How would I do the test through Linux as mentioned? I have never installed Linux, but this probably doesn't need a fancy distro, right?

 

[Udyr]

I'm tempted to think my PSU is the problem (it's tier C in the tier list, my setup seems to require something in at least tier B). Before I sink my money into yet another PSU, I need to know first, can Cache Hierarchy Errors be caused by PSUs?

I believe the other kind members have provided sufficient troubleshooting guide for your issue, so I'll give you some word of advice:

Take that list as a reference, not a PSU bible. Don't base your system requirements on whatever the list says your system "needs". Get a professionally reviewed unit that covers the power draw and a little extra.

 

[emothxughts]

While reading up on my unfortunate situation, I came across this Reddit thread claiming that RAM is to blame for PC builds crashing in games. Seems to line up with my own experience, where the crashes and restarts only happen in games, never in stress tests and Stable Diffusion runs. While I still can't discount motherboard or PSU as the problem area, this is another lead in my unfortunate mystery. What does TPU think?

Also, another thing I tried is decreasing power limit in AMD Adrenalin so the card consumes 75W or less power. It is indeed as slow as you'd expect in Heaven and Superposition (haven't actually played a proper game yet), but at this point I'd rather value stability over blazing speed.

 

[Kodehawa]

While reading up on my unfortunate situation, I came across this Reddit thread claiming that RAM is to blame for PC builds crashing in games. Seems to line up with my own experience, where the crashes and restarts only happen in games, never in stress tests and Stable Diffusion runs. While I still can't discount motherboard or PSU as the problem area, this is another lead in my unfortunate mystery. What does TPU think?

Also, another thing I tried is decreasing power limit in AMD Adrenalin so the card consumes 75W or less power. It is indeed as slow as you'd expect in Heaven and Superposition (haven't actually played a proper game yet), but at this point I'd rather value stability over blazing speed.

The 5600XT can probably be undervolted to use a lot less power than it does stock, and you can probably lower clocks a bit to achieve even more performance (efficiency curve moment).

Try Y-cruncher for RAM if memtest86+ doesn't find anything. I'm still on the PSU camp, though. (Also if it crashes it very much could be the PSU too)