Advice please - AMD Radeon RX 5700 XT - PC Restarts

[iameatingjam]

Another thing you can try, that helped me diagnose my cpu as the problem, was disabling cores, first I just disabled all ecores, and that fixed the problem, and eventually I was able to nail it down to a specific cluster of ecores (4 ecores). Mind you, this was 14700k, not an alder lake, which is known to be more stable. Still, something you can try. Like try enabling only one or two pcores and disable all ecores and see if you can boot and stay stable. If you can, keep enabling more and more until you find the point where you can't. If it does indeed work out this way, its very likely its the cpu.

If disabling cores doesn't help at all, then its probably not that.

 

[RJARRRPCGP]

Update... Tried another game (Diablo), all fine. But this wasn't stretching the GPU at all.


After this, I decided to try some new settings on the COD game... Resolution at 1080p instead of 4k and quality settings at Minimum. I didn't even load into the game, just left it on the main menu and the PC restarted after about 3 minutes.

Sounds like those fatal WHEA "Cache Hierarchy Error" crashes. There's a tendency for that to happen when the room is warm. Some CPU cores seem to be flaky at more than 55C! It can also happen, due to VRAM being OC'ed too much.

Update2: lots of different blue screens, even after removing the GPU. I'm now going through the process of reinstalling windows again!

You may have bad secondary caps on your PSU! Different error codes, and video corruption, makes it a possible PSU problem!

Topcat over at badcaps. net, likely would tell you to replace the caps or the PSU.

 

[iameatingjam]

Sounds like those fatal WHEA "Cache Hierarchy Error" crashes. There's a tendency for that to happen when the room is warm. Some CPU cores seem to be flaky at more than 55C! It can also happen, due to VRAM being OC'ed too much.

Sounds like the issues are continuing even with the GPU removed though, or thats what I thought I understood from skimming the discussion so far.

You may have bad secondary caps on your PSU! Different error codes, and video corruption, makes it a possible PSU problem!

Topcat over at badcaps. net, likely would tell you to replace the caps or the PSU.

Yes, perhaps.

 

[RJARRRPCGP]

Sounds like the issues are continuing even with the GPU removed though, or thats what I thought I understood from skimming the discussion so far.

The BSOD screenshots, make me suspicious about the PSU. I suspect bad/irregular output voltages!

I also saw video corruption like that, except it was on power up, after I added an HDD to a weak tiny PSU, back in the late '00s, LOL. I was like, seriously!? The CPU didn't even take much power!

It was with a tiny PSU that wasn't even 150W, IIRC, but that was still insane!

New problem... When I load windows up now, it takes ages.

Yikes! That does sound like a known bad cap symptom that Topcat says to look out for! (Hint: It's not just Windows slowness, it's different error codes and video corruption combined with a slow boot)

I suggest that you get HWiNFO! And try to see if you can catch voltages that are off-looking when under load. The BIOS screen stats, are worthless in this case.

Very likely, the +12V rail, +5V rail and the +3.3V rail will have voltage drop that normally isn't there!

 

[DAPUNISHER]

Samsung 2tb NVMe

If it's an older 980 pro without the firmware patch, it may be faulty. All the BSODs followed by long boot times I have seen with failing OS drives. That'd be my next step. And definitely keep the system barebones using iGPU until the troubleshoot is done.

 

[alexburn]

If it's an older 980 pro without the firmware patch, it may be faulty. All the BSODs followed by long boot times I have seen with failing OS drives. That'd be my next step. And definitely keep the system barebones using iGPU until the troubleshoot is done.

Typo... It's a Toshiba NVMe.
I've got a new SSD to try a fresh install of W11 today.
I'm leaving the GPU disconnected for now until I figure what's going on. Next step will be PSU.

 

[DAPUNISHER]

Yeah, you have narrowed the list down quite a bit already. Does your friend mix new and used parts on these builds, besides the GPU?

 

[alexburn]

Yeah, you have narrowed the list down quite a bit already. Does your friend mix new and used parts on these builds, besides the GPU?

Nope they were all brand new and I've still got all the boxes.

 

[Zach_01]

Good lord...
It hurts even to see this, and also hurts that no one noticed it...

The GPU card needs immediate repasting on the GPU

Nothing wrong with any thermal pad as VRAM and VRM temps are well within temp specs.
45C delta (110 - 65) between GPU and Hotspot temp is a red flag warning that is "making noise"

Notice how the GPU clocks and the power consumption decreases once the hotspots hits 110+C
This is a 220~230W card and already dropped to 205W during that test
Typical throttling.



Not quite sure if it will resolve the PC issue but it should be addressed nonetheless

 

[ratirt]

Good lord...
It hurts even to see this, and also hurts that no one noticed it...

The GPU card needs immediate repasting on the GPU

Nothing wrong with any thermal pad as VRAM and VRM temps are well within temp specs.
45C delta (110 - 65) between GPU and Hotspot temp is a red flag warning that is "making noise"

Notice how the GPU clocks and the power consumption decreases once the hotspots hits 110+C
This is a 220~230W card and already dropped to 205W during that test
Typical throttling.

View attachment 380314

Not quite sure if it will resolve the PC issue but it should be addressed nonetheless

The 5700xt's hotspot going to 110 in some cases (bench-marking mostly) is nothing unusual.

 

[JustBenching]

If cpu and ram are stock, the pc restarting is a PSU issue 99% of the time. A faulty gpu in most cases would just crash the game, lose signal and then come back after a few seconds. High gpu temps shouldnt restart your pc either unless the temps affect some power delivery, I had a 3090 thermal throttling to 105c and it worked "fine", no restarts.

 

[Zach_01]

The 5700xt's hotspot going to 110 in some cases (bench-marking mostly) is nothing unusual.

It got to 110C and stay there... while power and clocks starting to drop. Daaa....
My old 5700XT after 2.5 years of intense usage started to closing at 100C and I repaste it with Kryonaut.
Never exceeded 80~85C again for the next 1.5years at 2.0~2.1GHz. Same as new and slightly better.

Cant you see that the GPU is throttling?
Far from normal or intended.
Dont let vendors or even AMD/nVidia or whoever feeding us BS about how normal the 110C hotpot is.
Its the highest operating limit and once reached throttling occurs just like I see it clearly on that screenshot.

Take or leave my 2 cent.

EDIT:
Just notice it...
Also whats up with the 37% of GPU fan speed?

 

[alexburn]

Good lord...
It hurts even to see this, and also hurts that no one noticed it...

The GPU card needs immediate repasting on the GPU

Nothing wrong with any thermal pad as VRAM and VRM temps are well within temp specs.
45C delta (110 - 65) between GPU and Hotspot temp is a red flag warning that is "making noise"

Notice how the GPU clocks and the power consumption decreases once the hotspots hits 110+C
This is a 220~230W card and already dropped to 205W during that test
Typical throttling.

View attachment 380314

Not quite sure if it will resolve the PC issue but it should be addressed nonetheless

Funnily enough, I repasted the GPU last night. I'm yet to test it back in the machine, but will find out today if it's any better. I've been using the iGPU for the last few days.

That being said... I replaced the SSD yesterday and did a fresh install of windows. I also did some OCCT tests and it's been stable all day. I then managed to play COD for a few hours last night without a single crash. So I'm hoping I found the culprit.

It got to 110C and stay there... while power and clocks starting to drop. Daaa....
My old 5700XT after 2.5 years of intense usage started to closing at 100C and I repaste it with Kryonaut.
Never exceeded 80~85C again for the next 1.5years at 2.0~2.1GHz. Same as new and slightly better.

Cant you see that the GPU is throttling?
Far from normal or intended.
Dont let vendors or even AMD/nVidia or whoever feeding us BS about how normal the 110C hotpot is.
Its the highest operating limit and once reached throttling occurs just like I see it clearly on that screenshot.

Take or leave my 2 cent.

EDIT:
Just notice it...
Also whats up with the 37% of GPU fan speed?

The GPU is stock, no OC or mods. Not sure why the fan would only be at 37%. I'll take a look at Gigabyte Control Centre and edit the fan curve. That being said, it's obviously based on the lower GPU core temp, not the hotspot.

I took the cooling block off the GPU today, cleaned up the old paste (which looked fine) and repasted with Arctic MX4. Just done another test and came out with the same results. The hotspot temp was still 110 even with the fan set to 100% with MSI Afterburner.

I also did an OCCT power test with the GPU back in. All voltages were very stable and within spec, but the PC still rebooted without warning after about 3 minutes, no blue screen, so no dump file. The test passed fine with the iGPU and the low spec spare GPU I have. Does this sound like the PSU is tripping out under the full load of the higher spec GPU?

I think my next step has to be trying a new PSU... Will probably upgrade from 850w to 1000+.

Update:
Just had a read that the hotspot on this particular graphics card is within spec at 110 when run on Extreme benchmarks.
AMD article

 

[eidairaman1]

Funnily enough, I repasted the GPU last night. I'm yet to test it back in the machine, but will find out today if it's any better. I've been using the iGPU for the last few days.

That being said... I replaced the SSD yesterday and did a fresh install of windows. I also did some OCCT tests and it's been stable all day. I then managed to play COD for a few hours last night without a single crash. So I'm hoping I found the culprit.


The GPU is stock, no OC or mods. Not sure why the fan would only be at 37%. I'll take a look at Gigabyte Control Centre and edit the fan curve. That being said, it's obviously based on the lower GPU core temp, not the hotspot.

I took the cooling block off the GPU today, cleaned up the old paste (which looked fine) and repasted with Arctic MX4. Just done another test and came out with the same results. The hotspot temp was still 110 even with the fan set to 100% with MSI Afterburner.

I also did an OCCT power test with the GPU back in. All voltages were very stable and within spec, but the PC still rebooted without warning after about 3 minutes, no blue screen, so no dump file. The test passed fine with the iGPU and the low spec spare GPU I have. Does this sound like the PSU is tripping out under the full load of the higher spec GPU?

I think my next step has to be trying a new PSU... Will probably upgrade from 850w to 1000+.

Update:
Just had a read that the hotspot on this particular graphics card is within spec at 110 when run on Extreme benchmarks.
AMD article

Check the max current output

 

[kiddagoat]

What power supply do you have currently?? I don’t see the full system specs listed.

 

[eidairaman1]

What power supply do you have currently?? I don’t see the full system specs listed.

He only Mentions a Seasonic 850W, not a specific model in that brand, but we never know. I know sometimes the best brands get a bad unit or a unit that has a shorter life than average...

 

[kiddagoat]

I just perused the Seasonic website. Most of their 850W have a 70A single rail.

The card at most pulls 225W. So I mean that’s just under 20A. Could be an issue with the PSU.

 

[RJARRRPCGP]

I've just got home to run a test. Ran Benchmark, no issues. Loaded up COD and it crashed before the game even started. BSOD states KERNEL MODE HEAP CORRUPTION. Restarted machine, and it crashed on boot up... BSOD states PAGE FAULT IN NONPAGED AREA.
Any ideas?!

Is your room very warm?! Like 80F+? Or next to a radiator?!

 

[alexburn]

Some major developments in the last few days... I'm pleased to report that I've figured out the problems and it wasn't a hardware related fault (fully)...

So it appears I had 2 very specific issues.

1. My setup, including after I had restored Windows, was the latest 24H2 version of W11. It turns out that there is a known bug/incompatibility with certain chips when rendering 3D. This resulted in DirectX problems causing a crash when I was running my spare GTX970 graphics card. The crash would cause corruption and the problems snowballed from there. The solution was to find a version of 23H2 and roll back using a W11 installation media.

2. One of my original issues was that my system would just restart without warning, no error messages, no dump files. This does indeed point to a power issue. Even though my rig was pulling less than the spec, it seems the PSU didn't like something. Again, each loss of power caused a risk of corruption. So I contacted Seasonic for an RMA replacement. They were very helpful and gave me a list of things to try, saying they would replace free of charge with the 10 year warranty if I tried everything on the list...
One of the suggestions was to do with the 8 pin PCIEx power cable. My GPU has dual PCIEx power inputs (so 14 pins in total). The supplied cable has a dual PCIEx output that daisy chains from one input to the next. The advice was to run a completely separate PCIEx cable from another output to split the load from 2 outputs. There must have been too much power draw for a single output.

I can now run everything at Max with no errors or reboots. Strange, as I used to be able to do that anyway, even with the power cable as it was. So something must have either updated or degraded for this to suddenly start failing. But I'm just thankful it's all now resolved!

Thank you to everyone for your persistent help and advice. Hopefully anyone else who comes across similar issues can find some assistance from this thread.

Is your room very warm?! Like 80F+? Or next to a radiator?

In answer to this... Nope, a very cold room... Probably around 60 degrees (my heating is in the process of being upgraded in the house renovation).

He only Mentions a Seasonic 850W, not a specific model in that brand, but we never know. I know sometimes the best brands get a bad unit or a unit that has a shorter life than average...

Is the Seasonic Platinum Plus Focus PX850.
I also just purchased the Corsair Gold Plus RM1000x Shift PSU, but didn't need it, so I'm looking at returning it now.

 

[DAPUNISHER]

Thanks for posting the solution for searchers.

One of the suggestions was to do with the 8 pin PCIEx power cable. My GPU has dual PCIEx power inputs (so 14 pins in total). The supplied cable has a dual PCIEx output that daisy chains from one input to the next. The advice was to run a completely separate PCIEx cable from another output to split the load from 2 outputs. There must have been too much power draw for a single output.

Probably due to transient power spikes.

IIRC you stated a friend built the system? If so, tell him it's best practices to never daisy chain the connectors when possible. DIYers have become over obsessed with cable management. I don't know that's the reason he did it, but it tracks.

 

[alexburn]

Thanks for posting the solution for searchers.


Probably due to transient power spikes.

IIRC you stated a friend built the system? If so, tell him it's best practices to never daisy chain the connectors when possible. DIYers have become over obsessed with cable management. I don't know that's the reason he did it, but it tracks.

Indeed it does. It looked much neater using just one cable, so can understand why it was done. But I'm not fussed about the extra cable clogging up the aesthetics!

 

[DAPUNISHER]

Indeed it does. It looked much neater using just one cable, so can understand why it was done. But I'm not fussed about the extra cable clogging up the aesthetics!

Function over form, always. I hope this experience has convinced you to build your own next time? You did a great job of troubleshooting. You will have no difficulty following a build tutorial on Youtube.

 

[alexburn]

Function over form, always. I hope this experience has convinced you to build your own next time? You did a great job of troubleshooting. You will have no difficulty following a build tutorial on Youtube.

Thanks! I used to build my own machines many years ago. This time around, I just didn't have the time or inclination to update myself on the latest/best components. I just wanted to outsource the hassle for a decent gaming machine. And to be fair, it's a great setup!

 

[eidairaman1]

Some major developments in the last few days... I'm pleased to report that I've figured out the problems and it wasn't a hardware related fault (fully)...

So it appears I had 2 very specific issues.

1. My setup, including after I had restored Windows, was the latest 24H2 version of W11. It turns out that there is a known bug/incompatibility with certain chips when rendering 3D. This resulted in DirectX problems causing a crash when I was running my spare GTX970 graphics card. The crash would cause corruption and the problems snowballed from there. The solution was to find a version of 23H2 and roll back using a W11 installation media.

2. One of my original issues was that my system would just restart without warning, no error messages, no dump files. This does indeed point to a power issue. Even though my rig was pulling less than the spec, it seems the PSU didn't like something. Again, each loss of power caused a risk of corruption. So I contacted Seasonic for an RMA replacement. They were very helpful and gave me a list of things to try, saying they would replace free of charge with the 10 year warranty if I tried everything on the list...
One of the suggestions was to do with the 8 pin PCIEx power cable. My GPU has dual PCIEx power inputs (so 14 pins in total). The supplied cable has a dual PCIEx output that daisy chains from one input to the next. The advice was to run a completely separate PCIEx cable from another output to split the load from 2 outputs. There must have been too much power draw for a single output.

I can now run everything at Max with no errors or reboots. Strange, as I used to be able to do that anyway, even with the power cable as it was. So something must have either updated or degraded for this to suddenly start failing. But I'm just thankful it's all now resolved!

Thank you to everyone for your persistent help and advice. Hopefully anyone else who comes across similar issues can find some assistance from this thread.


In answer to this... Nope, a very cold room... Probably around 60 degrees (my heating is in the process of being upgraded in the house renovation).


Is the Seasonic Platinum Plus Focus PX850.
I also just purchased the Corsair Gold Plus RM1000x Shift PSU, but didn't need it, so I'm looking at returning it now.

Thanks for posting the solution for searchers.


Probably due to transient power spikes.

IIRC you stated a friend built the system? If so, tell him it's best practices to never daisy chain the connectors when possible. DIYers have become over obsessed with cable management. I don't know that's the reason he did it, but it tracks.

Function over form, always. I hope this experience has convinced you to build your own next time? You did a great job of troubleshooting. You will have no difficulty following a build tutorial on Youtube.

Thanks! I used to build my own machines many years ago. This time around, I just didn't have the time or inclination to update myself on the latest/best components. I just wanted to outsource the hassle for a decent gaming machine. And to be fair, it's a great setup!

Im glad the issue was resolved, it was unstable current/voltage draw, anyways when wires are split like that 1 value gets a drop in voltage or current.

I do recall even 10 years ago it was a bad idea to use a pig tail on the end of a pcie 6 or 8 connector.

 

[RJARRRPCGP]

Your PSU issues, normally would have me suspecting bad secondary filter caps.

But, it's possible for a primary filter cap to be bad.