Years of Hard Crashes with BIOs Resets

[freebie]

Hi,

 

I've been trying, unsuccessfully, to diagnose this for quite some time. Seeing if anyone has any ideas or ways to diagnose I've not thought of.

 

I'll start with the symptoms. When playing games, my PC will sometimes hard shutdown (not a BSoD, just power off). On restarting, the bios is wiped and I need to load my old profile.

 

Here are some notes of stuff I've been through and some points of interest I've picked up on:

 

1. It's not all games. Though it seems to be consistently when I using the PC for games. I recently started playing Deaths Door (great game fyi) and this is probably the most consistent game I've ever had produce this problem, and will crash the PC within 5 minutes of playtime. That's what inspired this post as now I have a somewhat consistent way to reproduce. Before other games would only crash once like every 10 or so sessions - so very irregular.
2. The motherboard on this machine has been replaced as the same board died a few months back and I replaced it with the same board (I was kind of hoping this was going to resolve the issue and was dodgy board and it had finally died).
3. I've tried running with default bios settings just to make sure it's not any of the virtualisation or XMP settings I have enabled being unstable.
4. I've got the latest, and continued to get the latest chipset drivers for the motherboard as they've been released. None have solved my issue yet. Maybe some have made it less frequent, but that's had to tell and my just be a skewed perception as it's irregular anyway.
5. I've got a Voltmeter out and tested the voltages on all the pins are what they should be. They are.
6. I've also sent the PSU back to the manufacture, and they tested it, said it was fine, and sent it back. No idea what their tests were though, and if they test under load.
7. I've looked through the Windows Event Viewer (though I'm by no means that familiar with it, so happy to hear more places I should check). There's nothing there that looks useful, other then a Kernel-Power (Event 41) logged after the PC has been rebooted.
8. I've ram a Memory Check, got the all clear.
9. I ran `sfc /scannow` to fix any windows corrupted files. It found stuff, but didn't solve this problem.
10. Finally, also my graphics card drivers are regularly kept up to date.

So my current thoughts are dialling in on the processor itself. Maybe some function on the processor that is commonly used in gaming APIs (if that's a thing)? That would line up with such a hard crash with a BIOs reset in my head. And I can be fairly certain it's not the motherboard, as it's been replaced now. I also know it's not a load related thing because it happens on quite undemanding games like Deaths Door, and has no problems with games like Control.

 

Here's my build:

 

OS: Windows 11 Pro 64bit

MOB: X570 I AORUS Pro WiFi (BIOS F35)

CPU: AMD Ryzen 5 3600

GPU: RTX 2070 Super

PSU: NJ450-SXL

MEM: 16GB (cant recall exactly the make/model but can get it if we care). 

 

Any pointers of what to try next... Also happy to upload any logs that may be of help.

 

Thanks

[Ashleyyyy]

have you stress tested the system> i mean running cinebench and heaven benchmark simultaniously to see if it turns off? 

 

it could be as simple as a psu that's too weak for the hardware. slamming the machine with a load should show that. 

 

if it does turn off, run cinebench or heaven on their own, to see if it still turns off when only the gpu or only the cpu is under load. if that's not the case, it's likely your psu is too weak to run both the gpu and cpu under load like in a game. 

[Kilrah]

The PSU is likely fine, but 450W is a little tight and it probably can't cope with the occasional spike.

[Godlygamer23]

Sounds like a power delivery issue with the power supply if it's just totally powering off like that. The load exceeds what the PSU is capable of handling for a brief moment, causing the OCP to trigger. Even if files were corrupt on your installation of Windows, that wouldn't cause a shutdown like that - it would more likely simply cause a BSOD.

[freebie]

16 minutes ago, Ashley MLP Fangirl said:

have you stress tested the system> i mean running cinebench and heaven benchmark simultaniously to see if it turns off? 

 

it could be as simple as a psu that's too weak for the hardware. slamming the machine with a load should show that. 

 

if it does turn off, run cinebench or heaven on their own, to see if it still turns off when only the gpu or only the cpu is under load. if that's not the case, it's likely your psu is too weak to run both the gpu and cpu under load like in a game. 

Hi Ashley,

 

I haven't ran those since I set up the machine a good few years ago. I just ran the two together now for 10 minutes, which pegged the cpu at 100% and the gpu at around 90-95%. 

 

That would line up with what I posted above, with this happing in games like Deaths Door that has like no CPU or GPU demand vs other games that run with no issues.

 

Let me know what you think.

[Godlygamer23]

5 minutes ago, freebie said:

Hi Ashley,

 

I haven't ran those since I set up the machine a good few years ago. I just ran the two together now for 10 minutes, which pegged the cpu at 100% and the gpu at around 90-95%. 

 

That would line up with what I posted above, with this happing in games like Deaths Door that has like no CPU or GPU demand vs other games that run with no issues.

 

Let me know what you think.

That test misses specific ways that the game itself could be stressing the hardware. The test is fairly consistent because they're benchmarks, whereas a game can have random spikes in loads that the power supply has to deal with...Transient spikes, which it has to deal with. This may cause, even just for a real quick moment, the power supply to exceed its capacity and trigger OCP. 

 

Even if the component is at "100%", that doesn't mean all the hardware is actually being stressed within those respective components, which is why different loads at 100% can cause varying temperatures.

 

 

[freebie]

14 minutes ago, Kilrah said:

The PSU is likely fine, but 450W is a little tight and it probably can't cope with the occasional spike.

 

6 minutes ago, Godlygamer23 said:

Sounds like a power delivery issue with the power supply if it's just totally powering off like that. The load exceeds what the PSU is capable of handling for a brief moment, causing the OCP to trigger. Even if files were corrupt on your installation of Windows, that wouldn't cause a shutdown like that - it would more likely simply cause a BSOD.

When I use a online power draw calculator and a plug socket watt monitor, I think I'm comfortably under the 450W rating of the power supply. i.e. in like the high 200W to low 300W range.

 

Unless you're saying that even though these are the ratings, that there can be spikes that dramatically (from my perspective) jump outside of these ranges?

 

I've also run the benchmarkers/stress testers above to try and peak my power draw.

 

I do admit the 450W power supply is probably lower then most people would go for, but I kind of think people over spec there anyway, and I don't have a bunch of external drives etc.

 

Let me know what you think.

[freebie]

24 minutes ago, Godlygamer23 said:

Sounds like a power delivery issue with the power supply if it's just totally powering off like that. The load exceeds what the PSU is capable of handling for a brief moment, causing the OCP to trigger. Even if files were corrupt on your installation of Windows, that wouldn't cause a shutdown like that - it would more likely simply cause a BSOD.

So when I run Deaths Door at the moment, as I'm using it as it's giving me consistent crashes. The CPU and GPU are maxing at 40% and 70% respectively. I get what you're saying about all loads are not equal, but they cannot be that unequal to take that load for a machine spec'ed at around 300W max reequipment to then exceed 450W.

 

Though it may be on to something though, and it's more of a firmware bug or a fab issue of the CPU/GPU that will draw way more then it's expected load to perform a certain operation?

 

Let me know if there's any other info I can provide that could rule out/in the PSU.

[Godlygamer23]

3 minutes ago, freebie said:

So when I run Deaths Door at the moment, as I'm using it as it's giving me consistent crashes. The CPU and GPU are maxing at 40% and 70% respectively. I get what you're saying about all loads are not equal, but they cannot be that unequal to take that load for a machine spec'ed at around 300W max reequipment to then exceed 450W.

 

Though it may be on to something though, and it's more of a firmware bug or a fab issue of the CPU/GPU that will draw way more then it's expected load to perform a certain operation?

 

Let me know if there's any other info I can provide that could rule out/in the PSU.

It's more about the transition that's causing the issue, again depending on how Deaths Door is loading the hardware, even at 40% and 70% usage respectively on each. That's the issue - the transient load spikes past what the unit is capable of, even for a single millisecond, and the internal hardware of the PSU sees it and goes "NOPE" and shuts off the unit. 

 

Do you happen to have another, potentially more capable, power supply that you can temporarily swap in to verify it? Do you also have a Kill-A-Watt that you can plug in to see how much power the system is actually pulling? 

[freebie]

37 minutes ago, Godlygamer23 said:

It's more about the transition that's causing the issue, again depending on how Deaths Door is loading the hardware, even at 40% and 70% usage respectively on each. That's the issue - the transient load spikes past what the unit is capable of, even for a single millisecond, and the internal hardware of the PSU sees it and goes "NOPE" and shuts off the unit. 

 

Do you happen to have another, potentially more capable, power supply that you can temporarily swap in to verify it? Do you also have a Kill-A-Watt that you can plug in to see how much power the system is actually pulling? 

I do not have a spare juiced up PSU around, unfortunately. But I do have a Kill-A-Watt.

 

I am unable to easily plug this in just to the PSU connection due to my cable management being under my desk and hard to get to, so I've taken these readings from the wall that feeds the extension cord that drives the PSU, but also my monitors (a 4K and a 2K), and a bunch of other little things, so the numbers will be larger due to that.

 

Here are some numbers:

 

* Windows desktop w/ monitors on: 206W

 

* Windows desktop w/ monitors off: 114W

 

* Monitor draw: ~92W

 

* Max Watt Draw seen during PC boot cycle to desktop, and then loading game save: 418.5W

 

* Max Watt Draw seen after PC shuts down from whatever I'm seeing playing game: 418.5W

 

I don't know how sensitive these Kill-A-Watt things can be to these instantaneous power draws, but potentially it never had chance to pick it up? I defer to you.

 

If you're feeling is it still could be a power spike, what rated PSU would you recommend I try, to be safely outside of whatever these spikes you feel/know can be?

 

(Also happy to get any other specific readings you want from that Kill-A-Watt)

 

 

 

[Godlygamer23]

1 hour ago, freebie said:

I do not have a spare juiced up PSU around, unfortunately. But I do have a Kill-A-Watt.

 

I am unable to easily plug this in just to the PSU connection due to my cable management being under my desk and hard to get to, so I've taken these readings from the wall that feeds the extension cord that drives the PSU, but also my monitors (a 4K and a 2K), and a bunch of other little things, so the numbers will be larger due to that.

 

Here are some numbers:

 

* Windows desktop w/ monitors on: 206W

 

* Windows desktop w/ monitors off: 114W

 

* Monitor draw: ~92W

 

* Max Watt Draw seen during PC boot cycle to desktop, and then loading game save: 418.5W

 

* Max Watt Draw seen after PC shuts down from whatever I'm seeing playing game: 418.5W

 

I don't know how sensitive these Kill-A-Watt things can be to these instantaneous power draws, but potentially it never had chance to pick it up? I defer to you.

 

If you're feeling is it still could be a power spike, what rated PSU would you recommend I try, to be safely outside of whatever these spikes you feel/know can be?

 

(Also happy to get any other specific readings you want from that Kill-A-Watt)

It's been a while since I've had a Kill-A-Watt. How fast are the readings? It'll likely be too fast for it to see it - it'll basically occur in between the reading intervals the vast majority of the time.

 

Is an SFX power supply an actual requirement for you? 

[freebie]

4 minutes ago, Godlygamer23 said:

It's been a while since I've had a Kill-A-Watt. How fast are the readings? It'll likely be too fast for it to see it - it'll basically occur in between the reading intervals the vast majority of the time.

 

Is an SFX power supply an actual requirement for you? 

I'm not sure the intervals. it's only a cheap one as well, so if there is variation there, then this one will be 'slow'.

 

As for the size, yeah, my build is in a Dan A4-SFX Case.

 

Cheers for the help by the way.

[freebie]

The other thing I can do, but I’ve not yet as I had originally written off the PSU, is order a monster PSU and return it if it doesn’t solve the issue. 
 

Should I do that? What rating would you suggest? 500W, 800W, 1000W?

[Godlygamer23]

26 minutes ago, freebie said:

The other thing I can do, but I’ve not yet as I had originally written off the PSU, is order a monster PSU and return it if it doesn’t solve the issue. 
 

Should I do that? What rating would you suggest? 500W, 800W, 1000W?

SFX power supplies I think are a bit harder to recommend something on, and as a result of the size, you may not get as robust circuitry since there isn't as much space(mostly dealing with capacitors). 650W should be fine. Probably anything modern from NZXT, Corsair, or Seasonic that's ATX.

Edited May 14, 2023 by Godlygamer23
Specified standard.

[freebie]

9 hours ago, Godlygamer23 said:

SFX power supplies I think are a bit harder to recommend something on, and as a result of the size, you may not get as robust circuitry since there isn't as much space(mostly dealing with capacitors). 650W should be fine. Probably anything modern from NZXT, Corsair, or Seasonic that's ATX.

Sorry, I see you added ATX. Are you saying to try a full size PSU external to the case instead of getting a small form factor one which a higher rating? Because the SFX ones are not robust enough?

 

EDIT: I've ended up ordering a Seasonic Focus SGX 650W, not a full sized ATX. My thought being 'if this does resolve it, I still need a PSU that'll fit in my case that can replace my current one. So might as well test with one I may be able to keep'.

[Godlygamer23]

55 minutes ago, freebie said:

Sorry, I see you added ATX. Are you saying to try a full size PSU external to the case instead of getting a small form factor one which a higher rating? Because the SFX ones are not robust enough?

 

EDIT: I've ended up ordering a Seasonic Focus SGX 650W, not a full sized ATX. My thought being 'if this does resolve it, I still need a PSU that'll fit in my case that can replace my current one. So might as well test with one I may be able to keep'.

The concern with SFX power supplies is they simply have limited space, so they'll either be missing capacitors or use smaller ones, compared to ATX units rated for the same output. It's just the nature of it. That means, coming back to transient spikes, they may not be able to handle them as well. I hope the new unit resolves the issue though. I suppose with the higher output, it should do better at handling transient spikes. 

[freebie]

@Godlygamer23 you were totally right; I think it was a transient spikes.

 

The new PSU does the job, at least in the Death Door game I was using to recreate the issue. I hope that 650W is enough of a headroom to handle transient spikes of other, more demanding games, as well...

 

I looked into this transient spike stuff because I hadn't heard of it until now, and seems to be a bit of a modern and growing problem, with these demands getting larger and larger from modern (I think mainly NVIDIA) GPUs. It's a bit obnoxious to be honest. Hard to know how to spec a PSU as I don't think people normally publish these peak demand numbers, or the online PSU calculators take them into account either.

 

Anyway, thanks again for the help. Hopefully this shows up for someone else with the same issue in a google search and helps them. It's a hard thing to debug if you've never heard of these spikes as can look like it could be mobo, CPU, PSU, or maybe drivers.

[Godlygamer23]

2 hours ago, freebie said:

@Godlygamer23 you were totally right; I think it was a transient spikes.

 

The new PSU does the job, at least in the Death Door game I was using to recreate the issue. I hope that 650W is enough of a headroom to handle transient spikes of other, more demanding games, as well...

 

I looked into this transient spike stuff because I hadn't heard of it until now, and seems to be a bit of a modern and growing problem, with these demands getting larger and larger from modern (I think mainly NVIDIA) GPUs. It's a bit obnoxious to be honest. Hard to know how to spec a PSU as I don't think people normally publish these peak demand numbers, or the online PSU calculators take them into account either.

 

Anyway, thanks again for the help. Hopefully this shows up for someone else with the same issue in a google search and helps them. It's a hard thing to debug if you've never heard of these spikes as can look like it could be mobo, CPU, PSU, or maybe drivers.

I'm glad that fixed your issue. I would definitely advise you to continue testing and trying other demanding games - the main issue I think is when the card isnt necessarily in a 'steady state'. That's when the caps are consistently being drawn from because the GPU basically needs power and needs it now, so less equipped power supplies have either missing output caps or has lesser capacity caps, so they don't have the ability to support the sudden energy demand. It's not so much about the total energy draw, but the sudden demands for power that cause the issues. It's very unfortunate, but also the current reality. 

 

Reviews from TechPowerUp do take into account transient loads, so I'd definitely keep them in mind for power supply reviews, among other reviewers.