Question random system restarts?

[darknopror]

my system has been suffering from random restarts that seem to not have a reason/ i cant recreate them at will.
so first of all here is my rig
cpu: ryzen 5 7500F (undervolted using PBO all cores negative 20)
ram: lexar thor ddr5 6000 cl32 (expo profile)
gpu: xfx rx 6950XT ( have multiple profiles recent crash happened on this one
min freq: 2200 MHz
max freq: 2300MHz
vram: stock
fans: 80%
power limit: stock)
mobo: asrock b650 pro rs
psu: 750 W gigabyte UD
i also must say that there are several kind of restarts/reboots that my system would go through and they are as follow.
1- the screen goes solid black/green with a digital buzzing noise for a second or two before rebooting, i have concluded that this is caused by the vram either overheating or being overclocked since i could somewhat reliably recreat this issue at will.
2- the system just reboots without a warning and the windows reliability manager simply states somethign along the lines of "the system shutdown was unexpected", cant recreate this issue but here are some more info, also it doesnt matter if its underload or if it isnt, it could crash after a stress test or while browsing steam
i got my system to randomly reboot/ crash before when i heavily undervolted the CPU, since then i undervolted it correctly and did a lot of stress tests (OCCT, prime 95 for 13 hours).
i have overclocked the ram timings but since then i have set them back to thier normal expo profile and stress tested them using OCCT and prime 95 and memtest and they all came back fine.
when i thought that i finally understood what the problem was and simply undervolted and underclocked my GPU and played a demanding game for a long time (about 2 hours) as soon as i exited the game the system reboot again without warning but also without the buzzing noise.
also i tried to recreate the vram problem again and decided to overclock it and overclock my entire gpu for that matter and it worked fine, it even let me set a record in 3dmark speedway stress test, so i am completely lost.
Ps: i suspected its the case restart/reset connected or button so i unplugged.
if any further information is required please let me know because this issue is really giving me a hard time to troubleshoot.
Thanks in advance.

 

[darknopror]

forgot to mention i never get BSODs at all, its almost always just a reboot.

 

[faalin]

The minimum PSU recommended for that GPU from XFX is 850w, and unless they have been able to fix the transient power spikes with driver updates i would recommend bumping that up to a very good 1000w psu.

While stress testing may seem ok with a high constant load, its the quick load spikes that is causing the issue.

 

[darknopror]

The minimum PSU recommended for that GPU from XFX is 850w, and unless they have been able to fix the transient power spikes with driver updates i would recommend bumping that up to a very good 1000w psu.

While stress testing may seem ok with a high constant load, its the quick load spikes that is causing the issue.

I have taken a look at the PSU and while you might be right at first, the fact that a lot of the times the PC would reboot while doing absolutely nothing causes concern and debunks the "PSU is a problem" claim.
My wattage never exceeds 80W for the CPU and 310W for the GPU.
It's highly unlikely that every single thing with the system power spikes all at once to exceed

 

[faalin]

heres a good video to watch,

View: https://www.youtube.com/watch?v=wnRyyCsuHFQ


Around the 9:40 mark he starts to explain transient spikes, the power draw for your card is rated at 335w, now double that for the transient spike at 670w plus the 80w for cpu and you're at 720w plus anything else in the system.

All it takes is running a game, closing the game witch will drop the card of any load because hey, there's nothing to render. Then boom i need to load the desktop, and pow transient spike. It only takes a split second to happen and most software doesn't refresh fast enough to catch it.



Trust me i know, why do you think i have a 1600w PSU in my system. My original EVGA 750w G2 succumbed to transient spikes from my 2080 Ti FTW3. Instead of buying a 1000w psu i installed a EVGA 1600w G2 i had laying around from another build.


If you want to nitpicking

GPU 670w with transient spike
CPU 80w (stated by you)
NVMe drive 10w when reading and writing
Ram 3w for every 8GB assuming your running 16GB 6w
fans? will call it 4, combo of case and cpu cooler, 3w per fan 12w

Thats 788w in a perfect case scenario, you also have to take into account the age of the PSU and degradation it has.

If your running an AIO your looking at 25-35w


Once again nitpicking here, tomshardware did a review on your PSU https://www.tomshardware.com/reviews/gigabyte-ud750gm-power-supply-review

and one of the cons was "Transient response could be better"

 

[Satan-IR]

I would say it's the PSU too. Not a good unit in coping with transient spikes either. That means sooner wear and tear caused by those spikes.

The fact that there's never been a BSOD makes it less likely to be a software or driver issue. The spikes happen so fast the software can't detect it. Talking about few miliseconds.

Easiest test can be run you system, as is, with a 850W known good working PSU if available.

 

[darknopror]

Unfortunately I don't have another PSU around neither can I get a new one at any time.
Also wouldn't a psu failure to respond to transient load lead to a system shutdown rather than a restart? Sorry for asking so many questions but I'm trying to learn and I don't have that much experience with PCs yet.
But I have undervolted and underclocked the GPU to the point where it only draws a max of 230 W and the CPU consumes 80 W at peak use which is very unlikely to be a game.
However I will be testing further, thanks for the replies
Edit: forgot to mention the PSU is only a few months old.

 

[darknopror]

Another thing to note is that the difference between GPU temp and GPU hotspot is usually around 15~20 degrees and at worst cases around 30.
GPU temp 67
Hotspot 97.

 

[Satan-IR]

Restarting after shut down due to power loss might be due to power settings in the BIOS. If set to POWER ON after AC power loss that might explain it. If you want PC to stay OFF you can look for a setting called something like 'AC power loss restart' or similar and disable it.

Spikes are sudden and for a fraction of a second the load goes almost twice the constant load you might expect to see while gaming or stress testing (the load PSU is rated for). Usually it's PSUs own OCP (Over Current Protection) triggering a shut down due to these spikes.

Edit: forgot to mention the PSU is only a few months old.

This eliminates the ageing factor somehow if you bought it brand new. From what you've described I agree that most probably it can't cope with the spikes. If you've noticed it happens under certain games more frequently one remedy for now would be not playing those. Until you can get a 1KW+ unit.

 

[DaleH]

I have taken a look at the PSU and while you might be right at first, the fact that a lot of the times the PC would reboot while doing absolutely nothing causes concern and debunks the "PSU is a problem" claim.
My wattage never exceeds 80W for the CPU and 310W for the GPU.
It's highly unlikely that every single thing with the system power spikes all at once to exceed

You've made a false assumption that the PSU is not the problem. The only conclusive way to prove it is to install a known good PSU. Rebooting while doing nothing, does not prove the power supply is not the cause of the problem.

 

[darknopror]

I have noticed something very strange now, I stated that my GPU silicon isn't very good when it comes to VRAM tuning, almost always causing a crash.
But recently I decided to overclock the VRAM over what I would usually do and it didn't crash, so I returned the settings to stock and moved on to my normal underpowered settings and about 10 mins later, the PC rebooted again.
I think tuning the VRAM causes a reboot, doesn't matter if I actually use those tuned settings or not. If I change the VRAM tuning and the system doesn't crash/reboot immediately it will inevitably regardless if I've been using those overclocked VRAM settings or not.

 

[darknopror]

Restarting after shut down due to power loss might be due to power settings in the BIOS. If set to POWER ON after AC power loss that might explain it. If you want PC to stay OFF you can look for a setting called something like 'AC power loss restart' or similar and disable it.

Spikes are sudden and for a fraction of a second the load goes almost twice the constant load you might expect to see while gaming or stress testing (the load PSU is rated for). Usually it's PSUs own OCP (Over Current Protection) triggering a shut down due to these spikes.


This eliminates the ageing factor somehow if you bought it brand new. From what you've described I agree that most probably it can't cope with the spikes. If you've noticed it happens under certain games more frequently one remedy for now would be not playing those. Until you can get a 1KW+ unit.

I'm using the ASRock b650 pro rs, can't seem to find this setting.
Also, fine I'll bite the bullet and get a bigger PSU but before I do ( mainly cause it's very expensive )
Is there a way to make sure it's not the GPU, CPU or board? I really don't wanna invest in a new PSU and still have the problem.
Also recently my PC rebooted right after exiting RDR2 and almost as soon as wallpaper engine started.
and this time i hadn't overclocked the VRAM at all ( at least in this session)

 

[Satan-IR]

In order to test those you'd need temporary replacements for each to be able to isolate the probable cause, eliminating the possible culprits one by one. Also the PC being able to endure stress tests (although depending on kind of tests run) shows motherboard, CPU and RAM are most probably OK.

I agree with above that the PSU that can't cope with the transient spikes. This is my perception of the issue from what info you provided and without being able to test PC hands-on. Can we say that with absolute cetainty, no not really. Trying the system as is with a good quality 850W or higher capacity unit can make it more certain.

Also trying PSU you already have with a less demanding GPU can also help determine if it's the PSU. But if by any chance GPU has problems lack of reboots with less demanding card can be indicative of a defective GPU too. If you can borrow one and test perhaps.

But you said it happens when system is almost idle and not under heavy load. You've never had a BSOD. These can also point to a PSU not able to cope.

Again, can we say with 100% certainty it's the PSU without having access to the system and PSU, no we can not.

 

[darknopror]

In order to test those you'd need temporary replacements for each to be able to isolate the probable cause, eliminating the possible culprits one by one. Also the PC being able to endure stress tests (although depending on kind of tests run) shows motherboard, CPU and RAM are most probably OK.

I agree with above that the PSU that can't cope with the transient spikes. This is my perception of the issue from what info you provided and without being able to test PC hands-on. Can we say that with absolute cetainty, no not really. Trying the system as is with a good quality 850W or higher capacity unit can make it more certain.

Also trying PSU you already have with a less demanding GPU can also help determine if it's the PSU. But if by any chance GPU has problems lack of reboots with less demanding card can be indicative of a defective GPU too. If you can borrow one and test perhaps.

But you said it happens when system is almost idle and not under heavy load. You've never had a BSOD. These can also point to a PSU not able to cope.

Again, can we say with 100% certainty it's the PSU without having access to the system and PSU, no we can not.

Thank you for replying, if there are any tests that I can do on your own behalf that will help identify or troubleshoot the problem better, I'll be more than happy to do it.

 

[kerberos_20]

Another thing to note is that the difference between GPU temp and GPU hotspot is usually around 15~20 degrees and at worst cases around 30.
GPU temp 67
Hotspot 97.

not really an issue, hotspot is on GPU die, gpu temp is on GPU PCB
you could probably repaste with high viscosity paste so it wont slide on that flat surface, but gpu would just throttle a bit, it wont reboot

 

[darknopror]

not really an issue, hotspot is on GPU die, gpu temp is on GPU PCB
you could probably repaste with high viscosity paste so it wont slide on that flat surface, but gpu would just throttle a bit, it wont reboot

Appreciate the reply.
I also tried heavily undervolting the GPU and stress testing to see what would happen and it didn't reboot, it crashed and froze for a bit then returned to default and continued running.
I guess the most consistent crash occurs when shutting down red dead and going to desktop, as soon as wallpaper engine starts running the PC reboots.
And again, if anyone knows any tests or methods to better isolate the problem and troubleshoot it, I'd be thankful.

 

[Satan-IR]

Thank you for replying, if there are any tests that I can do on your own behalf that will help identify or troubleshoot the problem better, I'll be more than happy to do it.

One was for testing the PSU is using the kind of testing equipment found in electronic laboratories, the kind that people who review PSUs use, which are not avilable to home users like us. So this one is out of question.

Only sure way is to test the system (and run the games you run and try to replicate the situations) with a good quality at least 850W PSU which is known to be able to handle spikes from your graphics card.

As I said before you can also try and check other components. As I said to make sure you each one of them work properly you have to apply a process of elimination.

Again, since the system can go through stress tests without problems (13 hours of Prime95?) the MOBO+CPU+RAM combo is working OK. There's no overheating either, otherwise you'd notice or it would show during stress tests.

 

[darknopror]

I will try to get a 850W PSU asap, but it will be a while before then.
Till then I'll try to eliminate the culprits one by one

 

[Satan-IR]

After all, it's your decision what to do. It's OK to think of other avenues of troubleshooting and to seek further info on the issue.

As @faalin said above, "The minimum PSU recommended for that GPU from XFX is 850w", in that case keep in mind that even a 850W might not be able to cope. Good quality more recent 850W ones might be able to manage.

You can search online or ask around maybe and see what PSUs other people are using with that card that's working. Persoanlly I'd go with a good 1KW if the card manuracturer says 850W minimum. Other calculations faalin made regarding power draw of other components makes sense to me too.

The OEM for this Gigabyte PSU is MEIC, which was responsible for the explosive P-GM series. I'm not saying these UDs are as bad or accusing Gigabyte or MEIC of anything. I, persoanlly, wouldn't buy a PSU from an OEM that made those explosive doostops a few years ago.

Aris's review here says it's not a bad PSU and indeed it passed all the tests. But the list of Cons is almost as long as the Pros and many can point to mediocre design or build quality. Usually not the case with best quality PSUs. Most of them have 2-3 Cons in such reviews and one is like 'short cables' or 'they could have used better fan'.

Anyway even if by a small chance the PSU is not the cause of the reboots, which with the info I have I think it is, if you get a unit with 7-10 years warranty (all good units are at least 7 years) you'd end up with a good quality unit that would outlast your current rig.

Not being able to cope with spikes like this puts a lot of stress on PSU internals and would probably cause it to die sooner than it would under less stress. It's a few months old so you should be able to sell it for a good portion of it's price when new. Some user that needs 600-650W with a card with no big spikes.

Anyway it's your call. You might want to see if others have any other input or suggestions which might help.

 

[darknopror]

I did a very small test to check things and now I'm baffled even more.
I tried my custom preset where I just have the GPU underclocked and undervolted, gamed on it for about 2 or 3 hours then turned off the game and, again, about 5 secs after wallpaper engine kicks in the system goes black screen with a buzzing sound then it reboots.
I tried to recreate it again but this time I ran with default settings ( stock gpu settings )
Opened the same game and waited for the GPU to get hot then turned off game and it didn't crash this time.
I will try to recreate the problem again using different settings, but I really disliked this black screen buzzing audio situation, since I thought it's almost exclusively a VRAM issue when that happens

 

[Satan-IR]

Have you tried to 'recreate the situation' without this Wallpaper Engine? No BOSDs usually means there's hardware issue and not software or drivers at play. Some hardware issue that happens beneath Windows and OS doesn't get a chance to detect and report it.

You mentioned this a few times that "after wallpaper engine kicks in" system reboots. Have you tried to see if uninstalling that application or preventing it from running (as soon as games cloes) can help by any chance? Is it running in background while gaming? Is it utilizing the graphics card to make these wallpapers or whatever it makes?

You've been under the assumption that this is a 'VRAM issue'. Graphics card or its VRAM is defective? It's not impossible and it's a hardware issue. Which in turn is in consistence with the nature of the problem, reboots without any BSOD or indication. Right?

One way of checking that is to get another known good working graphics card of same model like yours, plug it in and see it reboots still happen or not. If not your card's VRAM (or other components on card) may be defective. If reboots happen again it's more likely to be a PSU issue.

Again, usually you can not run tress tests for hours on a PC with defective components. Those stress test as far as I know don't cause those transient spikes that are caused in cards by some games.