(another) kernel power error 41 troubleshooting

[QuantumCoral99]

Hi all!

I'm sure you've seen this exact post before, but in an attempt to troubleshoot efficiently, I am going to post here with my specifics.

issue: in gaming, computer will randomly black screen, and then almost instantly force restart. This has been happing on and off (every few days) since i made a pc upgrade.
specs (after upgrade)

B550 Aurous Elite (rev 1.0)
Ryzen 7 5700x3d
Thermalright Phantom spirit 120mm EVO
32gb (4x8gb) 3200mhz ram (g skill ripjaws)
sk hynix 500gb ssd
wd blue 1tb hdd (5400rpm I believe)
rtx 3080 (critically this was a cert refurbish card) PNY XLR8 10gb
cooler master 750w gold
2x lian li 140mm fans
3x lian li 120mm fans (both sets of fans are standard ARGB, not the onefan thingy)

From what I can tell I am not reaching power limit. under full load- GPU 320w and CPU 50w-ish

My main confusion is unless i force a heavy overclock overload, I cannot force it to replicate the issue in stress testing. Even temps are fine with the GPU around 72C and CPU about 68C under full load in 3d mark.

I do have an overclock applied (in afterburner) and I have PBO enabled, but from what I can tell this does not change the frequency of the shutdown.

It usually happens after a decent duration of gaming, a couple hours maybe, but again it does not seem to be temps, as they stay pretty steady.

I might also test the voltage on the PSU, but I do not have a multimeter, nor do I see anything off in Hardware Monitor.

I ran memtest and got zero errors, and I let it run every pass. So, my two options now are essentially the GPU and the PSU. The disks seem fine, and under stress testing the GPU is fine. It will do 3dmark and Furmark with a baseline overclock without even breaking 72 degrees under 100 percent usage and 345W. So at this point I 100% think that it is most likely transient spikes from the 3080. 

Please let me know if you have any advice, or any more tests I should run. I can provide minidump files if need be, as well as HWM values.

thanks guys I appreciate the help!

[Sawa Takahashi]

3 hours ago, QuantumCoral99 said:

So, my two options now are essentially the GPU and the PSU.

Easy choice. The PSU costs far less than the GPU. If you can borrow those components, its even more affordable

3 hours ago, QuantumCoral99 said:

So at this point I 100% think that it is most likely transient spikes from the 3080. 

I agree with you. When stress tests do not generate an issue but gaming does, it is usually power spikes. That's why I'd try with a more powerful quality PSU to see if the issue disappears (or at least becomes less frequent).

What was your system before upgrade ? This info could help tell what changed and introduced issues in your system.

[QuantumCoral99]

On 8/28/2024 at 2:10 PM, Sawa Takahashi said:

Easy choice. The PSU costs far less than the GPU. If you can borrow those components, its even more affordable

I agree with you. When stress tests do not generate an issue but gaming does, it is usually power spikes. That's why I'd try with a more powerful quality PSU to see if the issue disappears (or at least becomes less frequent).

What was your system before upgrade ? This info could help tell what changed and introduced issues in your system.

Before the upgrade, the different parts were

 

RTX 2060 

Ryzen 5 3600XT

16gb 3200mhzs ddr4 GSkill ripjaws

4 total 120m case fans (2 CPU cooler, 2 case)

 

everything else was the same

 

The 2060 draws a max 185W so it would make sense for power spikes not to crash system before the upgrade. 750w was a ton of overhead for that system.

 

and I agree i'm thinking its the PSU, i am considering just going with the Seasonic 1000W gold atx3.0, which should give me plenty of headroom along with a much beefier system to handle transient spikes/overvolt/etc.

Amazon.com: Seasonic Focus V3 GX-1000-1000W - 80+ Gold - ATX 3.0 & PCIe 5.0 Ready - Full-Modular - Low Noise - 10 Year Warranty - Nvidia RTX 30/40 Super & AMD GPU Compatible (Ref. SSR-1000FX3) : Electronics

 

It might be a few days (maybe a week or two) before i can grab the new PSU and test.

 

I will update when I can

 

thank you for the help.

[ST1CK]

QuantomCoral I've had the exact same issue, I just switched my to a 1000W 80 gold certified from Gigabyte. My pc keeps crashing and then restarts after 3-5 seconds.
My theory as to what i've read online is maybe a motherboard issue (don't exactly know which), but my temps are still the same, and not a single component is having problems from the outside, what I'm gonna do next month is to buy a new CPU,Mobo and RAM to see if that helps which I hope. 
For me it happens in specific games such as CS2, BF 2042 and Fortnite.
 

[QuantumCoral99]

32 minutes ago, ST1CK said:

QuantomCoral I've had the exact same issue, I just switched my to a 1000W 80 gold certified from Gigabyte. My pc keeps crashing and then restarts after 3-5 seconds.
My theory as to what i've read online is maybe a motherboard issue (don't exactly know which), but my temps are still the same, and not a single component is having problems from the outside, what I'm gonna do next month is to buy a new CPU,Mobo and RAM to see if that helps which I hope. 
For me it happens in specific games such as CS2, BF 2042 and Fortnite.
 

Is your new PSU ATX3.0 certified? I assume it is based on your answer but that detail is very important.

 

It also only happens to me in certain games. From my research, certain PSU, GPU, MOBO combos are worse at handling transient spikes than others, but there is really no way to know which is super unfortunate, as well as transient spikes being virtually undetectable without sepific monitoring hardware. I would also run memtest86 if you haven't already, as if it is an issue with the ram, a psu upgrade wont help.

 

Amazon.com: Seasonic Focus V3 GX-1000-1000W - 80+ Gold - ATX 3.0 & PCIe 5.0 Ready - Full-Modular - Low Noise - 10 Year Warranty - Nvidia RTX 30/40 Super & AMD GPU Compatible (Ref. SSR-1000FX3) : Electronics

[ST1CK]

8 minutes ago, QuantumCoral99 said:

Is your new PSU ATX3.0 certified? I assume it is based on your answer but that detail is very important.

 

It also only happens to me in certain games. From my research, certain PSU, GPU, MOBO combos are worse at handling transient spikes than others, but there is really no way to know which is super unfortunate, as well as transient spikes being virtually undetectable without sepific monitoring hardware. I would also run memtest86 if you haven't already, as if it is an issue with the ram, a psu upgrade wont help.

 

Amazon.com: Seasonic Focus V3 GX-1000-1000W - 80+ Gold - ATX 3.0 & PCIe 5.0 Ready - Full-Modular - Low Noise - 10 Year Warranty - Nvidia RTX 30/40 Super & AMD GPU Compatible (Ref. SSR-1000FX3) : Electronics

The PSU is ATX3.0 certified, what does that have to do with the crashes? (I don't quite follow)
I've ran memtest86 and it did not amount to anything new info about the crashes.
But thanks for the response

[Mark Kaine]

On 8/28/2024 at 6:04 PM, QuantumCoral99 said:

32gb (4x8gb) 3200mhz

try A) without xmp

B) with only 2 sticks 

C) with only 1 stick (test each separately on each slot)

 

[Mark Kaine]

1 hour ago, ST1CK said:

The PSU is ATX3.0 certified, what does that have to do with the crashes? (I don't quite follow)
I've ran memtest86 and it did not amount to anything new info about the crashes.
But thanks for the response

kernel power is typically  =/= psu problem. 

 

it basically just means "your pc shutdown and we don't know why"

 

which is typically what happens when there are ram issues. 

 

 

ps: memtest =/= real life, so rather irrelevant when there are known widespread issues with Ryzen memory controllers, *especially* with more than 2 sticks.

 

lotsa stuff to test for you!

 

 

[Mark Kaine]

1 hour ago, QuantumCoral99 said:

Is your new PSU ATX3.0 certified? I assume it is based on your answer but that detail is very important.

 

It also only happens to me in certain games. From my research, certain PSU, GPU, MOBO combos are worse at handling transient spikes than others, but there is really no way to know which is super unfortunate, as well as transient spikes being virtually undetectable without sepific monitoring hardware. 

 

 

exactly... im not sure who wrote this, but there are only a few possibilities here.

 

RAM (4 sticks, why?)

 

PSU (eh)

 

GPU, specifically the memory (uh, 3080?? refurbished?? oof)  

 

 

 

On 8/28/2024 at 6:04 PM, QuantumCoral99 said:

cooler master 750w gold

this PSU should *not* have issues with RTX 3000 series,  so I'd test this last  

 

[Mark Kaine]

1 hour ago, QuantumCoral99 said:

It also only happens to me in certain games. From my research, certain PSU, GPU, MOBO combos are worse at handling transient spikes than others

confusing thread,  but just like OP, your issue is most likely ram related,  although i really wouldn't use a focus with 3000 series either... it might work,  but its not ideal  (corsair rmx would be better for example) 

[QuantumCoral99]

2 hours ago, ST1CK said:

The PSU is ATX3.0 certified, what does that have to do with the crashes? (I don't quite follow)
I've ran memtest86 and it did not amount to anything new info about the crashes.
But thanks for the response

 

Ah yeah sorry then mate. I have heard from other people that ATX2.0 was not designed with the iffy power design of the 30/40 series taken into proper account. But again I am no expert, i'm just passing along what people have told me. Apologies that I couldnt help

[QuantumCoral99]

1 hour ago, Mark Kaine said:

exactly... im not sure who wrote this, but there are only a few possibilities here.

 

RAM (4 sticks, why?)

 

PSU (eh)

 

GPU, specifically the memory (uh, 3080?? refurbished?? oof)  

 

 

 

this PSU should *not* have issues with RTX 3000 series,  so I'd test this last  

 

not a fan of my choices haha. oh well i thought it made sense. it was an upgrade not a new pc, so it was all cost to performance. I am in university and don't have a ton of money, so I wanted to get the most bang for my buck.

the psu was sold to me from a buddy and was not new.

i couldn't afford to buy a significant gpu upgrade new, and my 2060 was starting to fade. i got a nice deal on the 3080 and its given me no issues (at least explicitly) so far. It is also still under warranty (about 9 months left on it).

4 sticks of ram bc i just bought another identical kit to the one i had, to double the capacity for cheap. thats the only reason- it was not for performance purposes, just capacity.

 

im aware it could be a lot of things, but I cannot pin point the issue so far. The way it shuts off with no warning and no graphical errors made me think it was power, and I know its not temps.

[QuantumCoral99]

1 hour ago, Mark Kaine said:

confusing thread,  but just like OP, your issue is most likely ram related,  although i really wouldn't use a focus with 3000 series either... it might work,  but its not ideal  (corsair rmx would be better for example) 

thank you for the power supply recommendation, i would probably at this point test the psu with a multimeter before buying another psu, if you think that its graphics or ram related that makes more sense.

 

i will try your ram tests, it is tricky though because there is not a set pattern or real indicatior of the crash- it just more or less happens. 

 

it also had not happened again since i made this post

 

would there be some sort of indicator that it is a problem with the memory controller? 

 

like i said i am not an expert but i am doing my best to gather info

[Mark Kaine]

7 minutes ago, QuantumCoral99 said:

im aware it could be a lot of things, but I cannot pin point the issue so far. The way it shuts off with no warning and no graphical errors made me think it was power, and I know its not temps.

this is precisely why im 99.99999% sure its ram.

 

its not defective (probably) its just not fully compatible with the (cpu) memory controller. 

 

easy way to fix/test.  NO overclocks,  NO xmp... test each stick separately (as i said)

 

tldr: you probably have to buy new ram: 2xkit Samsung B-die preferred.

 

[QuantumCoral99]

Just now, Mark Kaine said:

this is precisely why im 99.99999f sure its ram.

 

its not defective (probably) its just not fully compatible with the (cpu) memory controller. 

 

easy way to fix/test.  NO overclocks,  NO xmp... test each stick separately (as i said)

 

tldr: you probably have to buy new ram: 2xkit Samsung B-die preferred.

 

okay thanks mate ill do this and get back to you

 

[Mark Kaine]

4 minutes ago, QuantumCoral99 said:

thank you for the power supply recommendation, i would probably at this point test the psu with a multimeter before buying another psu, if you think that its graphics or ram related that makes more sense.

similar issue as above... your psu is probably not defective,  it just doesn't fully support the gpu (if that's even the issue) 

 

 

AS SAID,  TEST RAM SEPARATELY IN A LIVE ENVIRONMENT OR SIMPLY BUY SAMSUNG BDIE  

 

 

[QuantumCoral99]

Just now, Mark Kaine said:

same issue as above... your psu is not defective,  it just doesn't fully support the gpu (if that's even the issue) 

 

 

AS SAID,  TEST RAM SEPARATELY IN A LIVE ENVIRONMENT OR SIMPLY BUY SAMSUNG BDIE  

 

 

another question:

 

could you give me a breif overview of how to test in a live environment? (or just point me where to look- i can figure it out from there)

 

also by samsung bdie would something like this work to find a compatible kit

B-Die Finder (benzhaomin.github.io)

 

thank you for the help, sorry this thread is so long

[Mark Kaine]

Just now, QuantumCoral99 said:

could you give me a breif overview of how to test in a live environment? (or just point me where to look- i can figure it out from there)

this is what im trying to tell you lol THE ONLY WAY TO TEST IS

 

test each stick separately... that means youll have to boot every time you install another stick and play a game that usually crashes...

 

 

you could also run PRIME95 but that's by far less effective than testing it with a real life application such as a game.

 

3 minutes ago, QuantumCoral99 said:

also by samsung bdie would something like this work to find a compatible kit

B-Die Finder (benzhaomin.github.io)

 

thank you for the help, sorry this thread is so long

yes. exactly.  like my gskill trident Z for example... but it has to be 2x kit BDIE. 

 

and no problem i hope it'll help!

 

 

[QuantumCoral99]

Just now, Mark Kaine said:

this is what im trying to tell you lol THE ONLY WAY TO TEST IS

 

test each stick separately... that means youll have to boot every time you install another stick and play a game that usually crashes...

 

 

you could also run PRIME95 but that's by far less effective than testing it with a real life application such as a game.

 

yes. exactly.  like my gskill trident Z for example... but it has to be 2x kit BDIE. 

 

and no problem i hope it'll help!

 

 

ahh okay mate i see what you are saying- that is my bad sorry for making you clarify a half trillion times lmao

 

ill test the ram like you said- and i was just looking at a gskill trident z 2x16gb 3600 kit that is reasonably priced so ill grab that. thank you again. ill update when all is said and done. hopefully the other guy reads the rest of this as well and that he can find a solution.

 

thanks again, appreciate the work you do on this platform helping people out.

[Mark Kaine]

4 hours ago, QuantumCoral99 said:

not a fan of my choices haha. oh well i thought it made sense. it was an upgrade not a new pc, so it was all cost to performance. I am in university and don't have a ton of money, so I wanted to get the most bang for my buck.

the psu was sold to me from a buddy and was not new.

i couldn't afford to buy a significant gpu upgrade new, and my 2060 was starting to fade. i got a nice deal on the 3080 and its given me no issues (at least explicitly) so far. It is also still under warranty (about 9 months left on it).

4 sticks of ram bc i just bought another identical kit to the one i had, to double the capacity for cheap. thats the only reason- it was not for performance purposes, just capacity.

 

im aware it could be a lot of things, but I cannot pin point the issue so far. The way it shuts off with no warning and no graphical errors made me think it was power, and I know its not temps.

don't get me wrong that's a great setup and i understand, why buy new when you can get basically better performance with used parts for less money, but as you may see now there are some issues with this combo - specifically the ram - which i understand not everyone knows but its a well known flaw with Ryzen cpus, AM5 is even worse in this regard (afaik)

 

but its not an unfixable issue,  theoretically you could probably tune the ram to work, or as said alternatively just run stock speeds, no XMP, maybe just 2 sticks instead of 4...

 

but tuning the ram could possibly take weeks, if you're not experienced this is definitely not worth it imho, especially because success isn't guaranteed either.

 

btw here's my B-die kit... 

 

 

see, it even says "compatible with RYZEN" and that's for a reason,  those rams just work... you can even do crazy overclocks without much effort... 

 

anyhow let us know how this all works out for you