Crashing to BIOS under load no BSOD - can't figure it out

[tribal canopy]

Will preface this by saying, I'm not booted to the bios menu upon crash but I'm booted to the manufacturer logo (ASrock).

Specs:
Win 11
5800x
b550i Phantom Gaming itx/ax
16x2 GB 3600MT CL18

• XMP profile is at 1.35 volts but I upped it to 1.39, stopped crashing but started up again
• Exact kit is: TTCED432G3600HC18JDC01 (T create)
  RTX 3080 10GB LHR (Galax 3080SG)
• Connected via a PCIE 4.0 Riser cable
  Lianli SP750

1TB 980 Pro
1TB Adata Falcon

Issue:
During graphic intensive games my computer will crash to bios without blue screening. If a blue screen appears it's only for a second and it's not clear / glitchy.
I can't recreate it in furmark, cinebench, etc.

I have all unnecessary log-in applications disabled.

My avg temp of my gpu is 83 degrees underload in furmark with the hotspot reaching 100 degrees (17 degree celcius delta). I guess this is bad but it's inconsistent and doesnt reach that temp when I'm actually playing games.

Typically crashes under high gpu load I suppose but I played NMS and Baldurs gate for hours without issue.
Interestingly crashes more frequentlythan anything else so far when I use kovaaks.
I mainly play valorant at the moment, there's not a pattern i see to crash such as larger maps, certain smokes, although I do notice that cypher crashes my game more often both when I do a custom (just me in a map testing stuff as cypher) or in matchmaking.

I dont have anything overclocked and have only ever used the XMP profile and PBO2 tuner to undervolt the cpu by like -5.
I have no OC software on my computer bar the nvidia beta app and MSI afterburner, which i use purely for performance metrics.

Things I have already tried (not ordered):

1. Disabling XMP
2. Upping DRAM voltage (below 1.4 volts tho)
3. resetting bios to older version
4. DDU (multiple times across windows installs)
5. Stress testing with:

• Cinebench
• Furmark

6. Memtest85, for both non-xmp and 1.39 volt configuration, both passed the full test
   -for the xmp config it spat out 500 errors in like 10 minutes
7. sfc scannow
8. checking the minidump in windbg
9. Clean Install of win11
10. Checking health of ssds in crystal disk

I havent done anthing physical to the computer yet such as only booting with 1 stick of RAM due to the limited time i have on my hands. I have to play games until it crashes which isn't always consistent.
Everything looks in place at a glance and I made sure cables and components were seated properly (reseated RAM).

[tawny charm]

Gut says PSU

Or RAM since memtest failed

In fact thats sorta a dead giveaway

Perhaps try one stick of ram, xmp enabled, dimm slot 1
Then try same ram, different dimm slot

Then try other stick, dimm 1 then dimm 2

Hopefully doing that will reveal if there is a specific stick or slot that fails

Then if you find a slot/stick combination that works, try running just that stick playing valorant for a week or so

If it seems to resolve the issue, then you know. If it happens again, then the ram is likely a red herring

[tribal canopy]

will try this when i have time, prolly not till wednesday

thanks 👍

[tribal canopy]

Both still crashing honestly

It passed the memtest again

these are my gpu temps right before the crash

avg, memory junc, hotpost and thermal limit

[tawny charm]

So unlikely gpu overheating

[tawny charm]

Wasn’t it failing with stock xmp enabled before? So it works now with only 1 stick?

Or does each stick still fail regardless of dimm slot, with xmp enabled (stock no voltage mods)

[tribal canopy]

tried both sticks

crashes with both

in each slot

doesn't really matter

[tawny charm]

Memtest you mean?

[tribal canopy]

nope just using

[tawny charm]

Ah

But im still curious about the xmp results

[tribal canopy]

like stick 1 to slot a, then b

stick 2 to slot a, then b

both configs crash at default, xmp and 1.38v

[tawny charm]

Yeah thats weird

[tribal canopy]

using spacemarine 2 to force the crash

works well

[tawny charm]

Honestly just buy another different ram kit from amazon and try it. If they also fail, just return them

[tawny charm]

Oh space marine crashes anyway

[tribal canopy]

it crashes the whole computer

[tawny charm]

So that could be a false positive

[tribal canopy]

ah i see

im still not getting BSODs

[tawny charm]

Oh it doesnt crash the whole pc nvm

[tribal canopy]

just crashes past bios

and my minidump doesnt get updated since no blue screen

super weird

[tawny charm]

Yeah thats odd

Im feeling psu or ram honestly

[tribal canopy]

sometimes I see a bluescreen but only for a second then it crashes

[tawny charm]

Imagine psu power cuts

[tribal canopy]

weird af

possibly

[tawny charm]

Youd get no logs

Cause it just dies instantly

[tribal canopy]

sigh

should I take it to a tech and ask them to check

[tawny charm]

Could be worth

Do you have another psu to use perhaps

[tribal canopy]

cause im gonna be building a new pc and need to know if i can reuse this psu or if i need a new one

nope

[tawny charm]

Yea then probably a good idea

Alternatively buy one on amazon and abuse their lax returns lol

[tribal canopy]

was thinking about that 💀

[tawny charm]

But if the techs price is reasonable then its probably worth it

[tribal canopy]

i dont have the box for my psu, but it's less than 5 years old so it should still be covered by the warranty

[tawny charm]

Im suspecting psu or ram

Last possibility would be the mobo but unlikely if both dimm slots fail

Usually there might be 1 bad dimm slot

[tawny charm]

Yeah likely

Do you overclock at all

[tribal canopy]

nope

[tawny charm]

I found with my 6950xt and 5600x i could hit 650w pretty easily

And your cpu is +80w on mine, gpu roughly the same

So you might be getting transient spikes above 750w

[tribal canopy]

i tried pbo2 tuner to undervolt a little but i dont have it installed

[tribal canopy]

should i try setting a curve in msi afterburner?

[tawny charm]

Hmm that may help a tad

One of these would be ideal

Theyre about $10

Plug it between your pc and the wall

[tribal canopy]

oh true

[tawny charm]

And itll show you the real wattage usage

Space marine will push it to the limits

And it should show highest usage as a stat

Theres a chance its just going past 750 at some point under heavy load

But even if thats not happening - it could still be a psu fault

[tribal canopy]

makes sense

one other thing I realized is that it was plugged into a surge protector, which has traditionally been fine for me

this one is rated at 2000 watts and I only have my monitor + my pc plugged in so it's not even reaching 1k watt power draw

i've plugged directly into the wall now though

2500w*

[tawny charm]

ah yep

[tribal canopy]

undervolted and it's all g

my max clock is like 1740 - 1790, factory spec puts it at 1725, i just kept it at 1740

undervolted to 875mv from 950 and it's stable in furmark and space marine 2

can prolly go lower but power draw is a lot less

leads me to believe it's psu related

[tawny charm]

I believe so yep

Next time just get an overkill wattage one imo

Ive learned over time it basically never hurts to go bananas with the PSU wattage

[tribal canopy]

Ye

Thought it was fine when I bought it cause it ran flawlessly, thought 250w overhead would be decent

[tawny charm]

Yeah 750w is pretty beefy