PC suddenly shutting down in certain games/programs

[LAA]

Hey,

So I've noticed a few days ago, my PC can suddenly turn off after playing fortnite for a little while. In a strange state too, all the fans and etc. stop inside the case and no signal display, but the power light around the power button stays on, pressing it again starts the booting sequence however. Weirdly, I at least haven't seen the issue playing other games, and I did look into CPU OCing after this doing multiple stress tests for hours, and even with that I haven't seen my PC turn off, crashed windows at best. Anyway, since then I've tried multiple things.. setting CPU back to stock settings, still have shutdowns Performed a bunch of stress tests like Cinebench, OCCT, Prime95, no issues. Repair install for Windows 10, same thing. I should mention while playing Fortnite, I haven't seen the CPU/GPU get over 80 degrees C, so not sure it's an overheating issue at least where they're concerned. Only thing I'm wondering now is whether it's a PSU issue, but no real idea how to investigate that. Though it's a 1000W supply and my CPU/GPU were barely reaching 500W during Fortnite, but that is just a theory at this point.

Since then I've found the power test from OCCT can cause the PC to shut down too, but it seems a little unpredictable, sometimes it during the first test, around 30-40 mins. Sometimes it takes 2 tests, again around 30-40 mins in.

I've also noticed doing a memory test did give me errors using OCCT memory test and Karhu RAM test, so I tried replacing with RAM that gave no errors from another computer, ran same memory tests, fine. Ran 2 power tests, eventually turned itself off.

Running out of things to try and diagnose what's causing the issue or how I should go about it.

 

Specs:

-AMD Ryzen 5950x

-H115i Platinum AIO

-MSI 3090 Gaming X Trio

-G.Skill 32GB (2x16) RAM

-Corsair RM1000X PSU

-X570 Aorus Master Motherboard (F34 BIOS)

-Coolermaster H500M case

-Sabrent Rocket 2TB SSD PCIe 4.0 NVME SSD (Windows installation here)

-2x Crucial 2TB MX500

-1 3TB HDD

 

Would really appreciate any suggestions, info or advice here.

Thanks

[LAA]

I've attached a log from HWINFO if this shows anything noteworthy?

OCCTTemp2.CSV

[Mark Kaine]

Do you have an older gpu you could try?

 

because i think this is either psu, or maybe even the gpu itself… 

 

yes, i know the psu should be good, but then the system shouldn't randomly shutdown either…

 

if you dont have anothor gpu, try undervolting your 3090.

 

power limit, 70/80% or so, core clock -100, memory maybe also -500 or so. Gotta experiment with that a bit perhaps.

 

Goal is to reduce possible spikes, so power limit and downclock is how you do that.

[Kisai]

13 hours ago, LAA said:

Hey,

So I've noticed a few days ago, my PC can suddenly turn off after playing fortnite for a little while. In a strange state too, all the fans and etc. stop inside the case and no signal display, but the power light around the power button stays on, pressing it again starts the booting sequence however

90% of the time, this is a PSU issue, as in insufficient power.

 

It can be a GPU issue, (or more likely a factory OC'd card), but generally if it's a driver issue, the GPU will reset and the OS will keep on trucking. If the GPU itself is failing there will be glitches like if the VRAM is too hot or malfunctioning, you'll start seeing corrupted textures. If the GPU itself stops malfunctioning if you "underclock" the card, then the problem is centered on the card.

 

You have to realize that OC'ing the card can easily spike the load the card has.

 

As for any other parts. RAM and CPU have specific malfunctions.

CPU will just, flat out BSOD, usually at a certain thermal point, and usually crashes with "freezing"

RAM on the other hand will usually fail at the same point, a lot.

 

 

[LAA]

46 minutes ago, Mark Kaine said:

Do you have an older gpu you could try?

 

because i think this is either psu, or maybe even the gpu itself… 

 

yes, i know the psu should be good, but then the system shouldn't randomly shutdown either…

 

if you dont have anothor gpu, try undervolting your 3090.

 

power limit, 70/80% or so, core clock -100, memory maybe also -500 or so. Gotta experiment with that a bit perhaps.

 

Goal is to reduce possible spikes, so power limit and downclock is how you do that.

Thanks for the post! So I did order another PSU which I could try replacing, (same model as mine), but want that to be last resort so I can return it if its not the issue after all.

Other question about that too is if I could keep the wires connected from my current PSU and just connect them to the new PSU?

I've read you shouldn't do that in a lot of cases but not sure in the case where its the same model, another RM1000X in my case. Though not sure if it's worth doing anyway just to eliminate wire issues too.

 

As for the GPU, I do have a 1060 I could try swapping in to see. Shame if I have to undervolt/downclock my 3090, I haven't personally OCd, so should all be default. I may need more advice with how to do this however as I haven't done it on a GPU before at least.

19 minutes ago, Kisai said:

90% of the time, this is a PSU issue, as in insufficient power.

 

It can be a GPU issue, (or more likely a factory OC'd card), but generally if it's a driver issue, the GPU will reset and the OS will keep on trucking. If the GPU itself is failing there will be glitches like if the VRAM is too hot or malfunctioning, you'll start seeing corrupted textures. If the GPU itself stops malfunctioning if you "underclock" the card, then the problem is centered on the card.

 

You have to realize that OC'ing the card can easily spike the load the card has.

 

As for any other parts. RAM and CPU have specific malfunctions.

CPU will just, flat out BSOD, usually at a certain thermal point, and usually crashes with "freezing"

RAM on the other hand will usually fail at the same point, a lot.

 

 

Thanks for the post also! So I haven't personally OCd my GPU, but possibly factory OCd. I haven't noticed any textures ot glitches or etc. either, so it doesn't appear to be noticeably failing at the least.

 

I did actually notice my PC freeze yesterday after swapping the new RAM (older RAM technically) in, just after booting OCCT actually. It is pretty rare to see that happen though.

[Mark Kaine]

56 minutes ago, LAA said:

So I did order another PSU which I could try replacing, (same model as mine), but want that to be last resort so I can return it if its not the issue after all.

Other question about that too is if I could keep the wires connected from my current PSU and just connect them to the new PSU?

I've read you shouldn't do that in a lot of cases but not sure in the case where its the same model, another RM1000X in my case. Though not sure if it's worth doing anyway just to eliminate wire issues too.

well, yes, you should always replace the cables, you cant know if manufacturer changed some pinouts (sometimes they even change the actual manufacturer…)

that said, if its a psu issue, replacing with the same exact psu might not be all that helpful…

 

58 minutes ago, LAA said:

As for the GPU, I do have a 1060 I could try swapping in to see. Shame if I have to undervolt/downclock my 3090, I haven't personally OCd, so should all be default. I may need more advice with how to do this however as I haven't done it on a GPU before at least

yeah, well then try the 1060 for a few days, that should give you some insight if its the psu/gpu combo, or maybe something else.

 

and sure i can try to help, but the undervolting/clocking is more a test than a permanent fix, however just swapping in your 1060 is easier, and imo a far better test too.

 

you could always try the undervolting with the 3090 later if you're still not sure. 

[LAA]

46 minutes ago, Mark Kaine said:

well, yes, you should always replace the cables, you cant know if manufacturer changed some pinouts (sometimes they even change the actual manufacturer…)

that said, if its a psu issue, replacing with the same exact psu might not be all that helpful…

 

yeah, well then try the 1060 for a few days, that should give you some insight if its the psu/gpu combo, or maybe something else.

 

and sure i can try to help, but the undervolting/clocking is more a test than a permanent fix, however just swapping in your 1060 is easier, and imo a far better test too.

 

you could always try the undervolting with the 3090 later if you're still not sure. 

OK, I switched to the 1060 and running OCCT power test, so see how it goes. I think with 1060 now with it not drawing as much power though whether it will stress the PSU as much though?

I could have got another PSU, but it's not like I'm using close to 1000W anyway and Corsairs still seem held in high regard, and not seen anyone else with issues with it anyway.

 

[Mark Kaine]

54 minutes ago, LAA said:

will stress the PSU as much though?

no, thats exactly the point. 

 

if its not crashing it means -most likely- the psu / power delivery isnt sufficient (for the 3090) for whatever reason.

 

hence, hopefully, replacing the psu would fix it. but, if you replace it with the exact same psu and the issue  persists, we're back to square one lol, so i do not recommend that.

 

get a rmi or hxi instead imo… 

 

[LAA]

38 minutes ago, Mark Kaine said:

no, thats exactly the point. 

 

if its not crashing it means -most likely- the psu / power delivery isnt sufficient (for the 3090) for whatever reason.

 

hence, hopefully, replacing the psu would fix it. but, if you replace it with the exact same psu and the issue  persists, we're back to square one lol, so i do not recommend that.

 

get a rmi or hxi instead imo… 

 

Gotcha thanks for info!

I was looking into those but according to this RMX/RMi are the same in terms of performance, just RMi has corsair link

https://www.anandtech.com/show/9625/the-corsair-rmx-and-rmi-1000w-power-supply-review/5

 

EDIT: Sorry one other thought I just had which I'm gonna feel really stupid about if this is the cause.

So my 3090 has 3 8 pin ports and I use 2 wires for two of them and the third is split from the second. Would I need to use 3 seperate wires for this?

[Mark Kaine]

31 minutes ago, LAA said:

Gotcha thanks for info!

I was looking into those but according to this RMX/RMi are the same in terms of performance, just RMi has corsair link

no prob! and yeah, thats possible , i wasnt sure if rmi is really the same, then a hxi, im just saying *not* exactly the same psu, chances are it would do nothing (since its the same )

[Mark Kaine]

48 minutes ago, LAA said:

EDIT: Sorry one other thought I just had which I'm gonna feel really stupid about if this is the cause.

So my 3090 has 3 8 pin ports and I use 2 wires for two of them and the third is split from the second. Would I need to use 3 seperate wires for this?

yeap. you should use 3 seperate cables in this case…

 

BTW, I just looked this up, you can find a million "posts" saying rmx/i are exactly the same except the "link stuff"… but, the rmx is single rail (apparently) the rmi is multi - single rail switchable… hence they *cannot* be *exactly* the same… !

 

btw i have a rmi and hell i will switch to single rail - only over my dead body

 

 

 

well…  then, try with 3 pcie cables, i think theres a good chance it fixes your issues *fingers crossed* !

[LAA]

5 hours ago, Mark Kaine said:

yeap. you should use 3 seperate cables in this case…

 

BTW, I just looked this up, you can find a million "posts" saying rmx/i are exactly the same except the "link stuff"… but, the rmx is single rail (apparently) the rmi is multi - single rail switchable… hence they *cannot* be *exactly* the same… !

 

btw i have a rmi and hell i will switch to single rail - only over my dead body

 

 

 

well…  then, try with 3 pcie cables, i think theres a good chance it fixes your issues *fingers crossed* !

Really hoped this could have been it.. but ran power test for an hour and a half and now it's turned off again.

Funny this time I did add new RAM which has RGB and I can see they stay lit, so everything turns off except the RAM lights and the light around the power button.

 

Is the next course of action replacing PSU?

[LAA]

As an update. Ended up replacing PSU (Same RM1000X), fingers crossed looks fixed.. ran a power test with the side panels off and another 2 with them on (which it hasn't been able to get past yet), and all fine.

I'll keep an eye on it and see how it goes

[LAA]

New update. So I haven't yet seen the PC shutdown, but I did just encounter the PC freeze while just having steam download updates and this was just shortly after turning it on after being off for a while.

I have seen it happen once before shortly after opening OCCT, which I thought may have been slow one off as I just installed new RAM, but this seems a little more concerning.