Strange random crashes, inconsistent performance, driver issues, computer haunted?

[GalaxyCHRIS]

I've been having random crashes recently on my system, about one every couple weeks I would say on average. When it crashes, it's a hard crash, it just cuts to black and restarts, no error logs. Some of my games (mostly valorant) have been having strange microstutters that they didnt have before. One of the crashes even went as far as corrupting the GPU driver and forcing me to DDU it. The PC is also no longer detecting my USB mic and not giving any error codes. Are these issues related at all? Any ideas on what could be causing it? The crashes started about 2-3 months ago with a crash in ANNO 1800, the usb mic issue is new.

 

Specs:

• Windows 11 Pro
• Gigabyte Z790 GAMING X AX (latest bios with updated microcode)
• Intel Core i7 13700KF
• ASUS TUF GAMING RX 6800
• 2 x 16GB G.Skill DDR5 6400MT CL 32 RAM
• Be Quiet! Pure Power 850W PSU
• Samsung 980 1tb as boot drive
• Samsung 990 pro 2tb for games

[esCaPade1]

Hi @GalaxyCHRIS

Might be worth checking your RAM - download memtest and put it on a usb boot drive with a tool like ventoy. Boot into it and let memtest run a full test at least twice.

If you do get errors, test one stick at a time.

Hope you find the answer.

[Sawa Takahashi]

Possible suspects are RAM (like other poster said), SSD and PSU. Try testing them and switching if you can get your hands on spares.

It may be a good idea to use monitoring software like HWinfo64 and generate a log of temps and voltages so you know you are not dealing with a bad cooler, dry thermal paste and an unplugged fan.

Good luck !

[GalaxyCHRIS]

1 hour ago, esCaPade1 said:

Hi @GalaxyCHRIS

Might be worth checking your RAM - download memtest and put it on a usb boot drive with a tool like ventoy. Boot into it and let memtest run a full test at least twice.

If you do get errors, test one stick at a time.

Hope you find the answer.

I ran memtest, finished with 0 errors, so for now I will rule out the RAM. Thank you for the suggestion though! Really appreciate the help.

[GalaxyCHRIS]

1 hour ago, Sawa Takahashi said:

Possible suspects are RAM (like other poster said), SSD and PSU. Try testing them and switching if you can get your hands on spares.

It may be a good idea to use monitoring software like HWinfo64 and generate a log of temps and voltages so you know you are not dealing with a bad cooler, dry thermal paste and an unplugged fan.

Good luck !

I don't really know how to test for integrity on an SSD, but I did check HWinfo for the info and the 980 is on 95% drive health with the 990 Pro on 99%, both have no drive errors reported. If this is the wrong way to test this, please let me know. 

 

I ran Cinebench multi core and furmark at the same time in order to load the entire system to the max. I did that for about 20 minutes and it did not cause any issues. The CPU did go up to 100°C but that it has been doing since i built the system. I'm running a be quiet 360mm AIO, this chip is just really hot. The thermal paste is about a month old. GPU temps were completely fine, the fan didnt spin up past 50%. The one thing i noticed is that in cinebench, whether furmark is running or not, it tends to have some squares that render extremely slowly compared to the others. Like maybe at a quarter of the speed. HWInfo does not show any cores throttling though, which is interesting. Is that normal or does that point to a CPU problem? I'm pretty confident it was not doing this when i built the system 2 years ago.

[esCaPade1]

It's pretty normal for some squares in cinebench to be slow - it's more about the complexity of that part of the scene than anything to do with your pc. Assuming it's the same square each time, that is.

If you built the system, maybe you can do so again? Pulling out each component and replacing it could conceivably fix a hard to trace issue.

It would also be good to see what happens with a fresh install of windows - if you can back up your current install, there'd be no harm to it. 

 

[Sawa Takahashi]

2 hours ago, GalaxyCHRIS said:

I don't really know how to test for integrity on an SSD, but I did check HWinfo for the info and the 980 is on 95% drive health with the 990 Pro on 99%, both have no drive errors reported. If this is the wrong way to test this, please let me know. 

The SMART monitoring you checked is a good approximation of the remaining life of your SSDs. There's been some problems with some Samsung SSDs so you can't completely rule them out yet.

2 hours ago, GalaxyCHRIS said:

The CPU did go up to 100°C but that it has been doing since i built the system.

I agree, these CPUs are rebranded toasters  Max temp is indeed 100C for that chip but after 20 minutes of stress testing, that's to be expected.

 

1 hour ago, esCaPade1 said:

It would also be good to see what happens with a fresh install of windows - if you can back up your current install, there'd be no harm to it. 

I must add that this is a good idea but if you can use a spare NVMe it would also rule out the SSD health at the same time. I personally use cheap WD blue drives and they never failed me.

 

Again, I hope this can help you.

[GalaxyCHRIS]

On 12/1/2024 at 9:52 PM, esCaPade1 said:

It's pretty normal for some squares in cinebench to be slow - it's more about the complexity of that part of the scene than anything to do with your pc. Assuming it's the same square each time, that is.

If you built the system, maybe you can do so again? Pulling out each component and replacing it could conceivably fix a hard to trace issue.

It would also be good to see what happens with a fresh install of windows - if you can back up your current install, there'd be no harm to it. 

 

sry forgot to reply - I ended up reseating the RAM yesterday and so far I haven't crashed again. The frame delivery feels significantly smoother also. I will wait a little bit to see if it stays like this or gets issues again, and if it breaks again, i will try a fresh windows install next.

[esCaPade1]

That sounds promising! Crossed fingers!