Question PC keeps hard freezing / crashing ?

[L0N3STARR]

Intel Core i9-12900k CPU
EVGA RTX 3090 FTW3 Ultra GPU
Corsair Vengeance 5600 MHz DDR5 RAM
Gigabyte Z690 Aorus Ultra MOBO
EVGA Supernova 1000 G+ 80 Plus Gold PSU
ROG Strix Helios Gundam Edition Case
Corsair iCUE H150i Elite LCD CPU Water Cooler
WD Black SN750 500 GB NVMe OS SSD
Samsung 970 EVO 2 TB NVMe SSD
2x Seagate Exos X14 12 TB HDD

When I say crashes I mean it hard freezes and is only recoverable by hard reboot.

I've tried debugging everything I can think of.

Troubleshooting steps performed:

• Stress testing CPU - no issues
• Memory check - no issues
• Installing OS on other SSD - still crashes
• Flashed MOBO BIOS - still crashes
• Reducing GPU Boost by 150 MHz, Memory by 200 MHz, power to 85% - still crashes
• Updating GPU BIOS - still crashes
• Utilizing CPU internal graphics - still crashes
• Switched power outlets - seems to crash less often, but still crashes.

It usually crashes when stress testing the GPU, so I thought that was the culprit, but since it's still crashing when using CPU graphics, I don't think that's the issue.

I am thinking of redoing the thermal paste on the CPU since maybe the stuff pre applied to the CPU cooler sucks, and also thinking of trying a brand new power cable but other than that I'm out of ideas.

 

[Misgar]

You stress tested the CPU with no issues, so the pre-applied paste on the cooler must have been doing its job. Still, it's one more thing to check I suppose. Reapply some decent TIM with a high W-mK rating, e,g. Kryonaut or your favourite high quality paste.

Have you overclocked the CPU. If so reset to default boosting.

Stress testing RAM seems OK, but have you run MemTest86+? Are you running with XMP overclock or at stock frequency?

You appear to have ruled out the RTX 3090 by switching over to the iGPU.

Do you have an alternative PSU you could try? If it's of much lower capacity, e.g. 500W, leave the RTX 3090 disconnected.

There's a slight chance there might be a slightly bent pin in the CPU socket. Remove the CPU and check.

Can you try the system in another room, or better still a different building? If the power outlet has badly worn contacts, you'll get slight arcing and mains spikes on heavy loads. Do not use cheap surge protectors or flaky mains distribution strips. Plug the ATX PSU directly into the wall outlet to prove it's not a faulty mains extension block.

Definitely try a different power cable. They're usually reliable, unless you've got a particularly nasty cheap badly manufactured item. If so, chuck it in the bin. They're dangerous.

 

[L0N3STARR]

You stress tested the CPU with no issues, so the pre-applied paste on the cooler must have been doing its job. Still, it's one more thing to check I suppose. Reapply some decent TIM with a high W-mK rating, e,g. Kryonaut or your favourite high quality paste.

Have you overclocked the CPU. If so reset to default boosting.

Stress testing RAM seems OK, but have you run MemTest86+? Are you running with XMP overclock or at stock frequency?

You appear to have ruled out the RTX 3090 by switching over to the iGPU.

Do you have an alternative PSU you could try? If it's of much lower capacity, e.g. 500W, leave the RTX 3090 disconnected.

There's a slight chance there might be a slightly bent pin in the CPU socket. Remove the CPU and check.

Can you try the system in another room, or better still a different building? If the power outlet has badly worn contacts, you'll get slight arcing and mains spikes on heavy loads. Do not use cheap surge protectors or flaky mains distribution strips. Plug the ATX PSU directly into the wall outlet to prove it's not a faulty mains extension block.

Definitely try a different power cable. They're usually reliable, unless you've got a particularly nasty cheap badly manufactured item. If so, chuck it in the bin. They're dangerous.

All good suggestions! Thank you!

I have not overclocked the CPU. I don't think this MOBO overclocked by default but honestly I'm no overclocking guru. I've done GPU overclocking in the past but never CPU. I looked at the MOBO's BIOS tools for overclocking and I'm not 100% sure if it's doing anything by default. Will send a pic today.

I did just use the built in Windows utility for stress testing the memory so I can try MemTest86+! XMP is disabled in the MOBO BIOS.

I don't have an alternate PSU. Do you think that's likely enough the issue that I should buy a different one, if trying all the rest of this doesn't solve it?

I used one of the contact frames for 12th gen Intel processors when I put it together. I'd hope that I wouldn't have ended up bending a pin, but will definitely check it. If a pin is bent, I'll be asking how to resolve that next!

I can at least try in a different room. I think the quality of the cable is okay, but it is pretty old so no harm in trying a new one.

Will report back.

 

[Misgar]

The contact frame shouldn't have bent a pin in the CPU socket, unless you dropped it into the socket during assembly.

If you have a bent pin in the socket, be VERY, VERY careful. Countless people have killed their motherboards by ham fisted attempts to straighten a pin. By all accounts, the pins snap offs with barely a nudge from a sharp sewing needle.

 

[L0N3STARR]

You stress tested the CPU with no issues, so the pre-applied paste on the cooler must have been doing its job. Still, it's one more thing to check I suppose. Reapply some decent TIM with a high W-mK rating, e,g. Kryonaut or your favourite high quality paste.

Have you overclocked the CPU. If so reset to default boosting.

Stress testing RAM seems OK, but have you run MemTest86+? Are you running with XMP overclock or at stock frequency?

You appear to have ruled out the RTX 3090 by switching over to the iGPU.

Do you have an alternative PSU you could try? If it's of much lower capacity, e.g. 500W, leave the RTX 3090 disconnected.

There's a slight chance there might be a slightly bent pin in the CPU socket. Remove the CPU and check.

Can you try the system in another room, or better still a different building? If the power outlet has badly worn contacts, you'll get slight arcing and mains spikes on heavy loads. Do not use cheap surge protectors or flaky mains distribution strips. Plug the ATX PSU directly into the wall outlet to prove it's not a faulty mains extension block.

Definitely try a different power cable. They're usually reliable, unless you've got a particularly nasty cheap badly manufactured item. If so, chuck it in the bin. They're dangerous.

After some more testing I'm thinking the issue is the circuit my wall outlet is on. I have a lot of electronics on that same circuit. I only have my PC and monitor (Spectre P30) on that outlet, but is it possible that I'm not able to draw enough power through that circuit and it's causing the crash? I'd have thought that would flip the breaker but maybe I'm wrong. I'm an electrical engineer, but not an electrician so I don't know enough about wiring in a home to be very useful.

I haven't had a chance to move everything to test elsewhere yet so I still need to report on that. Running MemTest86+ now.

Thoughts?

 

[Misgar]

The electronics labs I've worked in have test benches with 12 or more switched mains outlets, so don't worry about lack of current, just because other equipment is plugged in nearby. As you said, if the total load exceeded the breaker limit, it would trip.

Of course the more mains cables and adapters you plug in series (daisy chain) between the wall outlet and the computer, the higher the impedance of the circuit.

Every series plug and socket connection adds contact resistance, even if it's only a few tens of milliOhms. Each metre or foot of Copper cable adds a small amount of extra resistance.

Depending on where you live, your property will be wired up with a ring main on each floor, or individual spurs back to the distribution panel. Check the regs for your country.

If you want to discount a motley assortment of extension leads, reconnect your PC directly to a wall socket close to the mains distribution panel, where mains enters your property.