Ryzen 2700X sytem suddenly getting hard resets after 2 years stable

[Kolath]

Specs:
CPU: AMD Ryzen 7 2700X
Mobo: ASRock X470 Master SLI
GPU: EVGA RTX 2070 XC
RAM: 2x8GB G.Skill F4-3200C14-8GVK
PSU: EVGA SuperNOVA 750 G3, 220-G3-0750-X1, 80+ GOLD
OS: Windows 10 Home (Version 2004)
BIOS is original: P1.00 from 3/28/2018

Summary
My gaming computer has been pretty much rock steady since I put it together in 2018 but in the last few days it has started randomly restarting (no BOSD, full instant crash and restart) while gaming. Sometimes immediately, sometime after 15-20 minutes.

Recent History
I have been running Cyberpunk 2077 on max settings (except only low ray tracing) successfully since January. In March I did notice a few random restarts, but chalked them up to the game. But now it crashes constantly even playing much less intense games like Tabletop Simulator and Stellaris.

What I've tried

Basic Troubleshooting

• Blowing out dust from the CPU cooler with compressed air + cleaning the intake filters - Did not appear to help noticeably
• Upgraded the RTX drivers to the latest release - No effect

Memory Testing

• Memtest86 - No errors, ran for 2 hours 4 cycles no problem

CPU Testing

• Prime95 - By itself ran tonight for a hour without issue (will try running overnight) - CPU Tdie temp at ~82 C in HWInfo (with case side ON)
• Prime95 Overnight run 8 hours no issues - temp 53 C Tdie with case side open

GPU Testing

• OCCT - VRAM test of 88% memory - GPU temp 57 C - No errors after 1 hour
• OCCT - GPU test default (shader level 3) - GPU temp 75 C - No errors after 1 hour
• OCCT - GPU test Shader level 8 - Crash after about 60 seconds
• FurMark - Ran for about 40 minutes until I stopped it - GPU temp at ~ 75C ("Hotspot" temp at ~85 C?)
• Update: FurMark now again causing it to restart - Ran for about 8 minutes then restart; Now immediate restarts.

GPU + CPU Testing

• Furmark + P95... mixed results:
• When I start Furmark first, then added P95... it ran for about 10 minutes then stopped it to try reverse order...
• P95 first then add Furmark = instant restart
• Then tried a restart to try Furmark first = instant restart

PSU?

• OCCT "PSU Test" - 40 minutes and counting no restart - CPU and GPU maxed. GPU 77C and CPU 72C

I have gone ahead and ordered some new thermal paste and an addition case fan in case CPU temps and general temps are an issue (I wanted to add another fan to my case anyway to get to positive pressure.

In reading a few other posts I've seen some folks be concerned potentially about PSUs when there are random restarts. Any suggestions on how to test or isolate that?

 

[00MJR]

Dumb thought... but did you check the BIOS(CMOS) battery voltage? I'd go into the BIOS(CMOS) and safe your settings into a OC profile.. remove the battery and replace if needed. A dead or 'dying' battery can cause all sorts of weird issues...

A new battery will show over 3v but I've seen issues arise when this voltage hits 2.8v... if you don't have a way of measuring it, I'd just replace it regardless.

it could be something else... but since your system is 2+... I've seen batteries last a few month to a many years. all depends on use patterns

 

[Kolath]

Huh, that did not ever occur to me. I can give that a try!

I am not overclocking, everything is factory stock.

 

[00MJR]

It's a CR2032. Get a duracell or energizer if you can.

 

[Kolath]

Update:

So I seem to again be able to reproduce the crash using Furmark. Doing the default stress test it ran for about 8 minutes then computer rebooted.Then running the stress test instantly rebooted.

 

[00MJR]

hmm.. any extras PSUs sitting around?

 

[Kolath]

Tried OCCT GPU test with Shader complexity 8 and it crashed.

So... looking like this might be a GPU Issue :-(

 

[00MJR]

Yeah GPU or PSU..

 

[Kolath]

I have an EVGA GTX 970 that I could swap in.

I also have an old Corsair TX650W PSU that I could disentangle from an older desktop.

Should I try swapping GPU and run stress tests on the older GPU? Would that isolate PSU or GPU?

 

[00MJR]

I'd do the PSU first.. that'll save you some time

 

[EarthDog]

I'd stop running Furmark. Both Nvidia and AMD called them 'power virus' and suggested not to use them. Worth noting, particularly on Nvidia GPUs, is that the stress test is so hard that commonly, the clocks lower hundreds of MHz just so it fits within the power envelope (so what is it really testing?). Test with looping 3DMark or something else other than Furmark.

As far as the issue, since it seems to always happen, in some manner, WITH a GPU test, I'd lean towards it being the GPU. I have that PSU, it's a high-quality unit that can output more than its nameplate. I run a 10980XE (overclocked to 4.6 GHz or so) and an RTX 3080 on it without issue. That doesn't mean the unit can't be bad, however. But I would think it's the GPU.

Did you update your GPU driver? If you're using the latest, did you try rolling back?
What does the Windows Event log say happened?

Unrelated, update your motherboard BIOS as well. Your sig says the original and... I'd want to be on something a lot newer if not the newest out there.

 

[00MJR]

I'd suggest not changing drivers, updates, etc. Keep that all the same. If you're wanting to test component failures, you don't want to change how they're currently running.. and my reason for the PSU swap first is that when you change the GPU out, you're changing the amperage pull and the PSU may not fail(if it's bad) and then your swapping the PSU anyways.

 

[EarthDog]

Buuuuuuuuut, we don't know if it is hardware or software issue.

Isn't it easy, less time-consuming, and free to update drivers/BIOS? If the issue still happens after the 2 minute driver update, then chances are it's hardware, then you take the time to swap out PSUs or w/e.

Note...I don't think there's a wrong way here...I'm just not sure if it is hardware or software and updating/rolling back GPU drivers is a painless way to test that out(rule out borked/bad driver) without affecting power testing/swapping out PSU/the next step.

 

[00MJR]

You could check event viewer and see if there's something showing in the hardware failures.. or if you got any GPU driver issues in there. If everything seems OK and the restart doesn't show a reason in event viewer.. you're most likely facing a hardware issue.

No right or wrong way, just trying to keep the troubleshooting straightforward and simplified. Adding things into the mix usually results in more testing and headaches.

This sort of work is part of what I do everyday, so I'm trying to share what steps I'd do when it comes to testing hardware failures.

 

[Kolath]

I was on the Nvidia drivers from January and tried updating to March to see if it would resolve the crash. The driver version didn't seem to matter.

But will definitely try the BIOS update as well as the Mobo battery replacement.

But I suspect GPU probably at fault here just because that seems to be the common denominator.

 

[Kolath]

Updates!

I definitely think the issue is the PSU

Also, I was able to restore back to my original OC forum username! This is the same guy as Bestekov.

Things I tested
Upgraded the BIOS to P3.4 (last recommended version for Pinnacle CPUs) - No effect
At EarthDog's suggestion, I started using 3DMark instead of Furmark. It would hard reset in the first 2 minutes of the first test :-(

BUT I swapped in my old Corsair 650W and 3DMark runs like a champ no errors at all!

So, I guess it is time to RMA the EVGA PSU?

 

[00MJR]

You're welcome! Yeah luckily evga is easy to work with. Are you within the stepup timeframe? I'd recommend a 850 or 1kW for your setup. I've been a EVGA fanboy but this go around I've shifted away. They've had some bad products this year and it's kind of going down after they lost their head engineer

 

[ATMINSIDE]

I'd recommend a 850 or 1kW for your setup.

I'm curious why? I'd run his setup on 500W all day every day.
My main system is a 6700K OC'd with a 2080Ti on 500W... I'm certain that eats more power than a 2700X and 2070.

 

[00MJR]

I'm curious why? I'd run his setup on 500W all day every day.
My main system is a 6700K OC'd with a 2080Ti on 500W... I'm certain that eats more power than a 2700X and 2070.

I'm recovering from the malware vaccine so bare with me.. I'm kinda everywhere with my thoughts.. also people probably won't like hearing this but I'm speaking from experience, in regards to PSUs. If someone doesn't have details regarding their thoughts, it's likely their just repeating something they heard down the grapevine.

I always recommend buying a PSU that's "overkill" meaning if you system has a gpu that recommends a 500w psu, I'd go 850w.. if the gpu you're using recommends a 750w then I'd go 1kW. I try to avoid 750W PSUs cause the quality goes down a bit at this level of a PSU. I'm not talking efficiency ratings.. but it's easy to make PSUs at this level.

Components like CPUs and GPUs will always have instances of going past their rated TDP. Example, a 65w TDP intel chip will 'turbo' and 'speedstep' past the rated TDP unless you force Intel restrictions in the BIOS. Same goes for AMD. Turbo core, blah blah.

Another thing about TDP ratings... TDP is no longer a good way of measuring power requirements since every cpu and gpu boost past their rated TDPs. When a cpu says 95w TDP.. that's if the boosting, etc are all turned off. Granted everything is 'smart' and will auto adjust per workload and temp restrictions but you're still pulling past the recommended PSU needed.

Point being a 65w CPU will pull more like 100w+ when it's boosting.

This applies for GPUs too.

Then you have times the system will pull larger amounts of current for a very short time but this adds up if you're using a PSU that's underrated. Those short duration pulls will damage components overtime.

You'll see people say.. you don't want to overbuy watts cause then you're not utilizing the PSU at it's nominal efficiency rating..

That is a good point but that mindset doesn't consider the short duration spikes. You're trusting the system will pull a consistent amount at all times, and no system will do this.

Ive done some testing with a EVGA 850w g3 PSU with a system that would occasionally pull close to 700w at random times ... That psu died after a year of use. And that system had a gpu rated of gold.

Soo that's my thoughts. I'm sure I'm correct with some things and wrong on others.. but I'm basing this info off experience vs inaccurate wattage ratings recommended by companies.

Lol it's funny... The ratings for PC components are becoming the ratings off of amplifiers and speakers. You got all these wattage ratings.. rms, porgram, music, peak lol . If you aren't talking RMS.. your most likely calculating off a made up equation

 

[EarthDog]

I always recommend buying a PSU that's "overkill" meaning if you system has a gpu that recommends a 500w psu, I'd go 850w.. if the gpu you're using recommends a 750w then I'd go 1kW. I try to avoid 750W PSUs cause the quality goes down a bit at this level of a PSU. I'm not talking efficiency ratings.. but it's easy to make PSUs at this level.

Components like CPUs and GPUs will always have instances of going past their rated TDP. Example, a 65w TDP intel chip will 'turbo' and 'speedstep' past the rated TDP unless you force Intel restrictions in the BIOS. Same goes for AMD. Turbo core, blah blah.

Another thing about TDP ratings... TDP is no longer a good way of measuring power requirements since every cpu and gpu boost past their rated TDPs. When a cpu says 95w TDP.. that's if the boosting, etc are all turned off. Granted everything is 'smart' and will auto adjust per workload and temp restrictions but you're still pulling past the recommended PSU needed.

Point being a 65w CPU will pull more like 100w+ when it's boosting.

If Nvidia/Asus/MSI/Giga/etc say 850W, you can probably run it on a lot less. They already build in headroom. The only benefit of going over the recommendation is if you don't want the fan to spin up and pull less than 50% of the rated power. A quality 750W PSU will run any mainstream single CPU and GPU without issue. Even with the CPU clocked to 5 Ghz+ (be it 10900K or 11900K) and a 3090 overclocked, you aren't going to peak up there.

This applies for GPUs too.

The GPU's listed power is really close... a lot closer than a CPUs TDP to mean mean power.

If that EVGA 850W G3 died after peaking at 700W, it was faulty. It's efficiency rating has little to do with it.

For the OP, I would get another quality 750W-850W unit. Anything over that hurts the wallet more than it helps the machine.