PC Diagnostic software recommendations?

[aaouviz]

My PC is causing me trouble. It's almost always had problems but I've just dealt with it. I'm on the verge of buying a new one (it's time anyway) but right now I just don't want that worry.

I'm wondering if anyone has good software recommendations to help pinpoint what the issue might be? I'm often getting the blue-screen crash, a memory load dump, if I recall correctly.

It's generally gotten pretty bad over the course of approx. 2 months to the point of being unusable at which point I replace the windows SSD and things seem to work all fine again. But this time it hasn't helped. I'm now thinking it might be my graphics card. I've replaced the RAM. I thought the RAM was a real problem as the PC could never handle 128gb of RAM, so I've always only had 64gb installed with another 64gb sitting on my bookshelf collecting dust.

It generally happens under moderate-high load, but sometimes not.

My full system:
CPU: AMD Ryzen Threadripper 2990WX
GP: NVIDIA GeForce RTX 2070 SUPER 8 GB
MB: Asus Prime x399-a
RAM: 4 x 16gb DDR4
SSD: Samsung 850 Evo 500gb

Any tips in pinpointing the problem appreciated!

[Juraj]

I am not aware of anything myself, but few things can be done:

1) Event Viewer (just type that into Windows Start menu). Particularly Windows Logs/System and then red-icon Kernel issues. But it is bit of like finding needle in hay since even the events themselves will simply tell you that certain driver caused crash for example, but nothing beyond that. It can sometimes provides good hint, esp. if it's driver related. And then you can do a bit of googling of that.

2) Hardware stability the biggest one being MemTest, tests stability of your memory. That was always massive Achilles heel of 2990WX. Can be run for hours ideally, overnight. Not much to do if it tells you that memory is corrupted. 2990WX is capable of running 128GB, even on low-end motherboards like Prime, even with low-end memory modules, but it requires very conservative setup. Like 2667 MT/s for example, instead of 2933/3200. I don't remember the tinkering fondly.

I have plenty of blue-screens with even the latest&greatest hardware. Sometimes it's random stuff like latest Windows 11 update no longer being friendly with ancient drivers for Realtek audio chip, or Intel LAN/Wi-Fi/Bluetooth driver, etc.. etc.. (Windows installs those drivers automatically, but those drivers are often dinosaur era)
And solution to that is endless uninstall/reinstall of random drivers.

[aaouviz]

Thanks Juraj, I thought you might be the right guy to answer :)

Yep, event viewer seems to have told me it was a Kernel Power error. Not much info other than that.

Welp, this problem has plagued me 3 years now. I might have to look into those endless "Help on new PC" threads I often ignore here on the Corona forums.

[balatschaka]

My 2990wx was also driving me nuts. It would sometimes randomly go black screen with fans roaring at 100%, resulting in me restarting manually..
The solution was to get a new machine :)

[Juraj]

I didn't want to say that, but I also thought that 2990WX is best relegated to history, it just wasn't good chip with its unique half-dies-only-mem-access architecture gimmick.

This is of course fully budget related, I also try to abide today with "if it works, keep it" instead of wasting money&time on latest&greatest stuff unless it's given to me for free.
But 2990WX was just mistake in retrospect.

[aaouviz]

I was actually ignorant to any 2990WX common issues. I guess this is the likely cause of my persistent problems then?

I wonder what I can upgrade to if I want to just replace the CPU (and maybe MB?)?

[Juraj]

Nothing, X399 was dead-end platform, so no further options. And so was TRX40. AMD didn't do good job with keeping continuity sadly.

Any upgrade you might want, you will need to jump to new platform (CPU+Motherboard+DDR5).

2990WX was mostly very inconsistent with performance, it was heavily memory latency dependent (only half the cores can access memory, the other half have to go through that first half first).
It would work great under one situation, then badly under other..

Nowadays, that's no longer the case with any option on market. Even mainstream non-workstation platforms, like AMD Ryzen 9 7950X demolish the 2nd gen Threadrippers with 2x single-speed core, super consistent performance in every situation and can even support 192GB DDR5 memory, though it's bit tricky to get it work as well.