0

I built a PC with the following parts: Asrock B450M motherboard. AMD Ryzen 3 3100 processor (no internal GPU). Asus GT 710 graphics card (basic Nvidia one). A single HYNIX RAM 8gb stick EVGA 400W (but it doesn't matter really).

I'm trying to install OS on it. Any OS will do (Windows 10, Ubuntu, whatever).

The problem I'm having is that

  • when I boot with Windows 10 USB (I've tried 3 already) see the "choose installation method" menu for a second and then I get Blue Screen Of Death (in both Normal mode and Safe mode).
  • when I boot Ubuntu Live USB it gets to the GNOME desktop, then goes black, then goes back to GNOME but now it's frozen.

I googled it and they say it's due to installers using some basic NVidia drivers that don't work for this card.

I ran memtest86 - the RAM is 100% functional. CPU is brand new. The motherboard is brand new.

Asus Nvidia card is used, but I've seen photos of it working.

I assume I don't have the right Nvidia drivers. I'm trying now to install Ubuntu Server Edition, will say how it goes, but please let me know if I can fix it.

Hopefully, I don't have to buy a new CPU with a built-in GPU or something.

Improve this question
3
  • 2
    This has nothing to do with drivers. You have a hardware problem. Try pulling out and reseating the GPU. If that doesn't work, pull out and reseat everything else. If that doesn't work, you will need to figure out what component is failing. It might be your GPU.. but it might not. Photos of something "working" mean nothing. Even brand new stuff has problems sometimes.
    – Señor CMasMas
    Commented Apr 29, 2021 at 14:53
  • Ok, so I noticed the card is hot. I took it apart and saw that the thermal paste is hard as a rock. I replaced the thermal paste with fresh paste. It seemed this could help (as it froze sometimes sooner sometimes later into the boot process, looked like a heating problem). But it didn't solve it... I tried to attach an active fan to it to actively cool it. Still Blue Screen. I'm stuck. I do see UEFI and can't see normal OS. The heating was a perfect fit. Now my choice is to either buy a PC to try it or buy a second card and risk that it will still not solve the problem. Options sound silly.
    – Serhyi Vynohradov
    Commented Apr 30, 2021 at 13:06
  • Is there at least a console suite that can run some NVidia tests or something? (for Ubuntu etc.)
    – Serhyi Vynohradov
    Commented Apr 30, 2021 at 13:09

2 Answers 2

Reset to default
1

To answer your question, yes, it's possible to use a discrete GPU for almost any operating system. For Linux, it's even possible to install with no GPU at all and use a serial console and network access.

However, if you continue to see crashes and hangs regardless of operating system, you probably have some sort of hardware problem. Check that everything is seated correctly, including your GPU and memory (pull it out and reinstall it) and all the power connectors are fully inserted and that the power supply is adequate (and you could try swapping with a different power supply as well to see if that's broken). You could try the GPU in a different machine as well to see if that's the problem.

Just because the component is brand new doesn't mean it isn't defective; for example, I've bought brand new memory which was very clearly and obviously broken from the moment I put it into the machine. Other failures are more subtle, though. The phenomenon of early failures with hardware is sometimes referred to as “infant mortality.”

Ultimately, you'll need to figure out what component is sad and replace it.

Improve this answer
2
  • Thanks. I would love to follow a "happy path" first. (assume that everything is functional). I will try to install NVidia driver on a different machine and swap an SSD with a pre-installed Operating System. Thanks for the help. Power Supply is adequate (EVGA, has 350W on a 12V rail, my CPU should be not more than 65W, that graphics card is like 20W, + SSD & RAM). 400W should be plenty. I think I do have a different PC to try the video card though, good idea, thanks.
    – Serhyi Vynohradov
    Commented Apr 30, 2021 at 9:09
  • Very useful answer. Guided me towards the real problem.
    – Serhyi Vynohradov
    Commented May 4, 2021 at 13:45
0

Okay, so if anyone ever has the same problem:

I ran Ubuntu Server Installation (which is good because it has a verbose log of errors during boot) and I received the following error:

invalid opcode 0000 #1 smp nopti

This means "null was submitted unexpectedly to some Kernel function".

This means RAM is losing data.

This means RAM is glitchy.

I went to the store and asked for a new RAM, and it worked.

Now my PC is working.

Huge thanks to everyone who responded.

You were right guys, it was a hardware problem.

But it wasn't NVidia. It was RAM.

Moral of the story:

  • If anyone ever has such a problem, use Ubuntu Server Installer ISO to see what hardware errors you are getting.
  • Don't buy used PC parts online from shady immigrants without making sure they work.

Cheers!

Improve this answer
2
  • It was surprising though how RAM passed many tests of Memtest86+ and still was glitchy when it tried to boot OS.
    – Serhyi Vynohradov
    Commented May 4, 2021 at 13:46
  • I'm renaming this question so that people could find the solution
    – Serhyi Vynohradov
    Commented May 4, 2021 at 13:48

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .