Crashes relating to the GPU removed error D3D. Completely removed GPU from PC

I am continously experiencing crashes relating to the GPU removed error D3D. When i start up an empty project, both my monitors go to black and receive no further input from my RTX 3090. I have switched to my internal graphics (AMD Radeon) to get an better idea of what the problem was, and it seems that the crashes completely remove my RTX GPU from the device manager, similar as to if its not even physically attached. I Have to restart to (usually) re-enable my GPU.

I have scoured the internet for about 3 days and have tried a number of things:

  • Complete reinstall Windows, Epic Games Launcher and all engines
  • Edited RegistryEditor with TDR Delay values
  • Updated drivers to the latest
  • Reverted drivers to earlier version
  • Tried using the Studio Drivers
  • Removed NVIDIA App, used Geforce Experience instead
  • Reseated the GPU in its slot and checked the PSU cables on both sides
  • Moved the GPU to anothe PCIE slot in the motherboard
    Edit:
  • tried installing Windows 10

All components besides the GPU and PSU are completely new. Everything worked fine for about a month and a half, until this suddenly started happening. It also happens when i play games made in Unreal Engine. I tested this with the Steam game Hydroneer, which was made in UE, and with Cities Skylines, which was made in Unity. Hydroneer froze on startup, leading to another crash related to the GPU multiple times, while Cities Skylines worked just fine.

The crashes happen indepently of which engine version i am starting up. It seems to just be anything UE-related that causes issues.

My specs:

AMD Ryzen 7 7800X3D (1,5 months old)
NZXT B550E motherboard (1,5 months old)
Inno3D RTX 3090 (3 years old)
64GB DDR5 RAM (1,5 months old)
1050W PSU( 3 years old)

Anyone experiencing anything similar and maybe has a fix? UE is literally unusable for me right now.

1 Like

How are your graphics cards memory temps while playing? The 3090 series had the architecture of sandwiching the dies on both sides of the cads due to not having high enough chip capacity per die.

So as a solution they put some of the vram on the back. This would cause them to heat up in comparison to the ones under the heat sink. I would check it with HWiNFO or GPU-Z if the crash isn’t hardware related.

The comparison between an unreal game and cities skylines isn’t exactly a 1:1 test.

Unity in CS doesn’t really push the pc hardware that much so it’s not taxing the GPU. You may be getting problems with unreal because it’s using your GPU to it’s maximum => then the problems seem to arise.

Try putting on another game that is more hardware intense if you have it in your library.

You could try a 3d benchmark too. (though I’d probably not try furmark).

Try to narrow down if it’s really unreal at fault

2 Likes

Hi there, Thanks for the reply. I had the same ideas exactly.

I have used HWinfo to generate a log from when it happens:

I have attached the entire log here:
log.CSV (417.6 KB)

Although to me it doesnt seem like there is an hardware issue:

The load is not high, the power draw is not huge, the VRAM is not exhausted.
I also kept an eye on the temps before the crash, it did not go beyond 70C.

Maybe try this solution. Similar hardware and situation

I would try this, but my GPU has given up completely since yesterday. Unfortunately I think I’m gonna have to replace either the GPU or psu

Tough time on the market to get something usable, unless it’s slightly lower end hardware.

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.