Both the scripts in the repository and the README seem to be fully AI-generated. The README in particular seems to be a summary of Claude's thinking process. Also both the script and the README are full of assumptions related to the specific host/configuration this was used on

what is this, gpu manufacturers provide an executable to update the vbios on the 3080. I did it years ago on my ASUS 3080.

When I did it (using windows) all I had to do is

1) run the .exe

2) reboot and go to the mobo bios, enable rbar

3) boot up windows and check in nvidia control panel if rbar is on

there are instructions for doing it through linux too (although I haven't tested myself)

see:

https://nvidia.custhelp.com/app/answers/detail/a_id/5165/~/n...

Not reading that AI slop, if you need to flash the VBIOS of 3080 (or any GPU), do the following:

  1. Get a new VBIOS from a trusted source. I recommend TPU[1], they have a HUGE collection.
  2. Backup the current VBIOS (e.g. GPU-Z has a on-click option to backup and upload it to TPU)
  3. Have a working 2nd GPU[2] (or another PC with a free PCIe slot) so that you have the option the flash back the orignal VBIOS in case things go south [3]
  4. Test that your backup GPU/PC works BEFORE flashing any VBIOS, goddamit!

[1] https://www.techpowerup.com/vgabios/?architecture=&manufactu...

[2] onboard/iGPU is sufficient if present, but again: test if it works before doing anything! (just remove the dGPU and see if you get video output)

[3] Ofc you can also try to memorize the key inputs to do it blindly w/o video output from your GPU. Not recommended.

This is basically telling Claude "Make Resizable Bar work on my RTX 3080" and publishing the summary of the conversation.

For context, the RTX 3080 got ReBAR support with a VBIOS update. Some models ship with an updated VBIOS that supports it, some don't. The author's didn't.

The GPU manufacturer only provides Windows executables to update the VBIOS, the author's on Linux.

There is some valuable exploration (e.g. the mismatch in supported args in nvflash on Windows vs Linux) and just about as much information that is only useful on OP's machine (e.g. Ollama and LightDM are stopped before flashing, which is great if you happen to use either of those) or not useful at all (e.g. the word-soup-slop of diagram after diagram including "CPU Overhead Breakdown" -- what are we doing here?)

Thanks for sharing, anyway.