ASUS TUF Gaming GeForce RTX 5070 Ti 📟 🚧

Quelle RTX 5070 Ti choisir ?

ASUS TUF Gaming GeForce RTX 5070 Ti OC Edition - 980€ / 01/2026 - before price explosion

  • 16GB GDDR7 OC Edition
  • PCIe 5.0, HDMI, Display 2.1, 3.125 Slot, Ventilateurs axiaux, revêtement Protecteur PCB
  • 329x140x62.5mm - takes 3 PC slots
  • 3dmodels

Motherboard A520I

Settings

IO Ports/ PCIe-3.0

  • Above 4G Decoding: Enabled
  • Re-Size BAR Support: Auto

MultiGpu
AMD CBS/NBIO Common Options/GFS Configuration -iGPU Configuration: Auto -GPU Host Translation Cache: Auto

IO Ports/ Initial Display Output: IGD Video Integrated Graphic: Auto

Testing that it works

$ lspci | grep -E "VGA|3D"
01:00.0 VGA compatible controller: NVIDIA Corporation Device 2c05 (rev a1)
07:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Cezanne [Radeon Vega Series / Radeon Vega Mobile Series] (rev c8)
  • Both GPU are seen
  • you can plug the hdmi cable in whateber socket => it will boot on it

OpenGL

Check iGPU is PRIMARY (Xorg / Wayland)

  • opengl works on iGPU (motherboard HDMI)
  • does not work on RTX (cinnamon wm failing) - RESOLVED
$ glxinfo | grep "OpenGL renderer"
=> AMD Radeon Graphics
$ xrandr --listproviders
- AMD provider (Sink, Source)
- NVIDIA provider (Source)

But…

PRIME offloading

$ __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxinfo | grep renderer
=> Fail the same way as if primary (X_GLXCreateNewContext)
=> diagnosed it fail because of lack of privilege
=> sudo __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxinfo | grep rendered 
=> WORKS!!

see Permissions for resolution

Linux Setup

The NVIDIA RTX 5070 is part of the 5000-series, which requires the Open Kernel Modules and requires driver 580 or newer. - Working 580-open Driver Guide

  • linuxmint 22.3 / ubuntu 24.04 LTS
  • kernel: 6.14.0-37-generic
  • nvidia driver
    • 570.211.01 - smi working
    • 590.48.01 (recommended) - smi work the same - CUDA Verson: 13.1

Removing previous driver:

$ sudo apt-get remove --purge '^nvidia-.*'
$ sudo apt autoremove
$ sudo reboot

Permissions

On a standard Linux system with proprietary NVIDIA drivers:

  • nvidia-smi should work as a regular user
  • glxinfo should work as a regular user

Neither command should require sudo

If encountering the permission issue check groups for:

$ ls -l /dev/nvidia*

They should be either

  • video
  • or vglusers (which is typical when VirtualGL or remote GPU sharing / visualization software is installed.)

In my case it was vglusers. I fixed it by adding user to this group

$ sudo usermod -aG vglusers $USER
# then logged out / and relog

It is certainly linked to my xpra install

CUDA

CUDA works as long as nvidia-smi sees the GPU and the driver is installed. OpenGL or graphics output isn’t required for CUDA compute tasks.

Testing CUDA

Check that CUDA toolkit is installed

$ nvcc --version

# if above is missing
$ sudo apt install nvidia-cuda-toolkit

Install CUDA Samples

$ git clone https://github.com/NVIDIA/cuda-samples.git
$ mkdir build && cd build
$ cmake ..
$ make deviceQuery

# test
$ ./Sample/1_Utilities/deviceQuery/deviceQuery
# If this require sudo (to fix nodevice found) there is a permission issue

see

Pytorch 🚧

Written on January 24, 2026, Last update on
pc-hardware gpu