Desktop and app performance regression with 5070 TI on Fedora 41 vs. AMD 6650 XT

UPDATE: the new stable prod driver release 570.124.04 fixed all the main issues (1-3) listed below for me. The generic card name output (issue 4-6) remains, but is non-critical for me. nvidia-smi now lists the card just fine:

+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.124.04             Driver Version: 570.124.04     CUDA Version: 12.8     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 5070 Ti     Off |   00000000:0A:00.0  On |                  N/A |
|  0%   40C    P8             25W /  300W |    2147MiB /  16303MiB |      1%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

Initial post below, for future reference:


I’m aware of how fresh this card is, however, the current beta driver 570.86.16 doesn’t seem to really support this card despite its claims. I’m encountering the following issues after switching to the 5070 TI from an AMD 6650 XT:

  1. JetBrains IDEs (Goland, PyCharm, IntelliJ) became unusable, selecting text, executing runs, just using the UI became unbearably slow, with any action in the UI mentioned above cranking up GPU utilization to 100 %. VS code does not exhibit this slowdown to make it unusable, but even just typing in the editor gets the GPU utilization to 60 %. Typing text in Chrome (editing this post) even gets the GPU up to 86 %…).
  2. Games are unplayable, Doom Eternal and Helldivers 2 ~ 15 fps @ WQHD, Unigine Heaven benchmark @ ~ 27 fps, some games crash outright when trying to start (Total War: Warhammer II)
  3. nvidia-smi output, the GPU name isn’t even pretending to be referring to the 5070 TI:
| NVIDIA-SMI 570.86.16              Driver Version: 570.86.16      CUDA Version: 12.8     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA Graphics Device         Off |   00000000:0A:00.0  On |                  N/A |
|  0%   42C    P8             26W /  300W |    1190MiB /  16303MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
  1. fastfetch system details, also same generic GPU name reference:
OS: Fedora Linux 41 (KDE Plasma) x86_64
Kernel: Linux 6.13.4-200.fc41.x86_64
Display (DELL G2724D): 2560x1440 @ 165 Hz in 27" [External] *
Display (DELL G2724D): 1440x2560 @ 165 Hz in 27" [External]
DE: KDE Plasma 6.3.1
WM: KWin (Wayland)
CPU: AMD Ryzen 9 5950X (32) @ 5.08 GHz
GPU: NVIDIA Device 2C05 (VGA compatible)
Memory: 9.75 GiB / 125.69 GiB (8%)
Swap: 0 B / 8.00 GiB (0%)
  1. lspci output following suit:
❯ lspci | grep VGA
0a:00.0 VGA compatible controller: NVIDIA Corporation Device 2c05 (rev a1)
  1. lsmod output (nouveau is not loaded):
❯ lsmod | grep nvidia
nvidia_uvm           4145152  0
nvidia_drm            147456  181
nvidia_modeset       2138112  83 nvidia_drm
nvidia              13074432  1526 nvidia_uvm,nvidia_modeset
drm_ttm_helper         16384  1 nvidia_drm
video                  81920  2 asus_wmi,nvidia_modeset

I have not yet tried any CUDA workloads on it, which was the initial purchase argument for this card…

Is there an ETA for a driver version that’s supposed to truly support this card?

I’ve already mentioned this issue and tried to ask for ETA there and in support requests, to no avail. Something has gone wrong and/or is under NDA.

That is terrible news :(! Thanks for sharing nonetheless.