GPU appears multiple times in the UI

Topaz Video AI Bugs and Issues
1

In TVAI my GPU (an RTX 4090) appears twice times in the UI.

When using the benchmark tool, the results list my hardware as my CPU + RTX 4090 + RTX 4090. When it should only be one RTX 4090.

In the preferences I can select between Auto, CPU, RTX 4090, RTX 4090 All GPUs. Once again, there should only be 1 4090.

My OS is Windows 11 (Version 22H2 22621.1413)
My GPU drivers are 531.41

Although, I should note this issue has been happening for me since the release of TVAI 3.0.0, has persisted through multi TVAI updates (I’m currently on 3.2.1), Windows updates, GPU driver updates, a GPU upgrade (RTX 3060 to a RTX 4090), and a factory reset of Windows.

My hypothesis is that having the CUDA toolkit installed may be triggering this. But this is just speculation.

Here are my logs from when I open TVAI.
2023-04-14-22-51-42-Main.tzlog (22.6 KB)

Images:

Preferences menu

Benchmark screen
Benchmark screen900Ă—468 54 KB

3

Hi Lee,

I’ve confirmed with our developers that this can happen with a few different GPUs including the 4090 and certain Ti models of other GPUs. It can be caused by multiple monitors being connected or multiple versions of DirectX being installed at the same time. This is purely a visual bug that will not impact performance, so selecting either of the GPUs listed or selecting “All GPUs” will result in the same performance. We plan to fix this in an update soon.

Thanks and let me know if there’s anything else I can help with,

6

Just wanted to let you know that it’s not purely cosmetic. If you select All GPUs in this scenario, the performance is actually lower.

All GPUs:

Topaz Video AI v3.2.6
System Information
OS: Windows v11.2009
CPU: Intel(R) Xeon(R) Gold 6226R CPU @ 2.90GHz 94.654 GB
GPU: NVIDIA Quadro RTX 4000 7.8213 GB
GPU: NVIDIA Quadro RTX 4000 7.8213 GB
Processing Settings
device: 2 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 03.16 fps 2X: 02.03 fps 4X: 00.63 fps
Proteus 1X: 02.27 fps 2X: 01.76 fps 4X: 00.61 fps
Gaia 1X: 01.93 fps 2X: 01.25 fps 4X: 00.15 fps
4X Slowmo Apollo: 06.19 fps APFast: 13.31 fps Chronos: 02.55 fps CHFast: 03.68 fps

Selected single GPU:

Topaz Video AI v3.2.6
System Information
OS: Windows v11.2009
CPU: Intel(R) Xeon(R) Gold 6226R CPU @ 2.90GHz 94.654 GB
GPU: NVIDIA Quadro RTX 4000 7.8213 GB
GPU: NVIDIA Quadro RTX 4000 7.8213 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 06.88 fps 2X: 04.03 fps 4X: 01.42 fps
Proteus 1X: 04.79 fps 2X: 03.10 fps 4X: 01.29 fps
Gaia 1X: 02.76 fps 2X: 01.84 fps 4X: 01.34 fps
4X Slowmo Apollo: 03.59 fps APFast: 20.70 fps Chronos: 06.27 fps CHFast: 07.60 fps

7

ALL GPUs mode uses ORT+DML this is slower than the TensorRT option used when selecting the single GPU.

Depending on the monitors connected, certain GPUs might show up twice, selecting either one will work and is recommended.

If you have two different GPUs then select All GPU otherwise stick with single GPU selection or AUTO.

8

Thanks. What about two or more identical GPUs?

9

If you have 2 or more identical GPUs then use ALL GPU

3 Likes
10

Still not fixed! And it’s now a full year later. Please fix this. Using with 2x2080Ti, and showing up with 4 devices. If I select All GPU, then it crashes saying “Can’t find backend” in the crash logs, which I presume means it can’t find the pretend GPUs. Please fix as it serious degrades the quality of the (purchased) product for me.