Windows Alpha for AMD & Nvidia GTX GPUs only

Topaz Video AI Video AI Beta
1

Hi All,

Here is an early version of app that should improve performance for AMD GPUs and hopefully older Nvidia GPUs(non RTX cards) that use onnx models.

Download: Win

Known Issues:

  • Proteus Auto will not work, use Proteus with manual settings
  • Models cannot be run on CPU
  • Models cannot be run on intel GPU
  • Models run slower on RTX GPUs

Thank you for testing

PS: Since this is an alpha version, it will not interfere with Beta or Release versions on the machine.

7 Likes
2

Thanks for this!

Is there any optimal value we should be setting the memory usage to? Or just leave it at 100%?

Which leads into my thought that it would be ideal if the program calculated and set the memory use sizes optimally on it’s own without user intervention (or even visibility to it).

3

100% in most cases should be fine unless running too many filters.
Eventually yes, it is lot more complicated due to the GUI allowing multiple simultaneous processes.

4

Notice: Lowering Memory to 50% did speed up the process by 70% on my system.

So 16 - 17 fps instead of 10.

look at your models folder, it should use the 288px models.

2 Likes
5

Proteus Manual (No setting moved)
tested on 1050ti GTX, 720p → 1080p 59.97 = usual speed, 0.79 seconds per frame. around 0.78 with the last beta.

6

I tested this version vs 3.0.15 release and didn’t see any notable difference in performance on GTX 1070, everything was with about 2% of each other between versions. Perhaps it made a bigger difference to AMD GPUs.

1 Like
7

Did you try to lower memory in settings below 50%?

8

Yes, and the performance with memory set lower was worse in every scenario I tested:

GTX1070 8GB + Intel 6700K + 16 GB RAM + Windows 11

Alpha v3.1.6.0.a+amd

720p source, 2x upscale to 1440p, max mem 45%:

Artemis HQ: 0.31spf
Gaia HQ: 0.88spf
Proteus: 0.31spf

720p source, 2x upscale to 1440p, max mem 100%:

Artemis HQ: 0.29spf
Gaia HQ: 0.82spf
Proteus: 0.29spf

1080p source, no upscale, max mem 45%:

Artemis HQ: 0.53spf
Gaia HQ: 1.39spf
Proteus: 0.53spf

1080p source, no upscale, max mem 100%:

Artemis HQ: 0.45spf
Gaia HQ: 1.19spf
Proteus: 0.46spf

Chronos Fast:

720p source, Chronos Fast 2x frame double: 0.17spf max mem 45%
720p source, Chronos Fast 2x frame double: 0.15spf max mem 100%

1080p source, Chronos Fast 2x frame double: 0.57spf max mem 45%
1080p source, Chronos Fast 2x frame double: 0.44spf max mem 100%

1 Like
9

There is an updated version posted here

2 Likes