Looks like the 5080 is slower than a 4080 in Blender and TVAI?

Topaz Video AI General
1

According to this Tech Powerup article the 5080 is slower than a 4080, not in just TVAI but also Blender.

2

As the article says, apps that use Tensor RT, like TVAI, will need an update.

1 Like
3

Read the article… It has also been confirmed by Topaz here on the forums. It’s not compatible with the new architecture yet.

1 Like
4

The next update should support the new 5000 series

5

Edit:
Can confirm that the 5090 is at the moment up to 40% slower with most models than the 4090 (I own both) but 30% faster in Gaia
see benches below

6

I guess quite some people here are waiting for a benchmark…

7

maybe we can expect a big improvement with native FP4 support? (for 5000 series)

8

here u go. i only have v4.1 installed (don’t like v5 or v6 and also don’t have updated license)

Topaz Video AI  v4.1.0
System Information
OS: Windows v10.22
CPU: AMD Ryzen 9 7950X 16-Core Processor              31.71 GB
GPU: NVIDIA GeForce RTX 5090  31.349 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	32.80 fps 	2X: 	13.79 fps 	4X: 	04.07 fps 	
Iris		1X: 	37.00 fps 	2X: 	17.28 fps 	4X: 	04.94 fps 	
Proteus		1X: 	40.67 fps 	2X: 	17.87 fps 	4X: 	04.81 fps 	
Gaia		1X: 	21.58 fps 	2X: 	13.32 fps 	4X: 	04.20 fps 	
Nyx		1X: 	17.56 fps 	2X: 	12.40 fps 	
Nyx Fast		1X: 	32.71 fps 	
4X Slowmo		Apollo: 	37.57 fps 	APFast: 	68.85 fps 	Chronos: 	21.01 fps 	CHFast: 	38.34 fps

Here is my 4090 (TVAI v5 , v4 and v6 are very much the same performance):

Topaz Video AI Alpha  v5.3.0.1.a.hyp.rxl
System Information
OS: Windows v10.22
CPU: AMD Ryzen 9 7950X 16-Core Processor              31.118 GB
GPU: NVIDIA GeForce RTX 4090  23.576 GB
GPU: AMD Radeon(TM) Graphics  0.47445 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	44.05 fps 	2X: 	18.96 fps 	4X: 	05.19 fps 	
Iris		1X: 	42.82 fps 	2X: 	22.31 fps 	4X: 	05.99 fps 	
Proteus		1X: 	41.98 fps 	2X: 	22.13 fps 	4X: 	06.17 fps 	
Gaia		1X: 	15.66 fps 	2X: 	10.96 fps 	4X: 	05.60 fps 	
Nyx		1X: 	17.64 fps 	2X: 	15.97 fps 	
Nyx Fast		1X: 	31.61 fps 	
Rhea		4X: 	05.50 fps 	
4X Slowmo		Apollo: 	44.50 fps 	APFast: 	91.33 fps 	Chronos: 	33.24 fps 	CHFast: 	39.70 fps 	
16X Slowmo		Aion: 	37.31 fps

Performance is actually up to 40% worse it seems (whenever scaling is involved, CPU + RAM will play a relatively larger role reducing the difference between GPU generations)
Gaia 1x is actually 30% faster (maybe cause it’s an older model and doesn’t use newer tensor cores but the more basic GPU pixel functions where the behaviour hasn’t changed and profits from much faster VRAM of the 5090)
Nyx seems to be on par

1 Like
9

That’s underwhelming. I think Gaia is faster because it uses CUDA still, but I am not exactly sure.

10

At least it works. I really don’t want to update TOPAZ, the new UI is just atrociously awful.

11

Yep, getting roughly 30% slower performance on my 5090 when using Proteus.

12

Not to worry. Have you tried rendering anything with the 5090? Did you notice something?

I’ve done some rendering on my 5080, and on certain models, it’s not able to use more than 60% of its max power. Which points towards that most likely the tensor cores aren’t being used since they haven’t implemented blackwell yet. Once the tensor cores gets going we’ll see some proper gains.

1 Like