Video AI v3.3.X - User Benchmarking Results

v.3.3.9 max. GPU power use was 437 W.

Topaz Video AI  v3.3.9
System Information
OS: Windows v11.22
CPU: 13th Gen Intel(R) Core(TM) i9-13900K  31.775 GB
GPU: NVIDIA GeForce RTX 4090  23.59 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	39.36 fps 	2X: 	16.35 fps 	4X: 	04.45 fps 	
Proteus		1X: 	32.13 fps 	2X: 	15.65 fps 	4X: 	04.43 fps 	
Gaia		1X: 	15.04 fps 	2X: 	10.43 fps 	4X: 	05.04 fps 	
4X Slowmo		Apollo: 	42.18 fps 	APFast: 	83.92 fps 	Chronos: 	32.04 fps 	CHFast: 	33.67 fps 	

Your CPU generally outperforms my 7950x due to a better memory controller :cry:

I bought into the AM5 platform because the 13900K socket is at the end of the line so I will have at least one more generation of CPU upgrades…hopefully 2

2 Likes
Topaz Video AI  v3.3.5
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 5900X 12-Core Processor              31.912 GB
GPU: NVIDIA GeForce RTX 3070  7.8428 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	14.20 fps 	2X: 	07.24 fps 	4X: 	01.99 fps 	
Proteus		1X: 	13.11 fps 	2X: 	06.98 fps 	4X: 	01.92 fps 	
Gaia		1X: 	04.84 fps 	2X: 	03.32 fps 	4X: 	01.97 fps 	
4X Slowmo		Apollo: 	18.00 fps 	APFast: 	38.80 fps 	Chronos: 	10.90 fps 	CHFast: 	17.36 fps 	

Personal computer (Intel + Nvidia):

Topaz Video AI  v3.3.9
System Information
OS: Windows v11.22
CPU: 13th Gen Intel(R) Core(TM) i9-13900K  63.576 GB
GPU: NVIDIA GeForce RTX 4080  15.688 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	29.75 fps 	2X: 	15.92 fps 	4X: 	04.36 fps 	
Proteus		1X: 	26.41 fps 	2X: 	14.71 fps 	4X: 	04.18 fps 	
Gaia		1X: 	10.67 fps 	2X: 	07.28 fps 	4X: 	04.60 fps 	
4X Slowmo		Apollo: 	36.16 fps 	APFast: 	73.63 fps 	Chronos: 	23.63 fps 	CHFast: 	29.93 fps 	

Work computer (AMD + AMD):

Topaz Video AI  v3.3.9
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 7950X3D 16-Core Processor            127.15 GB
GPU: AMD Radeon PRO W7900  44.942 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	16.07 fps 	2X: 	09.26 fps 	4X: 	02.55 fps 	
Proteus		1X: 	15.21 fps 	2X: 	08.60 fps 	4X: 	02.58 fps 	
Gaia		1X: 	07.25 fps 	2X: 	04.73 fps 	4X: 	02.83 fps 	
4X Slowmo		Apollo: 	18.22 fps 	APFast: 	54.99 fps 	Chronos: 	10.76 fps 	CHFast: 	16.14 fps 	

No secret that TVAI is highly optimized for Nvidia, but kind of striking to see the disparity.

One thing to note: the all AMD machine, while not as fast, handles work much more elegantly. Barely even breaks a sweat, whereas the Intel + Nvidia machine sounds like a jet trying to takeoff…

Similar to Apple silicon. Not really as fast as Nvidia, but MUCH less power consumption leading to much less noise.

Seems to be a micro performance gain when you reduce the Vram from 1.0 to .8. Lower numbers tend to reverse the gain.

Mac Mini M1 - 512GB SSD - OS is Ventura 13.5 (not 13.05 as listed)

Topaz Video AI  v3.3.8
System Information
OS: Mac v13.05
CPU: Apple M1  16 GB
GPU: Apple M1  10.667 GB
Processing Settings
device: 0 vram: 0.8 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	03.71 fps 	2X: 	02.48 fps 	4X: 	01.02 fps 	
Proteus		1X: 	03.73 fps 	2X: 	02.48 fps 	4X: 	00.98 fps 	
Gaia		1X: 	01.22 fps 	2X: 	00.90 fps 	4X: 	00.71 fps 	
4X Slowmo		Apollo: 	03.87 fps 	APFast: 	14.53 fps 	Chronos: 	01.31 fps 	CHFast: 	01.98 fps 	

For HD upscales this is mostly correct (depending on the HW config 100% can still be the fastest).

But for SD upscales setting RAM to 10% still leads to a massive performance gain.

Interesting - I have not seen this gain. I mainly deal with 480p video and see no changes when upscaling to 720p when the RAM settings is changed to 10% or 100%.

Perhaps this is only the case when upscaling to 1080p?

I guess the speed gain by reducing RAM usage heavily depends on how many GPU cores you have (someone told that setting RAM to 10% forces the app to use the GPU while at 100% it uses the Neural Engine instead).

So, of course your mileage will vary on the “bare” M1 compared to the M2 ultra I have here (with 60 GPU cores).

Anyway this is what I’m getting here upscaling a 524x360 video to 1280x720:

The upper preview was done with RAM set to 10% and the middle and lower one with 80% and 100%, respectively.
So, with many GPU cores you get a massive boost by setting RAM to 10% for anything lower than 1080p input resolution and up to 2x upscales.

Topaz Video AI  v3.3.9
System Information
OS: Windows v10.22
CPU: AMD Ryzen Threadripper 3970X 32-Core Processor   127.87 GB
GPU: NVIDIA GeForce RTX 3090  23.77 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	24.37 fps 	2X: 	10.04 fps 	4X: 	02.79 fps 	
Proteus		1X: 	22.37 fps 	2X: 	09.58 fps 	4X: 	02.74 fps 	
Gaia		1X: 	08.50 fps 	2X: 	05.76 fps 	4X: 	02.68 fps 	
4X Slowmo		Apollo: 	30.39 fps 	APFast: 	47.54 fps 	Chronos: 	18.89 fps 	CHFast: 	24.31 fps 	

So I’m guessing that you’re getting a bit better results from windows 11?

I only have been using 11. Not a fan of 11 but I built the new system with a 4090 and figured it’s optimized better than 10 which I am most likely wrong cause 10 is very stable. I haven’t had any issues with 11 so I don’t have benchmarks from 10. I did debloat 11 so only the basics and maybe that’s why it hasn’t crashed yet. How about you?

Topaz Video AI  v3.3.9
System Information
OS: Windows v11.22
CPU: 13th Gen Intel(R) Core(TM) i9-13900K  127.75 GB
GPU: NVIDIA GeForce RTX 4090  23.59 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	40.23 fps 	2X: 	16.39 fps 	4X: 	04.30 fps 	
Proteus		1X: 	32.26 fps 	2X: 	15.39 fps 	4X: 	04.30 fps 	
Gaia		1X: 	15.85 fps 	2X: 	10.92 fps 	4X: 	04.71 fps 	
4X Slowmo		Apollo: 	41.92 fps 	APFast: 	73.20 fps 	Chronos: 	32.45 fps 	CHFast: 	34.61 fps 	

Topaz Video AI  v3.3.9
System Information
OS: Windows v11.22
CPU: 13th Gen Intel(R) Core(TM) i9-13900K  127.75 GB
GPU: NVIDIA GeForce RTX 4090  23.59 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 720x480
Benchmark Results
Artemis		1X: 	199.20 fps 	2X: 	79.91 fps 	4X: 	24.50 fps 	
Proteus		1X: 	151.91 fps 	2X: 	69.15 fps 	4X: 	22.52 fps 	
Gaia		1X: 	84.64 fps 	2X: 	57.48 fps 	4X: 	30.38 fps 	
4X Slowmo		Apollo: 	216.67 fps 	APFast: 	450.87 fps 	Chronos: 	171.72 fps 	CHFast: 	207.38 fps 	

Has anyone experimented with turning on/off the CPU core parking feature? Dependend on the personal use of the machine users report either better fps in games or benchmarks by turning off the core parking while other applications may benefit from the core parking turned on. Here are my results from the latest Alpha version with the CPU core parking turned on:

Topaz Video AI Alpha  v3.4.0.1.a
System Information
OS: Windows v11.22
CPU: 13th Gen Intel(R) Core(TM) i9-13900K  31.775 GB
GPU: NVIDIA GeForce RTX 4090  23.59 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	37.93 fps 	2X: 	19.46 fps 	4X: 	05.27 fps 	
Proteus		1X: 	35.37 fps 	2X: 	16.33 fps 	4X: 	04.91 fps 	
Gaia		1X: 	15.18 fps 	2X: 	10.51 fps 	4X: 	05.07 fps 	
4X Slowmo		Apollo: 	39.92 fps 	APFast: 	96.31 fps 	Chronos: 	33.24 fps 	CHFast: 	34.00 fps 	

Topaz Video AI  v3.3.9
System Information
OS: Windows v10.22
CPU: 13th Gen Intel(R) Core(TM) i7-13700K  31.776 GB
GPU: NVIDIA GeForce RTX 4090  23.59 GB
GPU: Intel(R) UHD Graphics 770  0.125 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	49.76 fps 	2X: 	19.62 fps 	4X: 	05.93 fps 	
Proteus		1X: 	40.87 fps 	2X: 	19.16 fps 	4X: 	05.03 fps 	
Gaia		1X: 	16.16 fps 	2X: 	11.72 fps 	4X: 	05.14 fps 	
4X Slowmo		Apollo: 	43.25 fps 	APFast: 	92.77 fps 	Chronos: 	34.20 fps 	CHFast: 	26.64 fps 	

Topaz Video AI  v3.3.9
System Information
OS: Windows v10.22
CPU: 13th Gen Intel(R) Core(TM) i7-13700K  31.776 GB
GPU: NVIDIA GeForce RTX 4090  23.59 GB
GPU: Intel(R) UHD Graphics 770  0.125 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1280x720
Benchmark Results
Artemis		1X: 	100.56 fps 	2X: 	42.20 fps 	4X: 	11.72 fps 	
Proteus		1X: 	82.95 fps 	2X: 	39.45 fps 	4X: 	11.95 fps 	
Gaia		1X: 	36.05 fps 	2X: 	26.35 fps 	4X: 	11.78 fps 	
4X Slowmo		Apollo: 	108.90 fps 	APFast: 	195.79 fps 	Chronos: 	73.28 fps 	CHFast: 	97.34 fps 	

1 Like
Topaz Video AI  v3.3.10
System Information
OS: Windows v11.22
CPU: 13th Gen Intel(R) Core(TM) i9-13900K  31.775 GB
GPU: NVIDIA GeForce RTX 4090  23.59 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	37.71 fps 	2X: 	16.12 fps 	4X: 	04.33 fps 	
Proteus		1X: 	32.04 fps 	2X: 	15.79 fps 	4X: 	04.60 fps 	
Gaia		1X: 	15.31 fps 	2X: 	10.42 fps 	4X: 	05.00 fps 	
4X Slowmo		Apollo: 	41.94 fps 	APFast: 	71.85 fps 	Chronos: 	32.22 fps 	CHFast: 	33.49 fps 	

We need a new benchmarking cathegory for versions 3.4.x.x.

Topaz Video AI Beta  v3.4.0.0.b
System Information
OS: Windows v11.22
CPU: 13th Gen Intel(R) Core(TM) i9-13900K  31.775 GB
GPU: NVIDIA GeForce RTX 4090  23.59 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	37.63 fps 	2X: 	17.46 fps 	4X: 	04.61 fps 	
Iris		1X: 	26.74 fps 	2X: 	14.31 fps 	4X: 	04.89 fps 	
Proteus		1X: 	31.97 fps 	2X: 	15.83 fps 	4X: 	04.63 fps 	
Gaia		1X: 	15.38 fps 	2X: 	10.45 fps 	4X: 	05.11 fps 	
4X Slowmo		Apollo: 	42.75 fps 	APFast: 	77.75 fps 	Chronos: 	32.24 fps 	CHFast: 	34.10 fps 	

Posting beta benchmarks is not a great idea since they are not meant to be directly comparable to official releases. I’m sure that category will be created once it’s released.

1 Like