Video AI 5.2.X - User Benchmarking Results

Topaz Video AI  v5.2.2
System Information
OS: Windows v11.23
CPU: AMD Ryzen 9 5950X 16-Core Processor              63.923 GB
GPU: AMD Radeon RX 7900 XTX  23.924 GB
Processing Settings
device: -2 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	22.62 fps 	2X: 	10.20 fps 	4X: 	02.77 fps 	
Iris		1X: 	24.75 fps 	2X: 	12.87 fps 	4X: 	03.60 fps 	
Proteus		1X: 	24.09 fps 	2X: 	13.68 fps 	4X: 	03.80 fps 	
Gaia		1X: 	00.22 fps 	2X: 	06.72 fps 	4X: 	02.99 fps 	
Nyx		1X: 	00.25 fps 	2X: 	08.24 fps 	
Nyx Fast		1X: 	18.05 fps 	
Rhea		4X: 	02.72 fps 	
4X Slowmo		Apollo: 	00.68 fps 	APFast: 	57.88 fps 	Chronos: 	16.31 fps 	CHFast: 	22.78 fps 	
16X Slowmo		Aion: 	28.38 fps 	

Still uses CPU for some tests on 5.2.2 despite having GPU selected in the settings.

Topaz Video AI  v5.2.2
System Information
OS: Mac v14.06
CPU: Apple M2 Ultra  128 GB
GPU: Apple M2 Ultra  96 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	19.25 fps 	2X: 	10.45 fps 	4X: 	03.23 fps 	
Iris		1X: 	11.74 fps 	2X: 	05.54 fps 	4X: 	02.66 fps 	
Proteus		1X: 	18.87 fps 	2X: 	11.92 fps 	4X: 	03.65 fps 	
Gaia		1X: 	05.82 fps 	2X: 	04.21 fps 	4X: 	03.13 fps 	
Nyx		1X: 	06.70 fps 	2X: 	06.06 fps 	
Nyx Fast		1X: 	20.28 fps 	
Rhea		4X: 	01.62 fps 	
4X Slowmo		Apollo: 	18.94 fps 	APFast: 	66.28 fps 	Chronos: 	06.49 fps 	CHFast: 	08.07 fps 	
16X Slowmo		Aion: 	20.72 fps 	

1 Like
Topaz Video AI  v5.2.3
System Information
OS: Windows v11.23
CPU: 12th Gen Intel(R) Core(TM) i5-12500  63.745 GB
GPU: NVIDIA GeForce RTX 4060  7.7705 GB
Processing Settings
device: 0 vram: 0.9 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	10.65 fps 	2X: 	07.32 fps 	4X: 	02.32 fps 	
Iris		1X: 	10.80 fps 	2X: 	06.38 fps 	4X: 	01.78 fps 	
Proteus		1X: 	10.17 fps 	2X: 	07.36 fps 	4X: 	02.44 fps 	
Gaia		1X: 	03.41 fps 	2X: 	02.28 fps 	4X: 	01.56 fps 	
Nyx		1X: 	04.15 fps 	2X: 	03.56 fps 	
Nyx Fast		1X: 	07.60 fps 	
Rhea		4X: 	01.36 fps 	
4X Slowmo		Apollo: 	15.28 fps 	APFast: 	40.10 fps 	Chronos: 	08.12 fps 	CHFast: 	12.59 fps 	
16X Slowmo		Aion: 	05.81 fps 	

Topaz Video AI  v5.2.3
System Information
OS: Windows v11.23
CPU: Intel(R) Xeon(R) w5-2465X  127.25 GB
GPU: NVIDIA GeForce RTX 4070  11.73 GB
GPU: NVIDIA RTX A4000  15.79 GB
Processing Settings
device: 0 vram: 0.9 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	18.12 fps 	2X: 	12.75 fps 	4X: 	03.85 fps 	
Iris		1X: 	18.54 fps 	2X: 	11.35 fps 	4X: 	03.28 fps 	
Proteus		1X: 	17.88 fps 	2X: 	12.35 fps 	4X: 	04.45 fps 	
Gaia		1X: 	06.18 fps 	2X: 	04.23 fps 	4X: 	02.89 fps 	
Nyx		1X: 	07.26 fps 	2X: 	06.15 fps 	
Nyx Fast		1X: 	13.54 fps 	
Rhea		4X: 	02.36 fps 	
4X Slowmo		Apollo: 	24.68 fps 	APFast: 	55.58 fps 	Chronos: 	13.16 fps 	CHFast: 	20.71 fps 	
16X Slowmo		Aion: 	22.71 fps 	

Topaz Video AI  v5.2.3
System Information
OS: Windows v10.22
CPU: Intel(R) Core(TM) i7-8700K CPU @ 3.70GHz  63.936 GB
GPU: NVIDIA GeForce RTX 3060  11.826 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	09.01 fps 	2X: 	06.62 fps 	4X: 	02.01 fps 	
Iris		1X: 	08.92 fps 	2X: 	05.75 fps 	4X: 	01.66 fps 	
Proteus		1X: 	08.87 fps 	2X: 	06.21 fps 	4X: 	02.15 fps 	
Gaia		1X: 	02.98 fps 	2X: 	02.07 fps 	4X: 	01.35 fps 	
Nyx		1X: 	03.86 fps 	2X: 	03.27 fps 	
Nyx Fast		1X: 	07.54 fps 	
Rhea		4X: 	01.16 fps 	
4X Slowmo		Apollo: 	12.99 fps 	APFast: 	31.47 fps 	Chronos: 	07.22 fps 	CHFast: 	11.94 fps 	
16X Slowmo		Aion: 	22.46 fps 	

I would be very pleased if any “Pro” user post its benchmarks, to verify the scaling statements made by Topaz.

Are there any “Pro” users here ?

That’s what you get with an AMD Radeon RX-7800 for about half the price of the Nvidia RTX 40xx.

Topaz Video AI  v5.2.2
System Information
OS: Windows v11.23
CPU: Intel(R) Core(TM) i9-14900K  127.77 GB
GPU: AMD Radeon RX 7800 XT  15.794 GB
GPU: Intel(R) UHD Graphics 770  0.125 GB
GPU: AMD Radeon RX 7800 XT [ZLUDA]  1.7163e+10 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis		1X: 	14.09 fps 	2X: 	09.64 fps 	4X: 	03.66 fps 	
Iris		1X: 	15.80 fps 	2X: 	09.41 fps 	4X: 	03.20 fps 	
Proteus		1X: 	14.36 fps 	2X: 	10.30 fps 	4X: 	04.58 fps 	
Gaia		1X: 	06.35 fps 	2X: 	04.25 fps 	4X: 	02.93 fps 	
Nyx		1X: 	06.10 fps 	2X: 	05.22 fps 	
Nyx Fast		1X: 	10.65 fps 	
Rhea		4X: 	01.77 fps 	
4X Slowmo		Apollo: 	19.35 fps 	APFast: 	56.53 fps 	Chronos: 	09.95 fps 	CHFast: 	16.36 fps 	
16X Slowmo		Aion: 	40.77 fps 	

But you have high end (bugged CPU) and much of RAM (>6000 MTs ?)
This help also.

Have you compared on the same PC a RTX 4090 ?

May be half the price but the performance ?

The 14900K is being held back by the 7800 XT.

You could get the same reults with a less-expensive CPU or you could double the performance of your PC by upgrading your GPU to a 4080S -Maybe even a 4070-Ti-Super.

Not sure what “bugged” CPU means?
RAM over >6000 MTs – what does that mean?

RTX 4090, I guess that’s the NVidia? that’s about 3 times the price of the Radeon. I don’t have those hardware laying around to try :-).

I understand that there are more powerful and (way) more expensive GPUs out there. I just posted to show what’s possible with a more modest - and very quiet - AMD Radeon 7800.

My problem is not performance, my problem is the quality of the output, which I lower than I hoped for. But I won’t complain that my 1960s and 1970s Super-8 footage doesn’t look like UHD Hollywood movie.

Quality has never been outputed through a GPU, because of its floating point approximations :slight_smile:

I don’t flame AMD or Nvidia or even your setting, but I did test many GPU on my side, both Nvidia and AMD, Nvidia are fare ahead in terms of power/price.

Even a GTX 1080 Ti that is Tensorless kick the ass, but you need 10 of it :sweat_smile:

Hope you’ll get the best out your 7800XT :slight_smile: And may be open an opporutnity for Topaz to port also OpenCL so Nvidia Tensors will be ‘less’ ahead

“Bugged” is probably referring to the voltage / degradation issue where Intel 13th and 14th gen CPU’s with TDP’s above 65w are degrading very quickly:

https://hardforum.com/threads/intel-has-a-problem-vast-majority-of-game-crashes-come-from-13th-14th-gen-intel-cpus-during-decompression.2035841/

The chips are requesting too much voltage and end up with very short life spans depending on how the CPU is used. I have limited my max VID to 1.4V in the BIOS as a temporary fix, but Intel has released a micro code update that is supposed to stop the chips from degrading. (But will not fix one that is already degraded)

2 Likes

thank you, I wasn’t aware of it.

Indeed I was thinking of that. To me getting such 13 or 14 gen CPu running a long time will lead to a crash and then drop out the encoded video to trash :frowning:

I haven’t owned an Intel for at least 15 years. Their hegemonic position and the will to kill AMD at all costs finally pushed them to do the opposite and that makes me happy for healthy competition.

I have recently owned a 7950X which will be pushed to its limits which are thermal. For that I unpacked my waterchiller which had been waiting for a real CPU for years :sweat_smile:

But since the weather is very hot and humid at the moment, my margin of progress is tight otherwise I have to isolate a lot of things. Fortunately there is 3D printing now which opens up so many possibilities.

Else, it seems that I learnt from Intel benchmarck is that memory is very important. I don’t know if TVAI uses DirectStorage or not but it would change things if so.
My actual Micron DDR5 kit is pure crap, going for a true low latency 6000 MTs kit, and forgetting also multiple rendering to save it. I will do my multi rendering on the 2990WX that love tons of jobs simultaneously.

So if you owners of Intel CPU + godlike DDR5, I would be pleased you check your DDR5 bandwidth use. (HWinfo 64 / Memory bandwidth read + write) to compare with AMD’s
I got some various tests around the GPU to find out that RTX 3070 will pump 3.5 to 5 GBps on PCIE bus, it fits on Gen3 x4 (aiming 7x RTX 3070 on 2990WX)
Also analysing the behaviour of mGPU, because it’s unclear for me that TVAI scales as advertised. It is closed to, but only in some scenarios.

I limited my max VID to 1.4V and also limited max boost frequency to 5.5 Ghz since I have never seen it do any real work above that frequency.

Performance is ever-so-slightly reduced, but well worth the peace of mind while I wait and see how the new AGESA fix plays out.

2 Likes

Ok thanks for this feedback. I will share mine soon

I am currently @ 5.2 max boost on air cooling Phantom with hyper noisy (but efficent) Delta 120 x2.
I’ve got some huge spikes in a setting 2x 3070, tomorrow I will check 3x 3070 aso

My render carshed after 2 days because I siimply “used” my PC, so i catch the opportunity to do some upgrades

If you think your CPU has already degraded, decompression seems to be the work load that most-consistently exposes the instability. Try installing two different Nvidia driver packages (back and forth between two different versions) multiple times. If it crashes at stock settings you probably need to RMA the CPU. (Intel will probably ask you to try the new BIOS first)

I don’t have any Intel CPU
I have 3900X / 5900X / 7950X / 1950X / 2990WX