Users can run benchmarking (Process > Benchmark, or, Ctrl/Cmd + B) to compare results across different machines. This option doesn’t transmit any data and is completely up to the user to share. Once benchmarking is finished you can copy and paste the results here.
Topaz Video AI v3.5.0
System Information
OS: Windows v11.22
CPU: 12th Gen Intel(R) Core(TM) i9-12900K 31.818 GB
GPU: NVIDIA GeForce RTX 4070 11.744 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 18.68 fps 2X: 12.12 fps 4X: 03.30 fps
Iris 1X: 19.83 fps 2X: 10.57 fps 4X: 03.32 fps
Proteus 1X: 17.49 fps 2X: 11.43 fps 4X: 03.50 fps
Gaia 1X: 06.36 fps 2X: 04.35 fps 4X: 02.96 fps
Nyx 1X: 07.51 fps
4X Slowmo Apollo: 25.62 fps APFast: 59.23 fps Chronos: 13.66 fps CHFast: 21.40 fps
1 Like
david.123
September 20, 2023, 2:54am
2
Topaz Video AI v3.5.0
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 7950X 16-Core Processor 63.732 GB
GPU: NVIDIA GeForce RTX 4090 23.59 GB
GPU: NVIDIA GeForce RTX 4090 23.59 GB
GPU: NVIDIA GeForce RTX 4090 23.59 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 41.74 fps 2X: 15.49 fps 4X: 04.40 fps
Iris 1X: 37.18 fps 2X: 19.05 fps 4X: 05.37 fps
Proteus 1X: 35.33 fps 2X: 15.74 fps 4X: 04.48 fps
Gaia 1X: 15.61 fps 2X: 10.75 fps 4X: 05.23 fps
Nyx 1X: 17.19 fps
4X Slowmo Apollo: 43.86 fps APFast: 89.99 fps Chronos: 31.28 fps CHFast: 38.17 fps
Note: there is a documented bug where multiple instances of a singe GPU show up in the list of GPU’s. I only have one 4090 card, not three.
Topaz Video AI v3.5.0
System Information
OS: Windows v11.21
CPU: 12th Gen Intel(R) Core(TM) i9-12900KF 127.78 GB
GPU: NVIDIA GeForce RTX 4060 Ti 15.745 GB
Processing Settings
device: 0 vram: 0.95 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 14.56 fps 2X: 09.15 fps 4X: 02.79 fps
Iris 1X: 15.11 fps 2X: 07.40 fps 4X: 02.19 fps
Proteus 1X: 12.44 fps 2X: 08.44 fps 4X: 03.04 fps
Gaia 1X: 04.49 fps 2X: 03.13 fps 4X: 02.09 fps
Nyx 1X: 05.39 fps
4X Slowmo Apollo: 18.11 fps APFast: 49.44 fps Chronos: 10.57 fps CHFast: 15.87 fps
Iris V2 actually improve the speed performance by around 50% @1X and 25% @2X per benchmark results.
2 Likes
Topaz Video AI v3.5.0
System Information
OS: Windows v11.21
CPU: AMD Ryzen 9 5900X 12-Core Processor 31.903 GB
GPU: NVIDIA GeForce RTX 3080 Ti 11.816 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 22.43 fps 2X: 09.74 fps 4X: 02.75 fps
Iris 1X: 18.40 fps 2X: 10.30 fps 4X: 03.36 fps
Proteus 1X: 18.13 fps 2X: 08.44 fps 4X: 02.57 fps
Gaia 1X: 08.02 fps 2X: 05.51 fps 4X: 03.31 fps
Nyx 1X: 09.92 fps
4X Slowmo Apollo: 30.05 fps APFast: 53.50 fps Chronos: 17.97 fps CHFast: 26.37 fps
Topaz Video AI v3.5.0
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 7900X 12-Core Processor 31.749 GB
GPU: NVIDIA GeForce RTX 4070 Ti 11.729 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 24.57 fps 2X: 13.10 fps 4X: 03.56 fps
Iris 1X: 26.16 fps 2X: 13.35 fps 4X: 04.01 fps
Proteus 1X: 22.69 fps 2X: 12.83 fps 4X: 03.78 fps
Gaia 1X: 08.28 fps 2X: 05.74 fps 4X: 03.78 fps
Nyx 1X: 09.59 fps
4X Slowmo Apollo: 35.58 fps APFast: 74.35 fps Chronos: 18.89 fps CHFast: 29.28 fps
1 Like
jo.vo
September 20, 2023, 7:53am
5
Apparently you still use Iris V1 for the benchmark. Why not V2?
lhkjacky
September 20, 2023, 8:03am
6
Topaz Video AI v3.5.0
System Information
OS: Windows v11.22
CPU: 13th Gen Intel(R) Core(TM) i7-13700K 31.773 GB
GPU: NVIDIA GeForce RTX 4070 11.744 GB
GPU: Intel(R) UHD Graphics 770 0.125 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 19.67 fps 2X: 13.01 fps 4X: 04.32 fps
Iris 1X: 20.34 fps 2X: 11.17 fps 4X: 03.53 fps
Proteus 1X: 18.14 fps 2X: 12.12 fps 4X: 04.57 fps
Gaia 1X: 06.66 fps 2X: 04.60 fps 4X: 03.15 fps
Nyx 1X: 07.95 fps
4X Slowmo Apollo: 26.86 fps APFast: 73.07 fps Chronos: 14.23 fps CHFast: 22.77 fps
kiprode
September 20, 2023, 8:40am
7
iMac 2019
Topaz Video AI v3.5.0
System Information
OS: Mac v11.071
CPU: Intel(R) Core(TM) i5-9600K CPU @ 3.70GHz 80 GB
GPU: AMD Radeon Pro 580X 8 GB
Processing Settings
device: 0 vram: 0.84 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 01.66 fps 2X: 01.07 fps 4X: 00.43 fps
Iris 1X: 01.54 fps 2X: 00.57 fps 4X: 00.26 fps
Proteus 1X: 01.68 fps 2X: 01.15 fps 4X: 00.49 fps
Gaia 1X: 00.57 fps 2X: 00.40 fps 4X: 00.31 fps
Nyx 1X: 00.67 fps
4X Slowmo Apollo: 01.46 fps APFast: 06.29 fps Chronos: 00.34 fps CHFast: 00.64 fps
GTOMAN
September 20, 2023, 9:04am
8
Topaz Video AI v3.5.0
System Information
OS: Windows v10.22
CPU: AMD Ryzen 7 3700X 8-Core Processor 63.952 GB
GPU: NVIDIA GeForce RTX 4090 23.59 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 29.09 fps 2X: 12.36 fps 4X: 03.16 fps
Iris 1X: 34.57 fps 2X: 12.68 fps 4X: 03.49 fps
Proteus 1X: 19.35 fps 2X: 11.18 fps 4X: 02.95 fps
Gaia 1X: 15.66 fps 2X: 09.78 fps 4X: 03.00 fps
Nyx 1X: 18.83 fps
4X Slowmo Apollo: 34.49 fps APFast: 51.00 fps Chronos: 31.29 fps CHFast: 28.55 fps
lemans99
September 20, 2023, 1:56pm
9
Someone can explain me this ?
3060 has Tensor FP16 compute power ~51.2 TFLOPS (Proteus 2X ~05.88 fps ) ~8.7tflops per frame 4060 has Tensor FP16 compute power ~60.0 TFLOPS (Proteus 2X ~06.89 fps ) ~8.7tflops per frame 4090 has Tensor FP16 compute power ~330.0 TFLOPS (Proteus 2X ~17.00 fps ) ~19.4tflops per frame
Bad programme optimization or where is the bottleneck ?!
Any mention of ‘trillion floating-point operations per second’ in compute power online, can only be used in the context of whatever program was used to measure that number. Is TVAI only doing pure floating-point operations? Doubtful. Even if it was, it still has to load and unload the GPU. Pretty sure those numbers are from the absolute best case scenarios with synthetic loads—Again, only useful for comparing other GPUs running the exact same load from the exact same program.
All that aside, people have confirmed, time and time again on these forums, that CPU computing power plays a big part in TVAI processing speeds.
1 Like
lemans99
September 20, 2023, 2:39pm
11
Understand this but the fact remains that VAI cant fully load more powerful GPUs like 4090 and use their full compute potential for some reason (CPU or memory bandwidth ) . So i think it`s kinda bad VAI optimization and GPUs like 4090 seems to be overkill for VAI …
Those TFLOPS numbers are like comparing a fruit to an animal, when comparing what program made those numbers to TVAI. They are very different workloads.
Can TVAI be optimized for better speed and utilization? Always. Is that easy? Not always.
benchmarks.json from v3.5.0 should be using Iris v2:
"Iris": {
"1X": {
"filter": "tvai_up=model=iris-2:scale=1[DEVICE_INFO]"
},
"2X": {
"filter": "tvai_up=model=iris-2:scale=2[DEVICE_INFO]"
},
"4X": {
"filter": "tvai_up=model=iris-2:scale=4[DEVICE_INFO]"
}
},
2 Likes
Imo
September 20, 2023, 3:36pm
14
Topaz Video AI v3.5.0
System Information
OS: Windows v11.22
CPU: 13th Gen Intel(R) Core(TM) i9-13900K
MEM: 31.775 GB DDR5 6000 MHZ CL32-38-38 @ 1.35 V XMP 3
GPU: NVIDIA GeForce RTX ASUS ROG STRIX OC 4090 22.096 GB
MAX CPU WATT: 176.2
MAX GPU WATT: 465.9
MAX CPU TEMP: 87 °C
MAX GPU TEMP: 67 °C
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 45.11 fps 2X: 21.79 fps 4X: 05.34 fps
Iris 1X: 41.09 fps 2X: 20.37 fps 4X: 05.90 fps
Proteus 1X: 37.70 fps 2X: 18.82 fps 4X: 05.56 fps
Gaia 1X: 14.50 fps 2X: 10.70 fps 4X: 05.31 fps
Nyx 1X: 17.78 fps
4X Slowmo
Apollo: 42.23 fps
APFast: 85.43 fps
Chronos: 33.26 fps
CHFast: 35.46 fps
I can confirm that I am getting almost a 50 percent speed increase compared to Iris V1.
david.123
September 20, 2023, 7:58pm
16
I read an article talking about the L2 cache on the 4090, being 72MB, which allows Video AI to fly on 720p and lower inputs because it can store many frames directly in L2 cache. Once you hit 1080p and up, it has to swap from VRAM to L2 cache a lot more which is why the performance drops so much on higher res. inputs.
That being said, there are probably specific optimizations on the 4xxx series that can still be implemented.
Perhaps you can research them and suggest them to the devs.
2 Likes
rfrankway
September 21, 2023, 3:04am
17
Topaz Video AI v3.5.0
System Information
OS: Mac v13.0502
CPU: Apple M2 Max 96 GB
GPU: Apple M2 Max 72 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 11.39 fps 2X: 07.28 fps 4X: 02.62 fps
Iris 1X: 07.17 fps 2X: 04.00 fps 4X: 01.01 fps
Proteus 1X: 11.24 fps 2X: 07.01 fps 4X: 02.40 fps
Gaia 1X: 03.30 fps 2X: 02.40 fps 4X: 01.82 fps
Nyx 1X: 03.39 fps
4X Slowmo Apollo: 12.07 fps APFast: 45.58 fps Chronos: 03.70 fps CHFast: 06.00 fps
y-komura
September 21, 2023, 7:25am
18
Topaz Video AI v3.5.0
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: 22.44 fps 2X: 11.42 fps 4X: 03.36 fps
Iris 1X: 21.11 fps 2X: 11.73 fps 4X: 03.71 fps
Proteus 1X: 22.29 fps 2X: 11.05 fps 4X: 03.33 fps
Gaia 1X: 07.82 fps 2X: 05.26 fps 4X: 03.21 fps
Nyx 1X: 09.56 fps
4X Slowmo Apollo: 27.48 fps APFast: 67.72 fps Chronos: 17.37 fps CHFast: 26.86 fps
TPX
September 21, 2023, 9:20am
19
Please post 4K too.
Topaz Video AI v3.5.0
System Information
OS: Windows v11.22
CPU: AMD Ryzen Threadripper 3960X 24-Core Processor 127.88 GB
GPU: AMD Radeon PRO W6800 29.956 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 10.27 fps 2X: 06.47 fps 4X: 02.11 fps
Iris 1X: 10.86 fps 2X: 06.26 fps 4X: 02.02 fps
Proteus 1X: 09.64 fps 2X: 05.76 fps 4X: 02.07 fps
Gaia 1X: 04.67 fps 2X: 03.15 fps 4X: 02.22 fps
Nyx 1X: 04.03 fps
4X Slowmo Apollo: 13.60 fps APFast: 46.22 fps Chronos: 06.26 fps CHFast: 11.38 fps
Topaz Video AI v3.5.0
System Information
OS: Windows v11.22
CPU: AMD Ryzen Threadripper 3960X 24-Core Processor 127.88 GB
GPU: AMD Radeon PRO W6800 29.956 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 3840x2160
Benchmark Results
Artemis 1X: 02.17 fps 2X: 01.34 fps 4X: 00.44 fps
Iris 1X: 02.31 fps 2X: 01.34 fps 4X: 00.43 fps
Proteus 1X: 02.07 fps 2X: 01.30 fps 4X: 00.43 fps
Gaia 1X: 01.00 fps 2X: 00.68 fps 4X: 00.48 fps
Nyx 1X: 00.68 fps
4X Slowmo Apollo: 03.34 fps APFast: 11.51 fps Chronos: 01.36 fps CHFast: 02.86 fps
glitterkill
September 21, 2023, 1:10pm
20
Someday maybe AMD GPUs will get some optimization love here. Until then, I will just keep using Iris.
Topaz Video AI v3.5.0
System Information
OS: Windows v11.22
CPU: AMD Ryzen 9 7900X 12-Core Processor 63.118 GB
GPU: AMD Radeon RX 7900 XTX 23.94 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 12.37 fps 2X: 06.12 fps 4X: 01.95 fps
Iris 1X: 16.63 fps 2X: 08.89 fps 4X: 02.78 fps
Proteus 1X: 10.13 fps 2X: 05.68 fps 4X: 01.96 fps
Gaia 1X: 08.53 fps 2X: 05.53 fps 4X: 03.22 fps
Nyx 1X: 08.04 fps
4X Slowmo Apollo: 23.10 fps APFast: 48.65 fps Chronos: 12.39 fps CHFast: 15.42 fps
1 Like
