Starting from Topaz Video AI v3.1.9, 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.
Fresh reboot (waited for everything to settle down), no other tasks running and Run1 is first pass at benchmarks, then Run2 is right after that to see if there is any caching effect. Fresh run on 3.1.11 before updating to compare with.
3.1.11 (fresh reboot):
Run1:
Topaz Video AI v3.1.11
System Information
OS: Mac v13.03
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: 10.42 fps 2X: 7.42 fps 4X: 2.64 fps
Proteus 1X: 10.83 fps 2X: 7.43 fps 4X: 2.19 fps
Gaia 1X: 3.39 fps 2X: 2.52 fps 4X: 1.9 fps
4X Slowmo Apollo: 11.19 fps Chronos: 3.31 fps Chronos Fast: 5.3 fps
Run2:
Topaz Video AI v3.1.11
System Information
OS: Mac v13.03
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: 11.06 fps 2X: 7.34 fps 4X: 2.5 fps
Proteus 1X: 10.89 fps 2X: 7.01 fps 4X: 2.36 fps
Gaia 1X: 3.28 fps 2X: 2.36 fps 4X: 1.71 fps
4X Slowmo Apollo: 10.57 fps Chronos: 3.34 fps Chronos Fast: 5.29 fps
3.2.0 (fresh reboot):
Run1:
Topaz Video AI v3.2.0
System Information
OS: Mac v13.03
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: 10.56 fps 2X: 7.64 fps 4X: 2.90 fps
Proteus 1X: 11.26 fps 2X: 7.29 fps 4X: 2.50 fps
Gaia 1X: 3.34 fps 2X: 2.44 fps 4X: 1.88 fps
4X Slowmo Apollo: 9.89 fps Chronos: 3.76 fps Chronos Fast: 6.22 fps
Run2:
Topaz Video AI v3.2.0
System Information
OS: Mac v13.03
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: 10.69 fps 2X: 7.37 fps 4X: 2.90 fps
Proteus 1X: 11.05 fps 2X: 7.34 fps 4X: 2.46 fps
Gaia 1X: 3.31 fps 2X: 2.29 fps 4X: 1.71 fps
4X Slowmo Apollo: 9.08 fps Chronos: 3.59 fps Chronos Fast: 5.99 fps
I know this is not a feature request forum, but going to put this out here anyway: A timestamp field in the results would be useful. Local or UTC, good either way.
Theoretical its more then twice faster then v3.1.9 - which is very good work ! But …
I decided to make “real life” test and the results unfortunately are pretty disturbing
By the benchmark i get about 30fps for 640x480p input upscaled 2X by Proteus ! But when i try to upscale real 640x480p@29.97fps vid 2X by Proteus (no other enhancements enabled) it takes about 4 minutes for 1 minute of the original vid . Or this is about 4 times slower then the data from the benchmark !!!
So my question is - is this some kind of bug or this is just a SCAM benchmark info to mislead the users that there is really great improvement in the newer Topaz software ?
Everybody can make such “real life” test for himself and compare with the benchmark info ! Will be nice to hear about other users experience . So please share it here!
I think the benchmark either writes the output to RAM or just ignores it. This can make a huge difference on speed. In the tests I’ve done before, using NVENC slows things down compared to PNG output. PNG output on a slow HDD goes fast until the cache gets filled, then the whole thing slows down to a crawl. I have SSDs that work the same way, but slow down even more once the cache is filled.
So yes. The speeds will not be the same. Benchmarks should try to eliminate as many extra variables as they can to test only the hardware in question. I believe they have removed the encoding impact from this benchmark.
You cannot compare like this. The video content and encoding is likely very different (10-bit vs. 8-bit, busy content like water scene vs. mostly still, the codec involved, fps, etc.). The only way to compare is to run the same real video on 3.1.x vs 3.2.0. What is that result? The world is not out to get you.
Proteus 2X: 20.64 fps At first looks like nice performance but i believe Topaz isnt yet good optimized for RTX 4xxx series … My RTX 3060 get Proteus 2X: 6.80 fps which is about just 3 times slower (even at 11 years old i5-2500K) . Tensor compute power of the 3060 is 51.2 TFLOPS while this of 4090 is 330.3 TFLOPS which is more then 6 times higher … So my point is that if the software is really good optimized to use the full power of 4090 the Proteus performance has to be about twice higher or about 40fps …
Also from 3090Ti results above get Proteus 2X: 14.73 fps (which seems logically compared to 3060) but the tensor compute power of the 3090Ti is just the half of 4090 or its just 160 TFLOPS My point is thats not realistic 3090Ti performance to be so close to 4090 at very same CPU and OS …