Video AI 5.4.X - User Benchmarking Results
Users can run benchmarking (Process > Benchmark, or, Ctrl/Cmd + B) to compare results across different machines.
Topaz Video AI v5.4.0
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: 09.51 fps 2X: 06.36 fps 4X: 01.68 fps
Iris 1X: 09.22 fps 2X: 05.81 fps 4X: 01.63 fps
Proteus 1X: 09.17 fps 2X: 06.42 fps 4X: 02.11 fps
Gaia 1X: 03.06 fps 2X: 02.12 fps 4X: 01.41 fps
Nyx 1X: 03.81 fps 2X: 03.24 fps
Nyx Fast 1X: 07.50 fps
Rhea 4X: 01.17 fps
4X Slowmo Apollo: 13.36 fps APFast: 33.35 fps Chronos: 07.48 fps CHFast: 12.43 fps
16X Slowmo Aion: 22.69 fps
I haven’t bothered doing this for quite a while, but I got a new PC today so I thought it might be a good idea to see how it does:
Topaz Video AI v5.4.0
System Information
OS: Windows v11.24
CPU: Intel(R) Core(TM) i9-14900KF 63.725 GB
GPU: NVIDIA GeForce RTX 4090 23.576 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 40.31 fps 2X: 18.01 fps 4X: 04.82 fps
Iris 1X: 38.64 fps 2X: 21.25 fps 4X: 05.65 fps
Proteus 1X: 40.69 fps 2X: 21.23 fps 4X: 05.70 fps
Gaia 1X: 15.57 fps 2X: 10.64 fps 4X: 05.36 fps
Nyx 1X: 17.22 fps 2X: 14.38 fps
Nyx Fast 1X: 30.89 fps
Rhea 4X: 05.19 fps
4X Slowmo Apollo: 40.76 fps APFast: 73.15 fps Chronos: 32.79 fps CHFast: 35.98 fps
16X Slowmo Aion: 34.47 fps
4 Likes
Nice specs on that new rig!
1 Like
This is what I am working with. Looks like it’s time for an upgrade like PaulM’s
Topaz Video AI v5.4.0
System Information
OS: Windows v10.22
CPU: AMD Ryzen Threadripper 3960X 24-Core Processor 127.88 GB
GPU: NVIDIA GeForce RTX 3090 23.77 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 24.26 fps 2X: 13.32 fps 4X: 03.74 fps
Iris 1X: 22.10 fps 2X: 13.91 fps 4X: 04.10 fps
Proteus 1X: 23.53 fps 2X: 15.45 fps 4X: 04.09 fps
Gaia 1X: 08.14 fps 2X: 05.55 fps 4X: 03.42 fps
Nyx 1X: 09.95 fps 2X: 08.28 fps
Nyx Fast 1X: 19.52 fps
Rhea 4X: 03.10 fps
4X Slowmo Apollo: 32.51 fps APFast: 62.63 fps Chronos: 18.52 fps CHFast: 28.47 fps
16X Slowmo Aion: 28.11 fps
Topaz Video AI v5.4.0
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: -2 vram: 0.9 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 18.12 fps 2X: 12.19 fps 4X: 03.62 fps
Iris 1X: 18.03 fps 2X: 11.24 fps 4X: 03.23 fps
Proteus 1X: 17.72 fps 2X: 12.05 fps 4X: 04.30 fps
Gaia 1X: 06.08 fps 2X: 04.24 fps 4X: 02.88 fps
Nyx 1X: 07.22 fps 2X: 06.12 fps
Nyx Fast 1X: 13.44 fps
Rhea 4X: 02.36 fps
4X Slowmo Apollo: 24.65 fps APFast: 55.08 fps Chronos: 13.08 fps CHFast: 20.50 fps
16X Slowmo Aion: 24.21 fps
oh wow, v5.4.0 released, seems similar to me.
Topaz Video AI v5.4.0
System Information
OS: Windows v11.23
CPU: Intel(R) Core(TM) i7-14700K 31.781 GB
GPU: NVIDIA GeForce RTX 4090 23.576 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: 41.45 fps 2X: 18.89 fps 4X: 05.39 fps
Iris 1X: 38.04 fps 2X: 22.94 fps 4X: 05.28 fps
Proteus 1X: 41.69 fps 2X: 22.41 fps 4X: 06.26 fps
Gaia 1X: 15.32 fps 2X: 10.56 fps 4X: 05.24 fps
Nyx 1X: 17.40 fps 2X: 14.82 fps
Nyx Fast 1X: 31.10 fps
Rhea 4X: 05.19 fps
4X Slowmo Apollo: 43.64 fps APFast: 84.85 fps Chronos: 33.47 fps CHFast: 37.98 fps
16X Slowmo Aion: 34.45 fps
Topaz Video AI v5.4.0
System Information
OS: Windows v11.24
CPU: Intel(R) Core(TM) i7-9700K CPU @ 3.60GHz 31.843 GB
GPU: NVIDIA GeForce RTX 3080 9.8174 GB
GPU: Intel(R) UHD Graphics 630 0.125 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 19.35 fps 2X: 10.33 fps 4X: 02.94 fps
Iris 1X: 18.93 fps 2X: 11.14 fps 4X: 03.22 fps
Proteus 1X: 18.68 fps 2X: 12.12 fps 4X: 03.22 fps
Gaia 1X: 06.54 fps 2X: 04.55 fps 4X: 02.83 fps
Nyx 1X: 08.35 fps 2X: 06.61 fps
Nyx Fast 1X: 15.65 fps
Rhea 4X: 02.63 fps
4X Slowmo Apollo: 22.28 fps APFast: 44.99 fps Chronos: 14.60 fps CHFast: 22.06 fps
16X Slowmo Aion: ... fps
May I ask - did you set the processing device to Auto on purpose? The reason I ask is that there was a bug in a previous version which set the processing device to Auto without the user knowing. This caused significant performance loss for Apple silicon users, but I’m not sure what effect is has on Windows.
Thanks.
Andy
Nothing changed for me.
Topaz Video AI v5.4.0
System Information
OS: Windows v11.24
CPU: Intel(R) Core(TM) i7-14700K 31.747 GB
GPU: NVIDIA GeForce RTX 4080 SUPER 15.671 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 30.41 fps 2X: 20.71 fps 4X: 05.73 fps
Iris 1X: 30.86 fps 2X: 19.91 fps 4X: 05.06 fps
Proteus 1X: 29.45 fps 2X: 19.99 fps 4X: 06.68 fps
Gaia 1X: 10.81 fps 2X: 07.43 fps 4X: 04.90 fps
Nyx 1X: 12.49 fps 2X: 10.64 fps
Nyx Fast 1X: 22.74 fps
Rhea 4X: 04.18 fps
4X Slowmo Apollo: 42.01 fps APFast: 106.25 fps Chronos: 23.75 fps CHFast: 36.06 fps
16X Slowmo Aion: 38.84 fps
“New PC”
You mean, “a hell of a PC”
You can do 99% of the things i do.
I’ve 128GB ram.
Had the 3960 b4 too.
Now 7950X.
Much faster in some use cases and much less energy waste.
Looking 4 to get an 9950X3D.
I hope that the cache will help with visual computing (image editing, realtime on screen processing), and some AI software processing that is performed on CPU (Capture One (Match Look, Masks, Smart Presets).
Hello.
Yeah, that would require an in-house engine rebuild, not just a new stack version build.
Hello All. As I own by myself a Mac Mini M1 with 16GB of memory, which I use for 7x24 Video Upscaling of all my old DVD’s and BD’s to FHD or even 4K. It did this now for quite a while very energy efficient and silent 7x24. As my brother now obtained a base M4 Mac Mini with it’s 16 GB of RAM, I took some time to run various TVAI benchmarks on the M1 vs. the M4 to estimate, how much a upgrade to a M4 (maybe even pro or max) may help speeding up TVAI tasks.
I list all the benchmarks below, most of them I did with Input Resolution of 768x576, as I’m mainly Enhancing my DVD collection, but also did some of the default 1920x1080 benchmarks to compare the (non Pro or Max) M1 vs M4 results and take some guesses out of it and also compare to M4 Pro or Max Results of other users here in the community.
I mainly use the RHEA or IRIS enhancers, sometimes Proteus, Gaia or Dione Variants too. Artemis makes too “perfect” upscales with plastic faces and artifacts in faces with low quality inputs, which is why I don’t use them too much anymore.
In terms of upscaling with a 768x576 input source I have following findings to share:
For my base M1 Mac Mini with 16 GB of RAM I found the “Low Power” enabled mode for memory settings between 50 and 100 % most effective - especially RHEA seems to benefit noticeable in that mode, when mainly Neural Engine and CPU gets used. The often in this community mentioned “tweak” to set memory usage to 10% to prefer GPU usage over Neural Engine seems in case of the base M1 processor counter productive, any Memory setting below 50% gives - depending on used upscaling model - a noticeable drop in fps.
Looks for me somehow, that on a base M1 processor 8 core GPU the Neural Engine outperforms the GPU in most cases.
For the base M4 processor with it’s super effective Neural core and the 10 GPU cores it behaves in widest parts comparable to the M1 behavior described above as memory settings between 50…100% seems to work best, any lower setting gives a noticeable fps drop. Where it starts to differ is obviously that the new Neural Engine with its 38 TOPS in M4 vs 11 TOPS in M1 really shines in a model as RHEA, which gets nearly double the fps in low power enabled mode compared to the way slower mode in Low Power disabled mode (0.45 fps vs. 0.23 fps with 1920x1080 input source, or 2.20 fps vs. 1.28 fps with 768x576 input).
If I compare this with Benchamark results from community members owning M4 PRO or MAX, their additional GPU’s are adding up and overtake the Neural Engines performance (which is in all M4 variants the same). The M4 PRO’s are coming from my knowledge with 16 or 20 GPU cores, and the MAX variants with up to 40 GPU cores, which is a whopping 4x GPU capacity of the base M4. This also seems to add up with the RHEA results for example (when Low Power mode is DISABLED) for a 1920x1080 input:
M4 (Base) 10 GPU cores / 16 GB RAM 0.27 fps
M4 PRO 16 GPU cores / 24 GB RAM 0.43 fps
M4 MAX 40 GPU cores / 128 GB RAM 1.13 fps
So for the Base M1 (and probably even M2 or M3 CPU’s) it seems to be the case, that enforcing their Neural Engine is not only the most performant and efficient variant in terms of performance and cooling fan noise, but also seems to outperform it’s low number of GPU cores. Iassume the whole M1, M2 and M3 PRO’s, MAX’s and ULTRA’s might start performing quite well using GPU instead Neural Engine - opposite to their base model which performs better with Neural Engine obviously.
For the Base M4 processor the Neural Engine and the base GPU cores seems to operate nearly on par with it’s 10 GPU cores…depending on the used upscaling model in most cases the “Low Power” enabled or disabled brings a comparable performance, solely the RHEA upscaler benefits overwhelmingly from enforcing the low power mode by nearly doubling the fps over the “Low Power” mode disabled.
In case of M4 Pro and MAX it seems, that their added GPU cores really starts outperforming the M4’s Neural core in TVAI tasks. In case of the 16 GPU core M4 PRO the fps results comes more or less on par with RHEA running on the Neural Core. I expect the 20 GPU core Variant to outperform the “Low Power” Neural Engine Performance slightly, and the M4 Max - as we see above - definitely outperforms the Neural Engine by far.
While I definitely want to upgrade my old M1 to a M4, I’m not sure in terms of TVAI upscaling performance, do I go with the “bang for the buck” entry level base M4 Mac Mini for around EUR 600, a 20 GPU Core M4 PRO Mac Mini which starts around EUR 1600 for 24 GB RAM and 512 GB Storage (with lots of upsell options), or one of the devices with M4 MAX as a MacBook Pro, which is available with the M4 MAX Processor with up 40 GPU cores, and pricing wise…yeah…let’s say “way out of reach or what I am willing to to pay”
So for the minimal gain I might get expect within TVAI by going with the faster M4 PRO with 20 GPU cores compared to the base M4 and Neural Engine…not sure, it’s worth the added price tag of a factor 2,5 for the PRO. The M4 MAX Version might bring a real benefit by a factor 3 to 4 in fps for upscaling performance. But here we’re entering price ranges, where we really also can start consider a NVIDIA RTX4090 based system instead a apple one, where - based on numerous community benchmarks here - a RTX4090 outperforms the highly priced M4 MAX by at least a factor 2,5 (Proteus) or even up to 4 (RHEA).
Considering all that factors Apple M4 wise specifically in terms of TVAI, I’d personally say the base M4 probably brings most value in relation the spent money, the next performance wise noteworthy big step on apple silicon seems to be a M4 MAX based system. For TVAI I don’t see too much benefit out of an M4 PRO in relation to its added price tag compared to base M4.
For me personally I consider a M4 PRO anyway, as I also started playing around with local LLM’s as GPT4ALL / OLLAMA etc., and the added GPU’s and larger Memory Options of 48 or 64 GB makes it really attractive for that local LLM/AI Stuff besides the TVAI upscaling for me. From a pure working perspective with Video watching, Mailing, Office etc. I’m still perfectly happy with the 16 GB M1 Mac Mini - Apple really hit the nail with their apple silicon.
Maybe to set the TVAI performance plus minus in relation to the NVIDIA / Windows based systems…The Base M4 performs in my eyes plus minus comparable to my Son’s entry level Gaming Laptop with a AMD Ryzen 7 5800H, 16GB RAM and its NVIDIA GeForce RTX 3050 Ti Laptop GPU with 4 GB.
Depending on used Upscaling Models the M4 PRO CPU’s seems to perform in the RTX3060 or 4060 region, The M4MAX in best case around a RTX4060…4070 region. The 3080, 3090, 4080 and 4090 are still too capable, Apple Silicon isn’t there yet ;o)
PS: To the M4 Pro and Max owners out there…I really would appreciate some comparative benchmarks for 100% Memory settings in TVAI with “Low Power” mode enabled and disabled to see the difference on the various models.
It’s really noteworthy, how much the various memory settings and low power mode setting can benefit or hit the result.
Here we go…btw. the “vram” tells the memory setting, 1=100%, 0.75 =75% and so on. Instances = is low power mode (mainly tries to use CPU and Neural Engine), 1 low power mode disabled (also uses GPU)
==> Baseline Benchmarks with my base M1 Mac Mini with 16 GB RAM and 768x576 input source with various memory and power settings
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 100%) = Low Power Mode Enabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 17.57 fps 2X: 09.93 fps 4X: 03.58 fps
Iris 1X: 12.77 fps 2X: 06.39 fps 4X: 01.90 fps
Proteus 1X: 16.63 fps 2X: 10.08 fps 4X: 03.12 fps
Gaia 1X: 05.94 fps 2X: 04.39 fps 4X: 03.23 fps
Nyx 1X: 02.43 fps 2X: 01.31 fps
Nyx Fast 1X: 05.37 fps
Rhea 4X: 01.16 fps
4X Slowmo Apollo: 20.05 fps APFast: 64.94 fps Chronos: 06.04 fps CHFast: 08.82 fps
16X Slowmo Aion: 13.47 fps
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 50%) = Low Power Mode Enabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 0.5 instances: 0
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 17.96 fps 2X: 10.35 fps 4X: 03.39 fps
Iris 1X: 10.16 fps 2X: 04.77 fps 4X: 02.07 fps
Proteus 1X: 16.67 fps 2X: 09.97 fps 4X: 03.16 fps
Gaia 1X: 05.95 fps 2X: 04.40 fps 4X: 03.19 fps
Nyx 1X: 02.44 fps 2X: 01.29 fps
Nyx Fast 1X: 05.37 fps
Rhea 4X: 01.30 fps
4X Slowmo Apollo: 20.16 fps APFast: 68.06 fps Chronos: 04.52 fps CHFast: 08.95 fps
16X Slowmo Aion: 13.44 fps
Mainly GPU (Memory Usage 10%) = Low Power Mode Disabled (Sole Setting where RHEA heavily uses GPU)
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 0.1 instances: 1
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 11.72 fps 2X: 07.41 fps 4X: 02.87 fps
Iris 1X: 09.89 fps 2X: 05.64 fps 4X: 01.45 fps
Proteus 1X: 09.96 fps 2X: 06.59 fps 4X: 02.52 fps
Gaia 1X: 03.61 fps 2X: 02.76 fps 4X: 02.07 fps
Nyx 1X: 03.18 fps 2X: 02.02 fps
Nyx Fast 1X: 06.98 fps
Rhea 4X: 00.88 fps
4X Slowmo Apollo: 12.56 fps APFast: 47.98 fps Chronos: 05.02 fps CHFast: 07.47 fps
16X Slowmo Aion: 13.43 fps
Mainly CPU (Memory Usage 10%) = Low Power Mode Enabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 0.1 instances: 0
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 11.85 fps 2X: 07.38 fps 4X: 02.93 fps
Iris 1X: 08.91 fps 2X: 05.45 fps 4X: 01.46 fps
Proteus 1X: 08.71 fps 2X: 06.21 fps 4X: 02.48 fps
Gaia 1X: 03.68 fps 2X: 02.72 fps 4X: 02.01 fps
Nyx 1X: 02.53 fps 2X: 01.61 fps
Nyx Fast 1X: 05.99 fps
Rhea 4X: 00.82 fps
4X Slowmo Apollo: 12.29 fps APFast: 45.97 fps Chronos: 04.11 fps CHFast: 06.20 fps
16X Slowmo Aion: 13.31 fps
==> Baseline Benchmarks with my brothers base M4 Mac Mini with 16 GB RAM and 768x576 input source with various memory and power settings
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 100%) = Low Power Mode Disabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 26.99 fps 2X: 16.49 fps 4X: 05.45 fps
Iris 1X: 19.82 fps 2X: 12.04 fps 4X: 02.76 fps
Proteus 1X: 27.27 fps 2X: 20.25 fps 4X: 04.84 fps
Gaia 1X: 09.44 fps 2X: 07.29 fps 4X: 05.50 fps
Nyx 1X: 01.90 fps 2X: 00.76 fps
Nyx Fast 1X: 08.37 fps
Rhea 4X: 01.28 fps
4X Slowmo Apollo: 29.24 fps APFast: 38.91 fps Chronos: 10.52 fps CHFast: 15.27 fps
16X Slowmo Aion: 23.31 fps
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 100%) = Low Power Mode Enabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 27.03 fps 2X: 15.84 fps 4X: 05.37 fps
Iris 1X: 19.84 fps 2X: 12.03 fps 4X: 02.79 fps
Proteus 1X: 27.24 fps 2X: 20.27 fps 4X: 04.88 fps
Gaia 1X: 09.42 fps 2X: 07.29 fps 4X: 05.52 fps
Nyx 1X: 01.90 fps 2X: 00.76 fps
Nyx Fast 1X: 08.34 fps
Rhea 4X: 02.20 fps
4X Slowmo Apollo: 29.28 fps APFast: 39.57 fps Chronos: 10.53 fps CHFast: 15.32 fps
16X Slowmo Aion: 23.33 fps
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 75%) = Low Power Mode Disabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 0.75 instances: 1
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 26.93 fps 2X: 16.23 fps 4X: 05.83 fps
Iris 1X: 19.83 fps 2X: 12.02 fps 4X: 02.76 fps
Proteus 1X: 27.18 fps 2X: 20.25 fps 4X: 04.97 fps
Gaia 1X: 09.43 fps 2X: 07.33 fps 4X: 05.51 fps
Nyx 1X: 02.07 fps 2X: 00.87 fps
Nyx Fast 1X: 08.44 fps
Rhea 4X: 01.28 fps
4X Slowmo Apollo: 29.29 fps APFast: 39.64 fps Chronos: 10.49 fps CHFast: 15.33 fps
16X Slowmo Aion: 23.33 fps
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 75%) = Low Power Mode Enabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 0.75 instances: 0
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 26.96 fps 2X: 15.70 fps 4X: 05.76 fps
Iris 1X: 19.81 fps 2X: 12.04 fps 4X: 02.76 fps
Proteus 1X: 27.28 fps 2X: 20.29 fps 4X: 05.01 fps
Gaia 1X: 09.44 fps 2X: 07.31 fps 4X: 05.52 fps
Nyx 1X: 01.90 fps 2X: 00.76 fps
Nyx Fast 1X: 08.42 fps
Rhea 4X: 02.20 fps
4X Slowmo Apollo: 29.26 fps APFast: 39.45 fps Chronos: 10.51 fps CHFast: 15.33 fps
16X Slowmo Aion: 23.26 fps
(Memory Usage 50%) = Low Power Mode Disabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 0.5 instances: 1
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 26.08 fps 2X: 16.42 fps 4X: 05.88 fps
Iris 1X: 16.50 fps 2X: 08.96 fps 4X: 02.76 fps
Proteus 1X: 27.17 fps 2X: 20.21 fps 4X: 04.96 fps
Gaia 1X: 09.41 fps 2X: 07.26 fps 4X: 05.50 fps
Nyx 1X: 02.07 fps 2X: 00.86 fps
Nyx Fast 1X: 08.40 fps
Rhea 4X: 01.28 fps
4X Slowmo Apollo: 29.23 fps APFast: 39.67 fps Chronos: 09.02 fps CHFast: 15.36 fps
16X Slowmo Aion: 23.32 fps
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 50%) = Low Power Mode Enabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 0.5 instances: 0
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 26.99 fps 2X: 15.87 fps 4X: 05.49 fps
Iris 1X: 16.25 fps 2X: 08.94 fps 4X: 02.75 fps
Proteus 1X: 27.18 fps 2X: 20.19 fps 4X: 04.95 fps
Gaia 1X: 09.43 fps 2X: 07.30 fps 4X: 05.51 fps
Nyx 1X: 01.90 fps 2X: 00.76 fps
Nyx Fast 1X: 08.41 fps
Rhea 4X: 02.20 fps
4X Slowmo Apollo: 29.23 fps APFast: 39.67 fps Chronos: 07.57 fps CHFast: 15.29 fps
16X Slowmo Aion: 23.39 fps
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 25%) = Low Power Mode Disabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 0.25 instances: 1
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 24.24 fps 2X: 14.01 fps 4X: 05.50 fps
Iris 1X: 18.39 fps 2X: 10.49 fps 4X: 02.82 fps
Proteus 1X: 24.67 fps 2X: 14.97 fps 4X: 04.49 fps
Gaia 1X: 07.41 fps 2X: 05.46 fps 4X: 03.87 fps
Nyx 1X: 04.86 fps 2X: 05.92 fps
Nyx Fast 1X: 15.11 fps
Rhea 4X: 00.83 fps
4X Slowmo Apollo: 25.86 fps APFast: 31.90 fps Chronos: 09.26 fps CHFast: 13.47 fps
16X Slowmo Aion: 23.29 fps
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 25%) = Low Power Mode Enabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 0.25 instances: 0
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 20.28 fps 2X: 12.72 fps 4X: 04.88 fps
Iris 1X: 16.23 fps 2X: 07.99 fps 4X: 02.36 fps
Proteus 1X: 20.22 fps 2X: 13.19 fps 4X: 04.11 fps
Gaia 1X: 06.33 fps 2X: 04.87 fps 4X: 03.63 fps
Nyx 1X: 02.44 fps 2X: 01.07 fps
Nyx Fast 1X: 09.90 fps
Rhea 4X: 01.49 fps
4X Slowmo Apollo: 21.29 fps APFast: 27.30 fps Chronos: 07.08 fps CHFast: 09.92 fps
16X Slowmo Aion: 23.33 fps
Mainly GPU (Memory Usage 10%) = Low Power Mode Disabled (Sole Setting where RHEA heavily uses GPU)
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 0.1 instances: 1
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 25.10 fps 2X: 15.37 fps 4X: 05.95 fps
Iris 1X: 19.38 fps 2X: 10.02 fps 4X: 02.89 fps
Proteus 1X: 24.14 fps 2X: 16.16 fps 4X: 04.85 fps
Gaia 1X: 07.30 fps 2X: 05.58 fps 4X: 04.08 fps
Nyx 1X: 04.85 fps 2X: 05.99 fps
Nyx Fast 1X: 15.10 fps
Rhea 4X: 00.79 fps
4X Slowmo Apollo: 22.79 fps APFast: 86.80 fps Chronos: 09.61 fps CHFast: 14.02 fps
16X Slowmo Aion: 23.23 fps
Mainly CPU (Memory Usage 10%) = Low Power Mode Enabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 0.1 instances: 0
Input Resolution: 768x576
Benchmark Results
Artemis 1X: 19.96 fps 2X: 12.21 fps 4X: 05.05 fps
Iris 1X: 14.91 fps 2X: 07.48 fps 4X: 02.33 fps
Proteus 1X: 19.29 fps 2X: 12.49 fps 4X: 04.11 fps
Gaia 1X: 05.75 fps 2X: 04.50 fps 4X: 03.52 fps
Nyx 1X: 02.44 fps 2X: 01.07 fps
Nyx Fast 1X: 09.81 fps
Rhea 4X: 01.43 fps
4X Slowmo Apollo: 18.76 fps APFast: 75.15 fps Chronos: 07.25 fps CHFast: 10.68 fps
16X Slowmo Aion: 23.38 fps
==> And just for fun some baseline Benchmarks with my brothers base M4 Mac Mini and my old M1 Mac Mini with 16 GB RAM and 1920x1080 input source with various memory and power settings
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 100%) = Low Power Mode Disabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 07.19 fps 2X: 04.40 fps 4X: 01.50 fps
Iris 1X: 05.23 fps 2X: 02.97 fps 4X: 00.80 fps
Proteus 1X: 07.41 fps 2X: 04.98 fps 4X: 01.33 fps
Gaia 1X: 02.29 fps 2X: 01.74 fps 4X: 01.20 fps
Nyx 1X: 01.30 fps 2X: 00.95 fps
Nyx Fast 1X: 04.37 fps
Rhea 4X: 00.27 fps
4X Slowmo Apollo: 08.13 fps APFast: 09.69 fps Chronos: 02.48 fps CHFast: 03.98 fps
16X Slowmo Aion: 05.13 fps
Mainly Neural/CPU / Model dependent also little GPU (Memory Usage 100%) = Low Power Mode Enabled
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M4 16 GB
GPU: Apple M4 10.667 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 06.39 fps 2X: 03.88 fps 4X: 01.53 fps
Iris 1X: 03.93 fps 2X: 02.23 fps 4X: 00.74 fps
Proteus 1X: 06.43 fps 2X: 04.18 fps 4X: 01.28 fps
Gaia 1X: 01.94 fps 2X: 01.48 fps 4X: 01.08 fps
Nyx 1X: 00.63 fps 2X: 00.25 fps
Nyx Fast 1X: 02.80 fps
Rhea 4X: 00.45 fps
4X Slowmo Apollo: 06.67 fps APFast: 09.45 fps Chronos: 01.95 fps CHFast: 03.10 fps
16X Slowmo Aion: 05.12 fps
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 03.70 fps 2X: 02.30 fps 4X: 00.83 fps
Iris 1X: 02.57 fps 2X: 01.21 fps 4X: 00.36 fps
Proteus 1X: 03.43 fps 2X: 01.99 fps 4X: 00.64 fps
Gaia 1X: 01.23 fps 2X: 00.87 fps 4X: 00.57 fps
Nyx 1X: 00.79 fps 2X: 00.44 fps
Nyx Fast 1X: 01.77 fps
Rhea 4X: 00.23 fps
4X Slowmo Apollo: 04.32 fps APFast: 16.61 fps Chronos: 01.23 fps CHFast: 01.77 fps
16X Slowmo Aion: 02.95 fps
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 03.69 fps 2X: 02.30 fps 4X: 00.83 fps
Iris 1X: 02.73 fps 2X: 01.37 fps 4X: 00.39 fps
Proteus 1X: 03.42 fps 2X: 02.06 fps 4X: 00.64 fps
Gaia 1X: 01.23 fps 2X: 00.87 fps 4X: 00.62 fps
Nyx 1X: 00.86 fps 2X: 00.42 fps
Nyx Fast 1X: 01.84 fps
Rhea 4X: 00.26 fps
4X Slowmo Apollo: 04.33 fps APFast: 15.54 fps Chronos: 01.31 fps CHFast: 01.97 fps
16X Slowmo Aion: 02.94 fps
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 0.5 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 03.30 fps 2X: 02.10 fps 4X: 00.81 fps
Iris 1X: 02.26 fps 2X: 01.22 fps 4X: 00.39 fps
Proteus 1X: 03.01 fps 2X: 01.86 fps 4X: 00.69 fps
Gaia 1X: 01.08 fps 2X: 00.80 fps 4X: 00.58 fps
Nyx 1X: 00.79 fps 2X: 00.45 fps
Nyx Fast 1X: 01.65 fps
Rhea 4X: 00.24 fps
4X Slowmo Apollo: 03.97 fps APFast: 14.87 fps Chronos: 01.23 fps CHFast: 01.63 fps
16X Slowmo Aion: 02.99 fps
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 0.5 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 03.27 fps 2X: 02.10 fps 4X: 00.79 fps
Iris 1X: 02.37 fps 2X: 01.36 fps 4X: 00.40 fps
Proteus 1X: 03.00 fps 2X: 01.92 fps 4X: 00.70 fps
Gaia 1X: 01.09 fps 2X: 00.80 fps 4X: 00.57 fps
Nyx 1X: 00.87 fps 2X: 00.41 fps
Nyx Fast 1X: 01.84 fps
Rhea 4X: 00.26 fps
4X Slowmo Apollo: 03.94 fps APFast: 13.94 fps Chronos: 01.30 fps CHFast: 01.86 fps
16X Slowmo Aion: 02.89 fps
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 0.1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 02.87 fps 2X: 01.79 fps 4X: 00.69 fps
Iris 1X: 02.27 fps 2X: 01.33 fps 4X: 00.35 fps
Proteus 1X: 02.45 fps 2X: 01.54 fps 4X: 00.60 fps
Gaia 1X: 00.92 fps 2X: 00.68 fps 4X: 00.51 fps
Nyx 1X: 00.57 fps 2X: 00.39 fps
Nyx Fast 1X: 01.38 fps
Rhea 4X: 00.21 fps
4X Slowmo Apollo: 03.24 fps APFast: 13.72 fps Chronos: 01.06 fps CHFast: 01.52 fps
16X Slowmo Aion: 02.95 fps
Topaz Video AI v5.4.0
System Information
OS: Mac v15.01
CPU: Apple M1 16 GB
GPU: Apple M1 10.667 GB
Processing Settings
device: 0 vram: 0.1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 02.88 fps 2X: 01.79 fps 4X: 00.69 fps
Iris 1X: 02.50 fps 2X: 01.37 fps 4X: 00.35 fps
Proteus 1X: 02.49 fps 2X: 01.60 fps 4X: 00.60 fps
Gaia 1X: 00.93 fps 2X: 00.68 fps 4X: 00.51 fps
Nyx 1X: 00.65 fps 2X: 00.41 fps
Nyx Fast 1X: 01.43 fps
Rhea 4X: 00.22 fps
4X Slowmo Apollo: 03.24 fps APFast: 13.74 fps Chronos: 01.24 fps CHFast: 01.82 fps
16X Slowmo Aion: 02.90 fps
…and yes, I had that evening obviously too much time and was super bored…so I did those benchmarks to help me decide, which M4 I’m going to buy…
2 Likes
If you want silence, which is something I really appreciate, then you will want to make use of the Neural Engine as much as possible. In full power mode, my Mac Mini M4 Pro is very noisy and very annoying when the GPU is at least 70% sustained usage. I’m almost forced to use noise cancelling earphones. I think the key is to force as much as possible into the Neural Engine. I have found the optimum for silent processing is to use 2 concurrent processes in low power mode. More than 2 doesn’t give any more gain.
It would be really useful to have some feedback regarding the fan noise of a M4 Max MacBook Pro when the GPU cores are in sustained use running TVAI upscaling. Given how loud my M4 Pro mini gets, I can only assume a M4 Max MB Pro with even more GPU cores would be at least as noisy.
I would seriously consider the performance / cost ratio of buying 2 or more base spec M4 minis with extra TVAI licence(s) as needed (1 licence per 2 machines), against the laptop - which you may well have to run in low power mode due to the noise, therefore reducing the benefit of the extra GPU cores.
Andy
EDIT: Correction - regarding 2 concurrent processes, that should be “at 100% max memory”, NOT “in low power mode”. This applies to upscaling SD using Artemis or Proteus, which will use just the Neural Engine alone if the Max Memory is at 100%.
Topaz Video AI v5.4.0
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: 10.49 fps 2X: 07.37 fps 4X: 02.27 fps
Iris 1X: 10.63 fps 2X: 06.29 fps 4X: 01.75 fps
Proteus 1X: 10.15 fps 2X: 07.17 fps 4X: 02.34 fps
Gaia 1X: 03.38 fps 2X: 02.27 fps 4X: 01.54 fps
Nyx 1X: 04.07 fps 2X: 03.54 fps
Nyx Fast 1X: 07.44 fps
Rhea 4X: 01.34 fps
4X Slowmo Apollo: 15.10 fps APFast: 40.01 fps Chronos: 08.06 fps CHFast: 12.43 fps
16X Slowmo Aion: 06.18 fps
Very useful as I’m facing the same decision!
Topaz Video AI v5.5.0
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: 10.37 fps 2X: 07.37 fps 4X: 02.29 fps
Iris 1X: 10.84 fps 2X: 06.12 fps 4X: 01.77 fps
Proteus 1X: 10.13 fps 2X: 06.97 fps 4X: 02.45 fps
Gaia 1X: 03.41 fps 2X: 02.33 fps 4X: 01.56 fps
Nyx 1X: 04.13 fps 2X: 03.53 fps
Nyx Fast 1X: 07.53 fps
Rhea 4X: 01.36 fps
4X Slowmo Apollo: 15.37 fps APFast: 41.01 fps Chronos: 08.17 fps CHFast: 12.55 fps
16X Slowmo Aion: 05.77 fps
M4 Max not as fast as I expected !
Topaz Video AI v5.5.0
System Information
OS: Mac v15.0101
CPU: Apple M4 Max 36 GB
GPU: Apple M4 Max 27 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 12.53 fps 2X: 07.24 fps 4X: 02.65 fps
Iris 1X: 09.97 fps 2X: 04.73 fps 4X: 01.74 fps
Proteus 1X: 13.27 fps 2X: 08.15 fps 4X: 02.64 fps
Gaia 1X: 03.96 fps 2X: 02.89 fps 4X: 02.10 fps
Nyx 1X: 03.96 fps 2X: 04.62 fps
Nyx Fast 1X: 09.67 fps
Rhea 4X: 00.86 fps
4X Slowmo Apollo: 14.05 fps APFast: 28.91 fps Chronos: 04.75 fps CHFast: 08.22 fps
16X Slowmo Aion: 13.53 fps
Very Interesting Using M4 Max.
The Real power of the processor is in Multi-Process
Could you try a real Test with 2/3/4 processes?
For a single process SET “MAX MEMORY” TO 10
But for multi-process better let in 100
So in My M4 Pro (12c 16c) results are much better in multi process.
1 - 16 fps*
2 - 20 fps* (in sum)
3 - 22 fps* (in sum)
Convert 720x576 to 1080 with Proteus V3.
As other important info:
1 process - If Set to MAX MEMORY 10 the processor gets VERY HOT
But in SET MAX MEMORY 10 the processor is a good temperature. It’s Worth.
In MultiProces the temperature always gets well with MAX MEMORY 100
But M4MAX can be very different. How many Cores it has?
Just Let us know.
Thanks.
Topaz Video AI v5.5.0
System Information
OS: Windows v11.23
CPU: AMD Ryzen 7 5800X3D 8-Core Processor 63.928 GB
GPU: NVIDIA TITAN RTX 23.687 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 17.82 fps 2X: 08.50 fps 4X: 02.49 fps
Iris 1X: 16.29 fps 2X: 09.05 fps 4X: 02.32 fps
Proteus 1X: 14.31 fps 2X: 11.52 fps 4X: 02.88 fps
Gaia 1X: 04.17 fps 2X: 02.68 fps 4X: 01.75 fps
Nyx 1X: 04.67 fps 2X: 03.78 fps
Nyx Fast 1X: 10.71 fps
Rhea 4X: 01.46 fps
4X Slowmo Apollo: 06.50 fps APFast: 46.36 fps Chronos: 09.70 fps CHFast: 17.73 fps
16X Slowmo Aion: 19.69 fps
