Well it could be very easy. Just detect the gfx chip and download separate models for the 50x0 but keep the existing ones for older NVIDIA GPUs.
I asked if it would be esier to do just that, but didnāt get an answer.
1 Like
Here is a video on Youtube about benchmarks and various Mac configurations
2 Likes
Well, i bought a second 5070 for my computer i use for video editing and upscaling and i havent opened it yet. Its looking like im going to return it and get a 9070xt
4 Likes
I have a 5080 I got a few weeks ago. I put it in for a few hours, was unimpressed with what it can do in games as wells as TVAI beta, then I put it back in the box. Iāll put it in the computer Iām building for a friend, who only plays new games.
2 Likes
5090OC, 9950X3D CCD0 no HT, 6400CL30 RAM
Topaz Video AI v6.2.0
System Information
OS: Windows v10.22
CPU: AMD Ryzen 9 9950X3D 16-Core Processor 31.084 GB
GPU: NVIDIA GeForce RTX 5090 31.349 GB
GPU: AMD Radeon(TM) Graphics 0.47438 GB
Processing Settings
device: 0 vram: 0.8 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 57.36 fps 2X: 24.71 fps 4X: 05.16 fps
Iris 1X: 50.87 fps 2X: 23.87 fps 4X: 06.13 fps
Proteus 1X: 58.21 fps 2X: 25.18 fps 4X: 06.17 fps
Gaia 1X: 19.11 fps 2X: 14.01 fps 4X: 05.92 fps
Nyx 1X: 17.41 fps 2X: 11.65 fps
Nyx Fast 1X: 42.43 fps
Rhea 4X: 05.59 fps
RXL 4X: 05.59 fps
Hyperion HDR 1X: 45.40 fps
4X Slowmo Apollo: 60.06 fps APFast: 102.24 fps Chronos: 48.00 fps CHFast: 59.63 fps
16X Slowmo Aion: 45.17 fps
2 Likes
Can you guys please share your experience running 2 or more identical cards on the same system?
Hmmā¦ not out the woods just yet, though; until the awful branch 572.xx has finally been (editā¦ ughā¦ sunset). 575 branch shows much better results in inference workloads across the board (well, not ~next weekās 1st deployment 575 driver, but later May/June 575 branch drivers) and MS upcoming thingyā¦ just hold on, folks! Thenā¦ itās Topazās turn to capitalize on all of this upcoming greatness!
Will the RTX 4000 cards also benefit from this?
So Iām not really impressed with Rheaās results with the RTX5000.
Topaz Video AI v6.1.3 572.83
System Information
OS: Windows v11.24
CPU: AMD Ryzen 7 9800X3D 8-Core Processor 31.654 GB
GPU: NVIDIA GeForce RTX 4090 23.576 GB
Processing Settings
device: 0 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 43.55 fps 2X: 19.99 fps 4X: 05.29 fps
Iris 1X: 39.36 fps 2X: 23.85 fps 4X: 06.17 fps
Proteus 1X: 43.00 fps 2X: 22.93 fps 4X: 06.11 fps
Gaia 1X: 16.41 fps 2X: 11.24 fps 4X: 05.89 fps
Nyx 1X: 17.93 fps 2X: 14.07 fps
Nyx Fast 1X: 32.49 fps
Rhea 4X: 05.78 fps
RXL 4X: 05.21 fps
Hyperion HDR 1X: 32.41 fps
4X Slowmo Apollo: 49.15 fps APFast: 102.32 fps Chronos: 32.53 fps CHFast: 43.02 fps
16X Slowmo Aion: 31.98 fps
Topaz Video AI v6.2.0
System Information
OS: Windows v10.22
CPU: AMD Ryzen Threadripper 3970X 32-Core Processor 127.87 GB
GPU: NVIDIA GeForce RTX 3090 23.756 GB
Processing Settings
device: -2 vram: 1 instances: 0
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 24.52 fps 2X: 11.58 fps 4X: 03.57 fps
Iris 1X: 23.13 fps 2X: 13.07 fps 4X: 03.78 fps
Proteus 1X: 23.54 fps 2X: 14.57 fps 4X: 03.77 fps
Gaia 1X: 09.01 fps 2X: 06.30 fps 4X: 03.33 fps
Nyx 1X: 08.22 fps 2X: 06.33 fps
Nyx Fast 1X: 20.86 fps
Rhea 4X: 03.08 fps
RXL 4X: 03.00 fps
Hyperion HDR 1X: 20.91 fps
4X Slowmo Apollo: 34.02 fps APFast: 58.48 fps Chronos: 20.03 fps CHFast: 29.71 fps
16X Slowmo Aion: 24.76 fps
From NVIDIAās and MS upcoming software stacksā¦ yes, sir (or maāam), on all RTXās cardsā¦ all RTXs tests have shown incredible performance uplifts in inference/creative workloads (especially May/June deployment stacks)!
Will devs seize on the opportunity, though, wellā¦
The latest Studio driver from Nvidia, and optimization from the Nvidia application.
Topaz Video AI v6.1.3
System Information
OS: Windows v10.22
CPU: AMD Ryzen 7 6800H with Radeon Graphics 63.192 GB
GPU: NVIDIA GeForce RTX 3060 Laptop GPU 5.8555 GB
GPU: AMD Radeon(TM) Graphics 0.4744 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 09.66 fps 2X: 07.27 fps 4X: 02.44 fps
Iris 1X: 09.58 fps 2X: 05.87 fps 4X: 01.82 fps
Proteus 1X: 09.32 fps 2X: 06.96 fps 4X: 02.45 fps
Gaia 1X: 02.99 fps 2X: 02.09 fps 4X: 01.41 fps
Nyx 1X: 03.64 fps 2X: 03.22 fps
Nyx Fast 1X: 07.38 fps
Rhea 4X: 01.23 fps
RXL 4X: 01.16 fps
Hyperion HDR 1X: 23.91 fps
4X Slowmo Apollo: 13.47 fps APFast: 37.09 fps Chronos: 07.35 fps CHFast: 12.04 fps
16X Slowmo Aion: 19.64 fps
Ofc, because Rhea is benched in 4X mode and any 2X or 4X mode wil be limited by CPU/RAM due to the image scaling performed by CPU/RAM.
Only 1X benches give a good base for comparing GPUs. If u donāt do 1X workloads, high end GPUs are a waste of money with Topas
But using small source material and an old model doing deinterlacing! (Dione)
I mean the 50 series looks like its great for NVencoding. I just reencoded a 57gb video I upscaled to 4k with my 5070, the output video file is only 7.6gb.
I ran it CQ: 21, Speed: Slower. It was getting around 70fps.
Thatās pretty impressive
1 Like
From my experience, AMD encoder gives higher quality than NVIDIAs. I havenāt checked the latest results from my 5090 but before up to the 4090, I frequently had banding issues in sensitive material which couldnāt be resolved through higher bitrate setting. So I am actually using the iGPU of my AMD CPU to encode the videos instead of the NVIDIA encoder.
I just returned my 5070 and got a 9070xt. Weāll see if those numbers are for real soon
2 Likes
Topaz Video AI v6.2.0
System Information
OS: Windows v10.22
CPU: AMD Ryzen 7 5700G with Radeon Graphics 31.798 GB
GPU: AMD Radeon RX 9070 XT 15.417 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 23.72 fps 2X: 08.07 fps 4X: 02.27 fps
Iris 1X: 27.06 fps 2X: 11.43 fps 4X: 03.20 fps
Proteus 1X: 30.08 fps 2X: 11.65 fps 4X: 03.41 fps
Gaia 1X: 16.07 fps 2X: 09.34 fps 4X: 02.98 fps
Nyx 1X: 11.61 fps 2X: 08.78 fps
Nyx Fast 1X: 23.38 fps
Rhea 4X: 02.59 fps
RXL 4X: 02.61 fps
Hyperion HDR 1X: 25.62 fps
4X Slowmo Apollo: 28.62 fps APFast: 48.88 fps Chronos: 18.85 fps CHFast: 26.63 fps
16X Slowmo Aion: 43.54 fps
Well Iāll be darnedā¦ Obviously a lot of models utilize the CPU more than others. Thatās why most of my numbers are lower
This is pretty crazy.
