I took the liberty to put your results into Claude to generate some charts. Hope u don’t mind. I am just having fun exploring what the AI can do 
And a summary of your findings as understood by Claude:
1 Like
I created a comparison between 5080 (9900X CPU) and 5090 (7950X CPU). Both with 6200CL30 RAM.
1 Like
Neat. Looks like i low-balling the gains!
i made good prediction for this.
Around 30% for tensorrt.
Ya, i saw that comment. I was assuming the same. 30% +/- 5% or so.
What i didn’t guess at was that the increased render speed wouldn’t require max power draw. Which is around 80% (can get the rare transient spike up to ~85%) since my 3080 wanted all power it could get. So that’s a very welcome bonus. At 80%, the card won’t need to run fans at more than some 50%.
Check @topaz257 claude summaries. It’s actually more than 30% average increase. The links seem to be broken atm but iirc it was around 33%. Ranging from %22 to 83% increase in render speed.
1 Like
Interesting. The Claude pages throw an ‘internal server error 500’. First time i used that feature. Seems like a problem by the Claude servers. Let’s see whether they come back online
Update: Content is back. Seems like Anthropic fixed it
Topaz Video AI v6.1.2
System Information
OS: Windows v11.24
CPU: AMD Ryzen 9 7950X3D 16-Core Processor 63.615 GB
GPU: AMD Radeon RX 9070 XT 15.374 GB
Processing Settings
device: 0 vram: 0.99 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 35.32 fps 2X: 15.88 fps 4X: 04.40 fps
Iris 1X: 42.69 fps 2X: 19.51 fps 4X: 05.13 fps
Proteus 1X: 37.95 fps 2X: 19.49 fps 4X: 06.04 fps
Gaia 1X: 19.80 fps 2X: 12.42 fps 4X: 05.34 fps
Nyx 1X: 16.10 fps 2X: 12.77 fps
Nyx Fast 1X: 32.24 fps
Rhea 4X: 04.14 fps
RXL 4X: 04.01 fps
Hyperion HDR 1X: 29.17 fps
4X Slowmo Apollo: 35.71 fps APFast: 70.88 fps Chronos: 22.61 fps CHFast: 35.56 fps
16X Slowmo Aion: 53.67 fps
Topaz Video AI v6.1.2
System Information
OS: Windows v11.24
CPU: AMD Ryzen 9 7950X3D 16-Core Processor 63.615 GB
GPU: AMD Radeon RX 9070 XT 15.374 GB
Processing Settings
device: 0 vram: 0.99 instances: 1
Input Resolution: 720x480
Benchmark Results
Artemis 1X: 176.48 fps 2X: 83.47 fps 4X: 20.04 fps
Iris 1X: 313.75 fps 2X: 83.75 fps 4X: 31.49 fps
Proteus 1X: 253.79 fps 2X: 118.35 fps 4X: 32.19 fps
Gaia 1X: 108.40 fps 2X: 66.45 fps 4X: 27.37 fps
Nyx 1X: 45.90 fps 2X: 44.11 fps
Nyx Fast 1X: 128.03 fps
Rhea 4X: 18.91 fps
RXL 4X: 17.44 fps
Hyperion HDR 1X: 105.19 fps
4X Slowmo Apollo: 198.01 fps APFast: 465.64 fps Chronos: 86.80 fps CHFast: 171.98 fps
16X Slowmo Aion: 153.61 fps
Not bad for the 9070 XT… It’s closer to a 5080 than I would have guessed possible! I know Gaia isn’t really used much anymore but the 9070 XT kills it in Gaia.
I bet if Topaz actually spent some time to optimize for the 9070 XT they could get some good improvements!
4 Likes
My Hellhound 9070 XT does not come anywhere near these figures 
GPU is at 40-50% utilization only, CPU at 70%, card is 45C cold.
This is what I have. Can you please share your secret?
Topaz Video AI v6.1.3
System Information
OS: Windows v11.24
CPU: 13th Gen Intel(R) Core(TM) i9-13900KF 127.83 GB
GPU: AMD Radeon RX 9070 XT 15.374 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 22.63 fps 2X: 08.57 fps 4X: 02.32 fps
Iris 1X: 29.60 fps 2X: 11.30 fps 4X: 03.12 fps
Proteus 1X: 28.31 fps 2X: 12.41 fps 4X: 03.16 fps
Gaia 1X: 15.42 fps 2X: 09.40 fps 4X: 01.08 fps
Nyx 1X: 13.12 fps 2X: 09.92 fps
Nyx Fast 1X: 22.89 fps
Rhea 4X: 00.29 fps
RXL 4X: 01.03 fps
Hyperion HDR 1X: 16.81 fps
4X Slowmo Apollo: 24.44 fps APFast: 41.99 fps Chronos: 20.52 fps CHFast: 21.91 fps
16X Slowmo Aion: 41.68 fps
1 Like
128G ？open your XMP?
Just something to try in case it makes a difference - in TVAI settings, change the AI Processor from “Auto” to the 9070…
Andy
1 Like
Here are the settings I have in Adrenalin.
Otherwise, I’d suggest checking the basics first. RAM is a big one for Video AI. I’m assuming your running at full RAM speeds and have made sure to check XMP or memory latency with something like Aida64.
I did notice that with my 7950X3D. I get better performance by using process lasso to pin FFMPEG to my frequency cores not the cache die. I don’t know if you can do the same with FFMPEG pinned to your performance cores. Video AI is allowed to use all of my cores but doesn’t really use much CPU.
Most of the time my GPU is around 50-70% utilization during the benchmark. Except for Rhea and RXL, those will push it to 90% or higher.
If you play a game does it fully utilize your GPU?
Also, I had did do a fresh install into a new\different folder and it performed better.
Let me know what you find.
1 Like
Copied your GPU settings to no visible result
XMP profile is on, RAM is DDR4 3200 C16
PCIE x16 is set to Gen5
Using fresh Windows install
STALKER2 runs at impressive FPS, up to 100% GPU utilization
Utilization shown is Proteus 1080 2x upscale of real video.
The FPS with the real video 2x upscaling is significantly less than 6900XT on the same system, about 5.5 vs. 6.8
I don’t have experience running 13th gen on DDR4, but I would expect it to be less than ideal for TVAI. (vs fast/low-latency DDR5)
1 Like
ddr4 3200 too slow
1 Like
EDIT: I re-read your post and see you are using Proteus.
What model are you running for the 1080p 2x upscale? What version of Windows are you running?
I know 24H2 provided some significant performance increases… Here is a video that talks about 24h2 and includes some Intel testing for gaming.
I would also suggest doing some testing with disabling CPU virtualization in BIOS. I don’t recall what Intel calls it. This will disable something call Core Isolation in Windows. You can google for a detailed explanation. The short is it is related to security. Windows creates AV type of services in a virtual sandbox. I disable AMD’s virtualization feature in BIOS. I practice safe browsing, don’t download bad shit or link on shady links, and I have Bit Defender AV.
Something isn’t right for sure… I noticed your CPU utilization is higher than mine but that might be because I lock FFMPEG to 8 cores on my side.
Intel can go faster but it shouldn’t be slower than his old GPU.
9 with proteus 2x is the least I’d expect.
My windows is 24h2. I tried to assign the performance cores only to FFMPEG, but to no effect. I also believe FFMPEG is doing frame-serving and maybe video encoding, not any AI functions, so did not expect much anyway.
Will test disabling virtualization later today, but honestly hopes are low for this.
I come closer to a conclusion that my system RAM is indeed too slow. I saw some 9070XT results in another thread here, they vary, but none are as low as mine. It is going to be a painful upgrade to DDR5 considering how much of it I need (for other tasks) plus a new MB
1 Like
Topaz Video AI v6.1.3
System Information
OS: Windows v11.24
CPU: 13th Gen Intel(R) Core(TM) i9-13900K 63.768 GB
GPU: NVIDIA GeForce RTX 5090 31.349 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: 34.35 fps 2X: 13.22 fps 4X: 03.23 fps
Iris 1X: 34.16 fps 2X: 17.13 fps 4X: 04.37 fps
Proteus 1X: 40.13 fps 2X: 17.65 fps 4X: 04.59 fps
Gaia 1X: 19.92 fps 2X: 13.24 fps 4X: 04.34 fps
Nyx 1X: 16.24 fps 2X: 12.39 fps
Nyx Fast 1X: 31.86 fps
Rhea 4X: 03.82 fps
RXL 4X: 03.97 fps
Hyperion HDR 1X: 28.96 fps
4X Slowmo Apollo: 30.36 fps APFast: 62.19 fps Chronos: 22.78 fps CHFast: 32.08 fps
16X Slowmo Aion: 34.53 fps
Topaz Video AI Alpha v6.1.1.1.a.trt
System Information
OS: Windows v11.24
CPU: 13th Gen Intel(R) Core(TM) i9-13900K 63.768 GB
GPU: NVIDIA GeForce RTX 5090 31.349 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: 40.02 fps 2X: 17.11 fps 4X: 04.44 fps
Iris 1X: 39.84 fps 2X: 19.84 fps 4X: 04.86 fps
Proteus 1X: 51.06 fps 2X: 20.81 fps 4X: 04.86 fps
Gaia 1X: 18.69 fps 2X: 13.32 fps 4X: 04.51 fps
Nyx 1X: 16.75 fps 2X: 11.81 fps
Nyx Fast 1X: 37.31 fps
Rhea 4X: 04.11 fps
RXL 4X: 04.41 fps
Hyperion HDR 1X: 29.16 fps
4X Slowmo Apollo: 45.11 fps APFast: 63.99 fps Chronos: 45.06 fps CHFast: 39.72 fps
16X Slowmo Aion: ... fps
