Mine runs at 2400MHz. 
But the quite huge amount (256GB) actually matters more for some future highly detailed photogrammetry projects of mine than the speedā¦ And the HP Z840 system is super-stable, so that matters a lot also. 
For me 1.1.1 runs a bit faster than 1.1.0 with all models in the benchmark. With Starlight Mini I get a little bit faster testults with 1.1.1 also (1.1.0 on top, 1.1.1 below):
My system is also PCIe 3 and the RAM even running slower, but it still runs a bit faster than his. You think 5070 Ti runs worse that 4070 Ti Super on a PCIe 3 system, because of the bigger generational gap?
Hi,
āDo you think the 5070 Ti will perform worse than the 4070 Ti Super on a PCIe 3 system because the generation gap is bigger?ā
First off, Iām not a developer.
But no, I donāt think so. Emphasis on āthink.ā
When I switched from RTX4070Super to RTX5080, the gain was 1:1.7.
(same task from 5 fps to 8.5 fps, both PCIe4)
I think PCIe3 could be sufficient for RTX40xx, with the possible exception of the 4090?
Thereās a benchmark here for a 5070 Ti (PCIe4 or 5 ?) that comes damn close, I think within 10%, to my 5080. That would probably give it the best price-performance ratio.
The overall package always has to be right!
Fastest possible RAM and PCIe.
The computing power of the CPU, on the other hand, seems pretty insignificant*, but it must have a fast RAM controller and support modern PCIe.
As you can see in the neighboring benchmark thread, my 11900 is just as fast as a 12900, 14900, 265, or 285ā¦
(Iām not arguing about 2 or 3 percent).
kind regards
Look at this thread:
The 5x00 series is said to cause unusually high traffic on the PCI bus - so the PCIe3 really might make a difference for the worse here.
ok, AI say:
Subject: Clarification on High PCIe Bus Usage Reporting for RTX 50-series GPUs
Regarding the concerns about unusually high PCIe bus usage on the new NVIDIA 50-series (Blackwell) cards:
NVIDIA representatives have confirmed that this is a reporting error rather than a physical performance issue.
The discrepancy stems from the fact that the monitoring API used by many third-party tools (like HWInfo or MSI Afterburner) had not yet been updated for the specific architecture of the 5000-series at launch.
As a result, these tools are misinterpreting the data flow, showing high bus saturation even when the card is idle or under light load.
NVIDIA has stated that there is no actual excessive data traffic on the PCI bus and performance in applications like Topaz Photo AI or Video AI is not negatively affected by this bug.
Recommendation: Make sure to update to the latest NVIDIA Game Ready/Studio Drivers and use the most recent versions of your monitoring software, as updates are being rolled out to address this API reporting glitch.
Dr. Marlboro has stated that smoking has no negative impact on peopleās healthā¦ 
SCNR
Another machine, now Ryzen 7 5700x with 3600 RAM and RTX 5080.
Topaz Video v1.1.0
System Information
OS: Windows v10.22
CPU: AMD Ryzen 7 5700X 8-Core Processor 63.928 GB
GPU: NVIDIA GeForce RTX 5080 15.604 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 22.87 fps 2X: 10.52 fps 4X: 02.26 fps
Iris 1X: 33.87 fps 2X: 12.02 fps 4X: 03.19 fps
Proteus 1X: 34.96 fps 2X: 11.97 fps 4X: 03.37 fps
Gaia 1X: 12.76 fps 2X: 08.72 fps 4X: 02.90 fps
Nyx 1X: 17.23 fps 2X: 13.60 fps
Nyx Fast 1X: 31.78 fps
Nyx XL 1X: 18.03 fps
Rhea 4X: 02.83 fps
RXL 4X: 02.84 fps
Hyperion HDR 1X: 12.61 fps
4X Slowmo Apollo: 29.15 fps APFast: 62.96 fps Chronos: 25.79 fps CHFast: 26.95 fps
16X Slowmo Aion: 45.83 fps
```
Topaz Video v1.1.1
System Information
OS: Windows v11.25
CPU: Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz 223.4 GB
GPU: NVIDIA GeForce RTX 3060 11.831 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 09.33 fps 2X: 06.63 fps 4X: 02.43 fps
Iris 1X: 10.91 fps 2X: 06.32 fps 4X: 02.12 fps
Proteus 1X: 09.53 fps 2X: 07.13 fps 4X: 02.66 fps
Gaia 1X: 03.17 fps 2X: 02.15 fps 4X: 01.61 fps
Nyx 1X: 04.63 fps 2X: 03.99 fps
Nyx Fast 1X: 09.97 fps
Nyx XL 1X: 18.93 fps
Rhea 4X: 01.60 fps
RXL 4X: 01.60 fps
Hyperion HDR 1X: 12.47 fps
4X Slowmo Apollo: 12.37 fps APFast: 42.64 fps Chronos: 06.68 fps CHFast: 10.51 fps
16X Slowmo Aion: 18.52 fps
```
The Proteus Natural in 1.2.0 is only 44% as fast as Proteus. Nyx XL and 4x Slomo took some big performance hits as well.
```
Topaz Video v1.2.0
System Information
OS: Windows v11.25
CPU: AMD Ryzen 7 9800X3D 8-Core Processor 63.556 GB
GPU: NVIDIA GeForce RTX 5070 Ti 15.62 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 30.53 fps 2X: 16.35 fps 4X: 04.44 fps
Iris 1X: 34.19 fps 2X: 19.39 fps 4X: 04.78 fps
Proteus 1X: 31.83 fps 2X: 19.26 fps 4X: 04.96 fps
Proteus Natural 2X: 08.53 fps
Gaia 1X: 09.41 fps 2X: 06.62 fps 4X: 04.72 fps
Nyx 1X: 13.70 fps 2X: 11.80 fps
Nyx Fast 1X: 33.80 fps
Nyx XL 1X: 02.17 fps
Rhea 4X: 04.46 fps
RXL 4X: 04.29 fps
Hyperion HDR 1X: 34.78 fps
4X Slowmo Apollo: 25.08 fps APFast: 36.90 fps Chronos: 22.13 fps CHFast: 23.20 fps
16X Slowmo Aion: ERR fps
```
Iā¦ I donāt understand whatās happening, but 1.2.0 is much slower with any model than 1.1.0
Topaz Video v1.2.0
System Information
OS: Windows v10.22
CPU: Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz 31.834 GB
GPU: NVIDIA GeForce RTX 5070 Ti 15.534 GB
GPU: Intel(R) UHD Graphics 630 0.125 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 10.39 fps 2X: 05.06 fps 4X: 01.54 fps
Iris 1X: 16.37 fps 2X: 08.17 fps 4X: 02.49 fps
Proteus 1X: 16.09 fps 2X: 08.51 fps 4X: 02.57 fps
Proteus Natural 2X: 06.01 fps
Gaia 1X: 06.60 fps 2X: 04.59 fps 4X: 02.37 fps
Nyx 1X: 09.75 fps 2X: 06.15 fps
Nyx Fast 1X: 16.95 fps
Nyx XL 1X: 02.14 fps
Rhea 4X: 02.20 fps
RXL 4X: 02.13 fps
Hyperion HDR 1X: 08.83 fps
4X Slowmo Apollo: 18.82 fps APFast: 24.14 fps Chronos: 15.51 fps CHFast: 14.77 fps
16X Slowmo Aion: 27.91 fps
Something doesnāt seem right here.
See the 5070Ti bench directly above yours which doesnāt have that.
Maybe itās the GPU setting (auto vs. single)?
regarding changes in performance, see my note here.
For Nyx XL specifically, the irregular frequency at which the model outputs frames was causing our old method of calculating FPS to vastly overrate the modelās throughput.
Topaz Video v1.2.0
System Information
OS: Windows v10.22
CPU: AMD Ryzen 7 5700X 8-Core Processor 63.928 GB
GPU: NVIDIA GeForce RTX 5080 15.604 GB
Processing Settings
device: -2 vram: 1 instances: 1
Input Resolution: 1920x1080
Benchmark Results
Artemis 1X: 24.00 fps 2X: 11.46 fps 4X: 02.99 fps
Iris 1X: 36.47 fps 2X: 13.95 fps 4X: 03.52 fps
Proteus 1X: 34.88 fps 2X: 14.22 fps 4X: 03.56 fps
Proteus Natural 2X: 10.97 fps
Gaia 1X: 12.79 fps 2X: 08.73 fps 4X: 03.15 fps
Nyx 1X: 17.44 fps 2X: 13.71 fps
Nyx Fast 1X: 34.88 fps
Nyx XL 1X: 02.78 fps
Rhea 4X: 03.02 fps
RXL 4X: 03.04 fps
Hyperion HDR 1X: 12.79 fps
4X Slowmo Apollo: 29.04 fps APFast: 30.96 fps Chronos: 26.79 fps CHFast: 29.91 fps
16X Slowmo Aion: 29.35 fps
Sometimes faster than 1.1.0, sometimes irregularly slower.
So, letās get to the real-world benchmarks.
One input, 1266x924. Output is FHD letterbox.
Three models: Proteus, Iris medium, Rhea. All settings are Dynamic.
TVAI 1.0.4:
Proteus 8m39s
Iris 8m55s
Rhea 25m11s.
TVAI 1.1.0:
Proteus 8m47s
Iris 8m51s
Rhea 25m25s.
TVAI 1.2.0:
Proteus 8m34s
Iris 8m51s
Rhea 25m21s.
The difference is negligible.
Testing my RTX A4000 in a HP Elite Tower G9 vs Arc A770 in a DIY box and , both running 32GB DDR5 Crucial ram.
I may need to swap video cards between the boxes with better cooling in the DIY box as the A4000 has a little fan on it compared to the larger fans on the A770 and lack of fans in the HP Elite tower.
Interesting that Nyx XL and Aion are both faster with the Arc card.
Iāve also found in the past that even putting all 4x 16GB ram sticks into one machine doesnāt seem to make much of a difference with the ARC card. I should probably try with the A4000 and see if what happens.
Kinda bummed that I missed out on 96GB ram before the price hikes.. *sigh
More testing today, but as we understand, actual processing performance is different from benchmarks.
Cheers Dave
Topaz Video v1.1.1
System Information
OS: Windows v11.25
CPU: 12th Gen Intel(R) Core(TM) i5-12500 31.677 GB
GPU: NVIDIA RTX A4000 15.79 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: 13.32 fps 2X: 09.68 fps 4X: 02.56 fps
Iris 1X: 15.60 fps 2X: 09.38 fps 4X: 03.05 fps
Proteus 1X: 13.39 fps 2X: 10.03 fps 4X: 03.63 fps
Gaia 1X: 04.35 fps 2X: 02.96 fps 4X: 02.20 fps
Nyx 1X: 06.25 fps 2X: 05.51 fps
Nyx Fast 1X: 13.54 fps
Nyx XL 1X: 25.53 fps
Rhea 4X: 02.19 fps
RXL 4X: 02.10 fps
Hyperion HDR 1X: 16.69 fps
4X Slowmo Apollo: 15.80 fps APFast: 57.94 fps Chronos: 09.48 fps CHFast: 14.47 fps
16X Slowmo Aion: 23.59 fps
Topaz Video v1.1.1
System Information
OS: Windows v11.25
CPU: 12th Gen Intel(R) Core(TM) i7-12700 31.773 GB
GPU: Intel(R) Arc(TM) A770 Graphics 15.875 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: 12.62 fps 2X: 08.22 fps 4X: 02.89 fps
Iris 1X: 09.51 fps 2X: 06.09 fps 4X: 02.19 fps
Proteus 1X: 13.22 fps 2X: 09.97 fps 4X: 04.08 fps
Gaia 1X: 06.12 fps 2X: 04.37 fps 4X: 00.68 fps
Nyx 1X: 00.53 fps 2X: 02.37 fps
Nyx Fast 1X: 04.84 fps
Nyx XL 1X: 69.91 fps
Rhea 4X: 01.66 fps
RXL 4X: 01.69 fps
Hyperion HDR 1X: 22.45 fps
4X Slowmo Apollo: 15.02 fps APFast: 47.40 fps Chronos: 07.17 fps CHFast: 10.78 fps
16X Slowmo Aion: 41.14 fps
