VEAI no longer uses CUDA from version 1.7.0 for processing on Nvidia GPUs. If you are having issues with processing and clean install of the studio driver doesn’t fix it, please create a support ticket and we can look into the issue in detail.
On GTX series cards, for most input resolutions the performance might be slightly slower than 1.6.1. We are aware of this issue and it will be fixed with a combination of VEAI, Windows and Nvidia driver update in the future.
Thank you. I am trying to obtain a RTX card anyway so that should fix my issue if nothing else…
I noticed you recommended a 3090 user to run multiple instances of VEAI. Is that a new feature? I was not able to run 2 when I tried (not that I need to as I am using 100% of GPU).
Another question I have is about CLI: Does running CLI commands still open GUI as well? Normally when I use CLI binaries, the apps provide no GUI and I prefer it that way.
If the GPU usage is lower than 90% and VRAM is available then we recommend running multiple instances of VEAI. VEAI 1.7.0 and above should be able to run multiple instances on the same GPU as long as the GPU has enough VRAM. Use the reduce VRAM usage preference and try running multiple instances.
Yes running CLI commands still opens the GUI.
Which one, please?
I’m having GTX 960, So it doesn’t support studio drivers. I managed to install the 1.7.1 early version and it worked perfectly, before I removed it (trial ended). I did a fresh install after that with the license, but it doesn’t work. Going back to 1.6.1 helps, but it’s not fast as 1.7.1 and artemis was better in in 1.7.1
Since Turing the Nvidia GPUs are very good at Async Compute. Also because the software uses Tiles to Compute and between two tiles it stays at almost idle and there it could compute another tile from another instance.
And then you have almost the double Throughput.
This reminds me to nvidias 2017 GTC when their CEO said “this is one GPU” and the V100 did crush through the images.
I would recommend joining the beta and trying out the 1.8.0.0b Beta. Both 1.7.1 and 1.8.0 should work with GTX 960. You can also try downloading 1.7.1 and installing, the installer should be around 1.5GB.
Where do I join beta?
The beta is actually available in two places, here in the community, and on Facebook. The link to apply here is Topaz labs video beta testing group - Topaz Discussion Forum and the facebook group is Topaz Video AI | Facebook
2 Likes
I think you may have set something wrong. 
GTX1070 should be faster card than my old RX580 (4gb).
6.463 TFLOPS vs 6.175 TFLOPS
CPU: Ryzen 7 1700
GPU: RX580 (4GB)
OS: Win 10, Video: 640x360, 29.97fps, 2x Gaia-HQ
VEAI 1.7.1: 0.31 sec/frame
Im not a fan of 1.7.1 , it seems that you changed the compression level of video to make it seem like it renders faster, which it does, but it also results in files being almost twice the size as before with no better quality at all, the only upside is the rendering speed being much faster.
To me this is a downgrade, having files that would be 3gb is now becoming massive 6-7gb with the exact same quality, they do render a lot faster but considering i am keeping those files forever it seems like i’d rather wait a little extra than wasting twice the space to house them later.
Can you please give us an option to specify the compression level so the files don’t turn out so massive, i’d rather have low speed renders than having high speed renders. And no this is not the same as changing the CRF level, the CRF between old versions and new are the same yet the new vesion renders all the videos 3x times as fast but the files also become 2x or more in size.
Sorry to go off-topic, but how is 1.8.0b compared to 1.7.1? In terms of speed and quality for Gaia-HQ
try upscaling to 4k and watch files go over 100gb
yeah completely useless, and to think i purchased a 3090 card to process video files and now i end up running on my old 2080 ti because the new version of VEAI is awful and files becomes so large they are not usable.
Waste of 1500$
Gaia-HQ is the same from 1.7.1, the performance improvements in 1.8.0.0b is only for intel hardware. There are 2 new models Artemis AA for Anti-Aliasing and Artemis MQ for enhancement of video with average noise and compression artifacts.
2 Likes
thanks 
I’ll keep on 1.7.1 then
I’m going nuts!! I tried every possible way, and now I suspect there is something wrong with my new Gigabyte rtx3090.
My system：
i9-10900K
Gigabyte Z490 AORUS MASTER
DDR4-3200 Corsair 32G
U.S. commercial pirate ship RM850x
On my systems:
720P to 2K(200%) Artemis v8
RTX2080:0.22~0.26sec/frame
RTX3090:0.24~0.28sec/frame
I tried the developer’s advice above,It’s only going to get slower and slower
I feel terrible. I think I wasted $1500
Why is your 3090 0.85sec/frame…
my systems1920x1080 to 8k GAIA-HQ5
RTX3090 1.25/frame （i9-10900K DDR4-3200 32G ）
I know you are a developer, so correct me if I’m wrong: Artemis LQ is vastly improved. Artemis MQ is an improved version of the old HQ, and Artemis HQ is a new “Super” Q.
