Might want to run the benchmark with different settings (1 GPU, the other GPU, both GPUs, CPU and Auto) and compare framerates.
1 Like
Will do (at least I know now that VEAI has a Benchmark option under Process).
1 Like
Had a conversation with someone at Topaz who advised me that TVAI does not use multiple GPUs for single operations, but just divides queued files between them. So I have to assume that any improvement in speed observed for a single operation is just the result of one GPU being assigned the processing calculations while most everything else GPU-related on the computer goes to the other.
2 Likes
âAll GPUsâ is an experimental option which was introduced with V2 ( see Video Enhance AI v2.x | Topaz Video AI ). Here we are with V4 and itâs still âexperimentalâ. What your Topaz contact says now isnât what they had intended when V2 was released.
From their V2 FAQ
Q: My machine has more than one Nvidia GPU, can I use all of them?
A: Yes, you can use the Experimental All GPU option.
The Experimental All GPU setting is intended for users with multiple similar Nvidia or AMD GPUs installed (it will not work for Intel GPUs). For example, if you have two NVidia RTX 2080 cards installed or one GTX 1070 and one RTX 2080, this setting is for you. If you have a dedicated GPU and an integrated Intel GPU, you should choose your dedicated GPU, the All GPUs setting may not work for you or perform more slowly. To enable both graphics cards you would want to select Preferences from the Video Enhance AI menu and change the AI Processor to All GPUs. VEAI can only use 8-1 2 GB of VRAM in one instance, and canât consume more than this at this time, you can launch VEAI again on a windows machine and that will allow the second instance to consume more. We are working on optimizations to use all ram
As far as I can tell from my GPU-Z monitoring, it can share the workload between 2 similar GPUs but that doesnât mean the job gets done any faster. The elapsed time to do an iris upscale export was about the same whether I selected one GPU or used âAll GPUsâ.
Also, I read elsewhere (sorry I canât remember where) that âAutoâ selects the more powerful GPU ⊠which is also what I saw on GPU-Z (ie. one GPU did most of the work while the other was idle).
1 Like
I think I can count on one hand the number of times Iâve seen VEAI use more than 4Gb of VRAM, much less 8-12. And that was when it was running two exports at the same time. I think thatâs the real boost, use two GPUs to run two exports simultaneously and theyâll get done faster.
1 Like
