Can i run video ai on my server on VM

Will topaz video ai work on a vm because i want to move it to my home nas to free up my gaming pc i will be using unraid on my home nas

it might work if your VM presents the Hardware GPU to the VM guest.

2 Likes

We do not support the use of Topaz Video AI on a VM environment, however, as @Akila said, this may work as long as the VM meets our system requirement.

I am not familiar with the capabilities of unraid; however, Topaz Video Ai most definitely works on a vmware esxi server vm with “Dedicated GPU Pass-through”. If unraid can likewise present the bare metal GPU to the your vm it ought perform well.

The ESXi server used here has 43 vms but the one vm has sole access to the RTX 2080 ti and it works a treat to upscale old VHS interlaced videos to 4K with commendable results.

unraid is able to virtualize, partition or passhtrough a GPU from a host to a VM.

BUT: It´s not “klick and go” and a. lot of “depends”… You have to know what you are doing :slight_smile:
But if you manage to end up with a windows virtualmachine that has full access to the GPU (or a partitioned set of it), you are of course good to go.

MOst people don´t have a big GPU In theyr NAS - they have it in the gaming PC…

Maybe you provide some more specific information about the setups of your gear?

My home nas have x790 moatherboard with 13600k without gpu and 64gb ram

I have space to add gpu because it is atx case but I didn’t buy the gpu yet before making sure it will work on vm

Currently i’m upsacling animes and moving them to the nas what i want is to upscale on the server so i can play game whlie the app running on the server

Out of curiosity, why is everyone talking about passing a GPU through to the VM?
I thought Topaz Video Enhance AI supported CPU processing (at a significant speed reduction compared to most GPUs). You could do CPU processing in a VM without a dedicated GPU. And once you’ve got that sorted, you could transplant your GPU from your gaming computer to the server and test to make sure GPU passthrough works, then shift the GPU back to the gaming computer and buy a GPU specifically for the server.

1 Like

Do you see any performance difference between doing it directly on the host or via the VM guest?

We know CPU would most likely work, but the whole idea is to get the Performance of the GPU processing the VAI video, on a VM guest

The question is the other way around…

Its not “will tvai run on a VM”, its “will you be able to create a VM where TVAI will run”…

If you get a VM with enough compute Power, dedicated GPU inside the VM - of course you can run it…

Personally, I´d strongly advice against it - if you expose your GPU into a VM and run hevay computation 24/7 on it - one crash of your GPU might crash the whole NAS with it. Its in the nature of a NAS to store important data - I personally would not mix these usecases into one machine…

after all. the speed of the upscale is mostly dependand on the GPU… If you just want to buy a medium GPU for scaling some stuff now and then - you could simply get a used PC plattform for a hundert bucks and throw a 3070 or so in it … Why bother with a VM?

If you aim for a 4090 in your NAS - you have all kinds of other problems - heat dissipation, power… Unless you basically have a heavy duty PC and call it a NAS… :slight_smile:

yes, you could do it on a CPU…

the issue is: its so much slower and more power consuming - unless we are talking about this occasional one minute 320x240 clip … You´d need a veeeery big CPU to even get into “i am done in days or weeks, not months” scenario… Don`t look at the power bill…

It absolutely makes no sense to do it on a CPU if one uses stuf like this on a regular basis. Every 100 buck used GPU will outpefrom even the beafiest CPUs…

Besides - who has a 4. gen Threadripper in the NAS??? (Yes, I know you can do that . but then you basically do not have a NAS anymore, but a big machine you call NAS and use it as a a NAS… Netwprk attached storage is NOT about paramount compute power, but about data integrity, efficiency, banwith, storage, sharing, etc… all things a high-end cpu is not needed for…)

We had the same discussion over at Facebook a few days ago…

TLDR: Of course, from a technical viewpoint, you can run it in a VM. In moste cases, it makes no sense. Its not a plug& play thingi… Not because TVAI does not “support it”, but because its not plug&play to have a VM with a full dedicated GPU…

To be fair, I meant the ESXi server used there as I’ve only remote access to it. Also, this is not in any way a home nas; rather an aged-out massive NLP compute server w/ dual platinum xeons, 2Tb ram, 3 water-cooled A100s. This vm has 32 cores, 128 gb, RTX 2080 ti, with Solidworks, Davinci Resolve and Topaz Video Ai as the only apparent workloads.

I‘ve installed TVAI on my esx 8 with my old GTX 1660 TI (PCI passthrough), after solving the annoying error 43 for GPU PCI passthough it work fine. Not really performant because of the GTX 1660 TI but with no issues. esxi is a AMD Epyc 7282 with 128 GB Ram but running quiet a lot other VMs during that „test“.

Don‘t know if this helps :slight_smile: