Just be aware that according to the system requirements, E-GPUs are “not supported”. That’s not to say it definitely won’t work; just don’t bank on it.
In addition, to the information from Andy you’re need a Pro Subscription for using multiple GPUs with the current Topaz Video Studio version although, I believe you can use the old Topaz Video AI with more than one GPU at the same time and with the previous Video AI unless you have a PC with a nvidia card you won’t be able to Render with the Starlight Mini Model
Having said that, there is another option if it’s a case of running multiple Renders at the same time then, you can also, run one Render on your Computer and at the same time run another in the Cloud however, this method will incur purchasing Cloud Credits but never the less still an option.
I use an eGPU every day. I use Video AI version 5.0.4. I believe that this is the last version that came out that allowed usage of multiple GPUs without the Pro subscription, but I’m not certain. It was around that time that my update subscription ran out, and I chose not to renew because I value the ability to use multiple GPUs more than additional AI models, etc.
My setup is: Alienware m18 r2, i9-14900HX CPU, RTX 4080 laptop GPU. Connected to this via Thunderbolt USB-C cable is a Sonnet Breakaway Box containing a RTX 3080 Ti. By using the RTX 3080 Ti, I’m able to cut processing time by about half. For example, instead of 4 hours for a ~50 minute TV show, (FHD to UHD using Gaia), I can do the processing in about 2 hours plus change. Usually, this works fine. However, sometimes there are glitches. You just have to be patient and be willing to learn and accept the idiosyncrasies.
I do not run two parallel files. I just queue the files.
My understanding is that Video AI processing is dependent on Cuda cores. So you need to look for GPUs that have a high number of Cuda cores. I had an Alienware with a 4070, and the upgrade to 4080 gave way better performance (more Cuda cores). The 3080 Ti has a high number of Cuda cores, too.