I just spent way too much time trying to get WAN2.2 running on a B200, and I’m not even sure why I bothered. These machines aren’t exactly built for this type of workload, and getting Comfy up and running can be a real pain. But I was curious to see just how far I could push the technology.
The results? Well, I was able to crank everything up to max and run full precision throughout my workflow, no main device offloading, and top resolution. No sweat. And the peak VRAM use? A whopping 96 GB. Not bad, but not exactly surprising either.
Of course, I didn’t exactly use this opportunity to create something amazing. Nope, I just went for the classic “generate bikini” request. I mean, zero imagination is a curse, right?
In hindsight, it was kind of a waste of a good machine. But hey, at least I learned something. And who knows, maybe someone out there will find this useful. Maybe.
## The Takeaway
Running WAN2.2 on a B200 is technically viable, but it’s not exactly the most efficient use of resources. If you’ve got a 6000, you’re probably better off using that instead, with its various optimizations for retail.
## The Experiment
I’m not going to lie, this was a bit of a fun experiment, even if it was a bit of a waste of time. It’s always interesting to see just how far you can push technology, even if it’s not exactly practical.
## The Future
Who knows what the future holds for WAN2.2 and other similar technologies? Maybe one day we’ll have machines that can handle this type of workload with ease. Until then, I’ll just keep experimenting and seeing what’s possible.