This sounds more like an LLM box than a desktop. High memory bandwidth, GPU cores, kinda like the Mac minis running Exos. With 128 GB it would beat the pants off a much more expensive nvidia.
It looking like an AI machine is more AMD’s doing than Framework’s. Framework just grabbed one of, if not the highest performing mobile chip, which happens to be shipped with a big ass gpu
Well, I mean they didn’t need to get full bandwidth on the memory, they could have just gone with socketed RAM. And 128GB max is a hell of a lot for a desktop, but not a machine that you want to load multiple large models into and have accessing as fast as possible.
I’m considering buying the bare board for a local processor after someone else takes the risks and benchmarks it.
This sounds more like an LLM box than a desktop. High memory bandwidth, GPU cores, kinda like the Mac minis running Exos. With 128 GB it would beat the pants off a much more expensive nvidia.
It looking like an AI machine is more AMD’s doing than Framework’s. Framework just grabbed one of, if not the highest performing mobile chip, which happens to be shipped with a big ass gpu
Well, I mean they didn’t need to get full bandwidth on the memory, they could have just gone with socketed RAM. And 128GB max is a hell of a lot for a desktop, but not a machine that you want to load multiple large models into and have accessing as fast as possible.
I’m considering buying the bare board for a local processor after someone else takes the risks and benchmarks it.