I’ve been looking into build a new crazy workstation configuration. I already got a Thermaltake Core W200 case that I was going to use PCI-E hacks to add a lot of GPUs using PCIE cables to utilize both sides of the case with one system rather than the two systems the case supports.
I am really interested in is how does the Xeon W2595 and others do in running stable diffusion and/or large language models with 4 and 8 memory channels on that platform. I’m not expecting the best performance but since the memory limit would be system ram any model should at least load and work even if it’s very slow.
I’m thinking that this could be a good platform for me since the system is being built for mostly AI projects coded in python and the PCIE lane count looks good.