lolol
Timeline
Post
Remote status
Context
5
@kaia@pleroma.soykaf.com And here I am think getting a second R9700 if my bonus allows me
@rain can AMD do everything CUDA does now? when I had AMD card it sucked so much getting AI stuff to work
@kaia@pleroma.soykaf.com If done with pytorch yeah as their is pretty good pretty good ROCM support there. Also it has really good Vulkan support in most ".CPP" inference engines. Is it one to one no but as far as I see it at those prices they might as well not exist. So it is is ROCM/HIP & Vulkan or nothing at all 
@rain are you planning to use two of those in one build? 
64GB VRAM?
64GB VRAM?
@kaia@pleroma.soykaf.com Yeah I am not 100% sure yet how I want to do it. Was playing with the idea of hooking it up to my Strix Halo machine which has 128GB of unified RAM so I would have 32GB of VRAM with 128GB unified for off loading
Replies
0Fetching replies…