cuda for amd on windows
speculative decoding in vllm on amd gpus
amd strix halo rdma cluster setup guide
performance per dollar is getting faster and cheaper