Intel Shares First Benchmarks of Its Next-Gen Crescent Island "D270" GPU, Claiming a 5.9x Throughput Over RTX PRO 6000 As Xe3P Goes Big On FP64
retained text
Not posting this as a threat to AMD MI350P, since MI350P has higher FP64 Performance and uses HBM3E with more total bandwidth. But this is a warning to the crowd and especially PC gamers that think the AI and memory trade will collapse and bring memory prices way down very soon. If Intel can make a card capable of sucking up to 480GB of LPDDR5X, others won't be far behind if price/performance and demand is there for such a product. This model D270, still looks server passive air cooled from the picture though, just like the MI350P. MI350P has a configurable TDP of 450 to 600W, while the Intel D270 is listed at 350W. The MI440X is supposed to be a air cooled capable of up to 8 units in a cluster or server, but it also has pared back its FP32/FP64 performance, so I don't know if there is an MI400 class product for air cooled FP64 work loads like the MI350P is capable of?
document fetched 2026-11-17 00:35 · canonical https://reddit.com/r/AMD_Stock/comments/1xy0mln/intel_shares_first_benchmarks_of_its_nextgen