呢的AI pc 嘅4~50 TOPS 係講緊 NPU 嘅速度,仲要係計4bits ,同的大型模型 LLM 要嘅 GPU 行 FP16/32 加速運 ...
javacomhk 發表於 2025-2-10 18:57 
我見AMD 話:
AMD Ryzen AI 300: XDNA 2 NPU with Up To 50 TOPS.
One of the major feature inclusions with the XDNA 2 architecture is support for the Block Floating Point (Block FP16).......Most AI applications require 16-bit precision, and Block FP16 addresses this requirement by simultaneously bringing high performance and high accuracy to the mobile market, at least from an AI standpoint......AMD brings (close to) 16-bit accuracy at 8-bit speed, making it more performant for some AI applications. Altogether, the integrated NPU is slated to offer up to 50 TOPS of compute performance.
https://www.anandtech.com/show/2 ... 5-igpu-xdna-2-npu/2
AMD Ryzen 7040 and 8040 Series processors are built on AMD XDNA NPUs
https://www.amd.com/en/technologies/xdna.html#xdna2
唔知intel 係點, 但今代AMD已經有去到50 TOPS既NPU, 仲有FP16支援. 好似發展得都唔差. |