News Intel's next-gen Arrow Lake GPU will have new Xe-LPG Plus Architecture with XMX

Status
Not open for further replies.
"...XMX extensions are designed for matrix multiplication operations in FP64, FP32, FP16, and bfloat16 formats"

think you should check your sources on FP64.
int8 and fp8 maybe yes. fp64 doubtful.

For example, Ponte Vecchio has only TF32, FP16, BF16 and Int8 formats. TF32 is only 19 bits, so even matrix FP32 on these GPUs would be doubtful. I vaguely recall that the consumer ARC GPUs supported only some subset of those on PVC.
 
Last edited:
Status
Not open for further replies.