Top GPUs & CPUs & AI accelerators Ranked by INT16 TOPS of Matrix(Tensor) cores

Bandwidth Capacity

Memory

TF32 FP16 BF16 FP8 INT8 INT4

Matrix Compute

FP64 FP32 FP16 INT32 INT8

Vector Compute

PC GPU Server GPU CPU All Processors

#1

GPU

32 GB

MLU 290

256 T

1.23 TB/s

#2

AIU

192 MB

T-HEAD Hanguang 800

205 T

#3

AIU

MLU 370

128 T

#4

AIU

MLU 270

64 T