Tensor Computing Processor
BM1684

As Bitmain’s third generation TPU for deep learning, BM1684 achieves 6 times better performance than previous generation.

top image for mobile

Key Parameter

Peak performance

17.6 TOPS INT8
2.2 TFLOPS FP32

Video Decoding

32-ch H264/H265 Video Decode

Internal memory size

32MBytes

Processing Unit

1024

Features

BM1684 performance

Provide powerful performance on both Single-Precision (FP32) and Integer Operation (INT8)

BM1684 performance

Integrated high-performance ARM core for secondary development

BM1684 performance

Integrated video and image decoding/encoding capability

BM1684 performance

Ethernet and PCIe ports available

BM1684 performance

TensorFlow and Caffe enabled

BM1684 VS Competing Products with Obvious Advantages

upgrade