Cambricon logo

Cambricon MLU370-X8

Type: NPUArchitecture: MLUarch03Form factor: PCIeReleased: 2022Process: TSMC 7nmSpec confidence: Official
FP32
24
TFLOPS
Memory
48 GB
LPDDR5
Bandwidth
614 GB/s
memory
TDP
250 W
96.0 FP32 dense TFLOPS/kW of TDP

Overview

The Cambricon (寒武纪) MLU370-X8 is a dual-chip Siyuan 370 (思元370) training/inference accelerator built on TSMC 7nm (MLUarch03), launched March 2022. It is a full-height full-length dual-slot PCIe Gen4 card pairing two Siyuan 370 chips via an MLU-Link bridge (200 GB/s aggregate). Cambricon's official spec table lists 48 GB of LPDDR5 (not HBM) at 614.4 GB/s, a 250W board, and dense peaks of 24 TFLOPS FP32, 96 TFLOPS FP16/BF16, and 256 TOPS INT8. INT16 (128 TOPS) is published but has no matching precision column in this database and is omitted; the card has no FP8 or TF32 datapath.

Performance

Peak theoretical throughput by precision type

PrecisionPeak
FP64
No verified data available
FP32
32-bit floating point
24TFLOPS
TF32
No verified data available
BF16
Brain Float 16
96TFLOPS
FP16
16-bit floating point
96TFLOPS
FP8
No verified data available
FP6
No verified data available
FP4
No verified data available
INT8
8-bit integer
256TOPS

The MLU370-X8 in the GPU landscape

Peak FP16 TFLOPS (dense) against TDP, single-GPU parts tracked by Flopper

06001,2001,8002,4003,0000 W250 W500 W750 W1000 W1250 W1500 WInstinct MI355XMLU370-X8

Higher and further left is better: more half-precision throughput for less power.

Specifications

Architecture

MLUarch03

Form Factor

PCIe

Launch Year

2022

Process Node

TSMC 7nm

Memory

48 GB LPDDR5

Bandwidth

614.4 GB/s

TDP

250 W

Max power (Flopper estimate)

~288 W est. Flopper estimate: 250 W TDP x 1.15. The vendor publishes no maximum board power for this part.

Interconnect

200 GB/s MLU-Link

direction and scope not stated by vendor

Spec Confidence

Official

Full Specifications

Memory
Memory 48 GB
Memory Type LPDDR5
Bandwidth 614 GB/s
Interface Width No verified data available
Interconnect & I/O
GPU-to-GPU MLU-Link
Interconnect Bandwidth 200 GB/s direction and scope not stated by vendor
Power & Thermal
TDP 250 W
Max power (Flopper estimate) ~288 W est.
General
Form Factor PCIe
Architecture MLUarch03
Process Node TSMC 7nm
Launch Year 2022

Datasheet & Resources

Flopper Datasheet

Cambricon MLU370-X8 specifications, generated from our database. Printable.

Vendor product page

Cambricon documentation for the MLU370-X8

View

Cambricon MLU370-X8 product page

Cambricon · 2022-03-21

View

Data Provenance

Every figure traced to a source

Primary Source

Publisher
Cambricon
Published
2022-03-21

Data Quality

Spec confidence
Official
Clock basis
Boost
Core precisions with figures
4 of 9
Normalization
All values in TFLOPS

Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.

Browse GPU Cloud Providers

Frequently Asked Questions

How many TFLOPS does the Cambricon MLU370-X8 have?

The Cambricon MLU370-X8 delivers 24 TFLOPS FP32 and 96 TFLOPS FP16 at peak. Flopper does not currently have a verified FP8 throughput figure for it.

What is the power consumption of the Cambricon MLU370-X8?

The Cambricon MLU370-X8 has a TDP (Thermal Design Power) rating of 250 watts.

How much memory does the Cambricon MLU370-X8 have?

The Cambricon MLU370-X8 is equipped with 48 GB of memory with 614.4 GB/s of memory bandwidth.

What architecture is the Cambricon MLU370-X8 based on?

The Cambricon MLU370-X8 is based on the MLUarch03 architecture, launched in 2022.

Stay Updated on GPU Releases

Get notified when new GPUs are added or specifications are updated.

Loading verification...

No spam, unsubscribe anytime.

Back to GPUs

© 2026 Flopper.io - Compare the GPUs Powering AI