• Home
  • Systems
  • MetaX
  • 曦云C550 Shanghai Cube 液冷整机柜
MX

MetaX 曦云C550 Shanghai Cube 液冷整机柜

rack
128× C550
MetaXLink within each 32-card supernode group, with 7U of the cabinet reserved for switching
Power
kW Total
Memory
8192
GB Total

MetaX's high-density liquid-cooled cabinet, and the largest single configuration it publishes. A 47U rack holds four 10U groups plus 7U reserved for switches, and each group is four 2U eight-OAM nodes and one 2U power shelf, giving 128 liquid-cooled 曦云C550 modules in one cabinet. Each 10U group forms a 32-card supernode, which MetaX describes as supporting flexible parallelism strategies for training and inference. MetaX claims the 128-card layout occupies 70 percent less floor area than comparable products, and that the GPUs, network switches and software stack are entirely domestic. No power, memory total or throughput figure is published for the cabinet.

We will point you at suppliers who have it. Free, and no signup.

System Details

GPU Configuration

GPU Model: MetaX C550
GPU Count: 128 GPUs
Architecture:
Interconnect: MetaXLink within each 32-card supernode group, with 7U of the cabinet reserved for switching

System Specifications

Form Factor: rack
Total Power:
Total Memory: 8192 GB
Memory Bandwidth:

Powered by MetaX C550

This system utilizes 128 × MetaX C550 GPUs, each delivering exceptional performance for AI and HPC workloads.

Per GPU TDP

450W

Per GPU Memory

64 GB

Process Node

Architecture

Documentation & Resources

Official Datasheet

MetaX 曦云C550 Shanghai Cube 液冷整机柜 technical specifications

Download PDF

MetaX (沐曦股份 688802) STAR Market IPO first-round inquiry response

MetaX • 2025-08-27

View Document ↗

Typical Use Cases

AI/ML Training
High-Performance Computing
Data Analytics

The MetaX 曦云C550 Shanghai Cube 液冷整机柜 runs 128× C550 GPUs over MetaXLink within each 32-card supernode group, with 7U of the cabinet reserved for switching.