MetaX · Accelerator Card
C588
MetaX C588 is a new flagship general-purpose GPU (GPGPU) from MetaX, engineered on its proprietary core GPU IP. It delivers powerful multi-precision computing power, large-capacity high-bandwidth memory, and advanced multi-GPU interconnection (MetaXLink) technology. Coupled with the MXMACA software stack compatible with mainstream GPU ecosystems, this combination delivers superior performance, making it highly effective for large language models (LLMs) training and inference, general-purpose computing, and data processing.
Key Highlights
- Single module equipped with 128GB GPU Memory, featuring end-to-end ECC protection
- Capable of running the full-featured (full-blooded) DeepSeek model smoothly on a single server, delivering outstanding performance
- MetaXLink interconnect technology supports 8-card full-interconnect topology with bandwidth up to 896 GB/s
- OAM 2.0 compliant form factor module with a Maximum Board Power of 850W
Technical Specifications
| GPU Memory | 单模组128GB高带宽显存，全链路ECC / 128GB per module, end-to-end ECC |
|---|---|
| 精度算力（FP16/FP8/INT8, TFLOPS/TOPS）/ Precision Compute: | |
| Interconnect | MetaXLink 8卡全互连，带宽最高896GB/s / 8-card full interconnect up to 896 GB/s |
| OAM2.0 | |
| Typical Board Power | 最大850W / Max 850W |
| Form Factor | 符合OAM2.0规范扣卡式模组 / OAM 2.0 mezzanine module |
| Referenced Capability | 单机8卡可流畅运行DeepSeek满血版 / single-server 8 cards run full DeepSeek |
Typical Use Cases
Model Training and InferenceData ProcessingGeneral-Purpose Computing
img1: main; alt=官网产品页主视觉(banner); 官方图右下角标注"部分图片由大模型生成"，为含绿色科技背景的产品渲染主图。; 精度算力、单卡显存带宽、制程官网未公开；无公开硬件Datasheet PDF。
