MetaX Xiyun C500/C600
Vendor: MetaX
Category: GPU Graphics Processor
Architecture: MetaX MACA
Introduction
MetaX domestic GPU. Xiyun C-series focuses on cloud training and inference. Xiyun C600 is based on fully domestic process, integrating large-capacity memory and multi-precision hybrid compute, supporting MetaXLink super-node expansion with built-in ECC/RAS security modules. Cumulative sales exceed 55,000 units, 2025 revenue 1.644 billion RMB.
Specifications
| Model | Compute | Memory | Interface | TDP | Process |
|---|---|---|---|---|---|
| Xiyun C600 | TBA (FP8) | 144GB HBM3e | OAM | TBA | 7nm (domestic) |
| Xiyun C500 | 280 TFLOPS (FP16) / 560 (INT8) | 64GB HBM2e (1.8 TB/s) | OAM | 450W | 7nm |
| Xisi N100 | 80 TFLOPS (FP16) / 160 (INT8) | 16GB GDDR6 | PCIe 4.0 | 75W | 12nm |
Official Website
Driver Downloads
Linux
Related Documentation
OS Support
| Windows | Linux | macOS | Android |
|---|---|---|---|
| ❌ | ✅ | ❌ | ❌ |
Version History
| Version | Release Date | Description |
|---|---|---|
| MACA SDK 1.0 | 2024 | First official release, CUDA-idiom compatible |
Performance Benchmarks
| Model | Task | Performance Metric |
|---|---|---|
| Xiyun C500 | BF16 Training | Near A100 efficiency (official data) |
| Xiyun C500 | General AI Inference | INT8 optimized high throughput |
Pricing
| Model | Reference Price | Notes |
|---|---|---|
| Xiyun C500 | Contact vendor | Enterprise AI training accelerator |
| Xisi N100 | Contact vendor | Inference dedicated |
Quick Installation
Linux
# 1. Install MetaX driver and SDK
sudo rpm -ivh metax-driver-*.rpm
# 2. Verify
metax-smi
Driver and SDK obtained from MetaX official.
Code Examples
Python (MetaX)
import torch
# MetaX GPU uses CUDA-compatible backend
assert torch.cuda.is_available()
print(f"GPU: {torch.cuda.get_device_name(0)}")
Architecture Highlights
- MACA Architecture: MetaX proprietary GPU compute architecture for AI training and inference
- CUDA Compatibility: API-level CUDA programming model compatibility, reducing migration cost
- Full-Stack Independence: Hardware, driver, compiler, and libraries all independently developed
Model Compatibility
| Model/Framework | Support | Notes |
|---|---|---|
| PyTorch | ⚠️ Compatible | CUDA-compatible backend |
| General AI | ⚠️ | Ecosystem developing |
| Large Model Training | ⚠️ | Gradually supporting |
Related Products
If you're evaluating alternatives, the following products may also fit your scenario:
- Iluvatar Tianguai 100 — Iluvatar (GPU Graphics Processor)
- Hygon Shensuan Z100 — Hygon (ASIC Dedicated Accelerator)
- NVIDIA GPU / CUDA — NVIDIA (GPU Graphics Processor)
- Biren Technology BR100/BR20X — Biren Technology (GPU Graphics Processor)
- Moore Threads MTT S5000 — Moore Threads (GPU Graphics Processor)
- Baidu Kunlunxin P800 — Baidu (GPU Graphics Processor)
- Alibaba T-Head Zhenwu PPU — Alibaba (GPU Graphics Processor)