Cambricon MLU690 vs Moore Threads MTT S5000: Spec Comparison & Buyer's Guide
In AI infrastructure selection, Cambricon MLU690 and Moore Threads MTT S5000 are two accelerators frequently compared. This article contrasts them item by item — architecture, compute, memory, power, and release cadence — to help you quickly judge which fits training or inference workloads.
Spec Comparison Table
| Vendor | Cambricon MLU690 | Moore Threads MTT S5000 |
|---|---|---|
| Vendor | Cambricon | Other |
| Architecture | MLUarch | MUSA |
| Process | 5nm -class | TSMC 6nm |
| Release Date | 2025 | 2025-02-12 |
| FP8 Compute | — | — |
| FP16 Compute | 700+ TFLOPS | — |
| FP32 Compute | — | 62.5 TFLOPS |
| INT8 Compute | 2,800+ TOPS | 2,000 TOPS |
| Memory Type | — | — |
| Memory Capacity | 196 GB HBM3 | 80GB GDDR6X |
| Memory Bandwidth | 3.35 TB/s | 1.6 TB/s |
| TDP Power | ~500 W | 300 W |
Key Differences
- Power: Moore Threads MTT S5000 has a TDP of 300 W, lower than Cambricon MLU690's ~500 W, friendlier to datacenter PUE and cooling.
- Memory capacity: Cambricon MLU690 packs 196 GB HBM3, more than Moore Threads MTT S5000's 80GB GDDR6X, more comfortable for single-card hosting of very large models.
Selection Advice
- When chasing extreme single-card compute and a mature toolchain, prioritize Cambricon MLU690; if budget, power wall, or local support are hard constraints, Moore Threads MTT S5000 often fits better. Use this site's AI Compute Card Comparison Tool to validate multiple chips side-by-side before deciding.
FAQ
What are the main differences between Cambricon MLU690 and Moore Threads MTT S5000?
The core difference is architecture and compute density: Cambricon MLU690 uses MLUarch, FP8 ~No public FP8 data, memory 196 GB HBM3; Moore Threads MTT S5000 uses MUSA, FP8 ~No public FP8 data, memory 80GB GDDR6X. See the comparison table above.
What is the TDP (power) of Cambricon MLU690?
Cambricon MLU690 has a TDP of ~500 W; actual whole-system power also includes board, fans, and PUE.
Which is better for large-model training / inference?
Training values memory capacity, bandwidth, and multi-card interconnect; inference values single-card throughput and power efficiency. Combine the "Key Differences" and "Selection Advice" above with your batch size, model size, and SLA.
How much do Cambricon MLU690 and Moore Threads MTT S5000 differ in memory capacity?
Cambricon MLU690 is 196 GB HBM3, Moore Threads MTT S5000 is 80GB GDDR6X; the gap directly affects loadable model size and context length.
Related Pages
- Cambricon MLU690
- Moore Threads MTT S5000
- AI Chip Comparison Tool — Compare 2–4 chips side-by-side online