Specifications
Sourced specs
| Spec | Value | Source |
|---|---|---|
| FP16/BF16 compute (K100 AI version) | 192 TFLOPS | via mirrorfrog.com ↗2026-08 |
| INT8 compute (K100 AI version) | 392 TOPS (FP32 49 TFLOPS, TF32 96 TFLOPS) | via emcreative.eastmoney.com ↗2024-11-12 |
| FP64 compute (original K100) | ~5.2 TFLOPS (dual-precision units replaced by AI units in K100 AI version) | via emcreative.eastmoney.com ↗2024-11-12 |
| Memory | 40GB HBM2e (64GB variant rumored), ~1.2 TB/s read | via mirrorfrog.com ↗2026-08 |
| TDP | 300-400 W (sources vary), PCIe dual-slot | via mirrorfrog.com ↗2026-08 |
| Software | DTK toolchain, ROCm/CUDA-compatible | via emcreative.eastmoney.com ↗2024-11-12 |
Pricing
Price history, newest first
| Date | Price | Type | Source | Confidence |
|---|---|---|---|---|
| 2025 | ~$7.8K from ¥56,000 at 2025 avg 7.19 CNY/USD per cardK100 AI version retail ~5.6万元/卡, claimed roughly half of comparable NVIDIA H20 pricing; date approximate | Street price | via Xueqiu (retail channel claim) ↗ | low |
vendor-official or multiple independent reports · single major report or several secondary · estimate or lone secondary source
Prices in USD, converted at each observation year's average exchange rate; the original figure and rate print under every converted price. How prices are sourced and graded ↗
Context
Notes
- Hygon has never officially published DeepComputing II specifications; all figures are third-party and version-dependent (standard K100 vs K100 AI).
Where it sits
In the stack
All specs and prices link to their original sources, dated as observed. Figures are market observations, not offers; treat each as of its date.