New Arrival
AI Compute.
Fully Reimagined.
256GB VRAM on a single card. 100% domestic supply chain. Zero export restrictions. From edge inference to data center training — one platform, full stack.
The AI Compute Problem
GPU shortages. Export controls. Skyrocketing prices. Building AI infrastructure shouldn’t be this hard.
Supply Chain Uncertainty
Export controls and geopolitical tension make GPU sourcing unreliable. Lead times stretch into months.
Exorbitant Pricing
Premium NVIDIA GPUs cost $15K-$30K each. Data center deployments run into millions of dollars.
Power Inefficiency
High-end GPUs draw 300-700W each. Cooling and electricity costs eat into ROI.
Three Product Lines. One Architecture.
From compact edge devices to 16-card rack servers — scale seamlessly with the same Jiangyuan D-Series architecture.
Edge AI Industrial PC
Compact, low-power AI inference for factory floors, smart cities, and edge deployment.
- AI Compute160 TOPS INT8
- VRAM64GB LPDDR5
- Video Decode128× 1080p
- Power≤150W
- Use CaseSmart City / Factory
Desktop AI Workstation
Liquid-cooled desktop AI power. For developers, researchers, and creative teams.
- AI Compute320-1280 TOPS
- VRAM128-512GB
- Cards1-4× D20/D20L
- CoolingLiquid cooling
- Use CaseR&D / Dev / Creator
Rackmount AI Server
Enterprise-grade AI servers. 8-16 GPUs per chassis. Data center ready.
- AI Compute1,280-5,120 TOPS
- VRAM512GB-2TB
- Cards8-16× D10/D20
- Form Factor4U Rackmount
- Use CaseData Center / Enterprise
256GB VRAM. Game Changer.
Run 70B+ parameter models on a single card. No model parallelism. No multi-card complexity.
Head-to-Head: D20 vs. NVIDIA L20
How our flagship accelerator stacks up against the industry standard for inference.
| Specification | Jiangyuan D20 | NVIDIA L20 | Advantage |
|---|---|---|---|
| VRAM | 256GB LPDDR5 | 48GB GDDR6 | 5.3X more |
| Memory Bandwidth | 409.6 GB/s | 864 GB/s | — |
| INT8 Compute | 320 TOPS | 239 TOPS | +34% |
| FP16 / BF16 | 160 TFLOPS | 119.5 TFLOPS | +34% |
| Context Window | 128K tokens | 32K tokens | 4X larger |
| Power Draw (TDP) | 150W | 275W | 45% less |
| Efficiency (TOPS/W) | 2.13 | 0.87 | 2.4X better |
| Video Decode | 256× 1080p30 | Limited | Superior |
| Form Factor | Single slot | Dual slot | 2X density |
| PCIe Generation | Gen 5 x16 | Gen 4 x16 | 2X bandwidth |
Built for Every Industry
From smart cities to healthcare to enterprise AI — our compute platform scales to your needs.
Smart City / Public Safety
Real-time traffic analysis, crowd monitoring, and anomaly detection across 128+ simultaneous video streams.
Healthcare & Medical
Medical imaging AI, patient monitoring, and drug discovery acceleration with full data privacy on-premise.
Enterprise AI
On-premise LLM deployment, document processing, and knowledge management with zero data leaving your network.
Education & Research
University AI labs and student training platforms with local model fine-tuning — no cloud dependency.
Manufacturing / Industry 4.0
Defect detection, predictive maintenance, and digital twin optimization on the factory floor.
Government & Defense
100% domestic supply chain, national crypto security built-in, and complete data sovereignty.
Full Product Lineup
All configurations. Find the right compute for your workload.
| Series | Model | Cards | Total VRAM | INT8 TOPS | CPU | RAM | Storage | Starting Price |
|---|---|---|---|---|---|---|---|---|
| Edge Series (Industrial) | ||||||||
| Edge | PYD10-Edge-FT-01 | 1× D10 | 64GB | 160 | Feiteng D3000 | 64GB | 2TB NVMe | $4,999 |
| Edge | PYD10-Edge-IT-01 | 1× D10 | 64GB | 160 | Intel i5-10500 | 64GB | 2TB NVMe | $5,200 |
| Station Series (Desktop) | ||||||||
| Station | PYD20L-Station-IT-01 | 1× D20L | 128GB | 320 | Intel i5-14400 | 48GB | 1TB NVMe | $7,999 |
| Station | PYD20L-Station-HG-01 | 1× D20L | 128GB | 320 | Hygon 3350 | 48GB | 1TB NVMe | $7,999 |
| Station | PYD20L-Station-04-W5 | 4× D20L | 512GB | 1,280 | W5-2455X | 64GB | 2TB NVMe | $19,800 |
| Station | PYD20L-Station-04-W7 | 4× D20L | 512GB | 1,280 | W7-2495X | 128GB | 2TB NVMe | $23,500 |
| Server Series (Rackmount) | ||||||||
| Server | PYD10-Server-FT-08 | 8× D10 | 512GB | 1,280 | Feiteng S5000C | 512GB | 2×480GB + 4×3.84TB | $68,000 |
| Server | PYD10-Server-HG-08 | 8× D10 | 512GB | 1,280 | Hygon 7380 | 512GB | 2×960GB + 4×3.84TB | $70,000 |
| Server | PYD20P-Server-FT-08 | 8× D20 | 1,024GB | 2,560 | Feiteng S5000C | 512GB | 2×480GB + 4×3.84TB + NVMe | $88,000 |
| Server | PYD20P-Server-FT-12 | 12× D20 | 1,536GB | 3,840 | Feiteng S5000C | 512GB | 2×480GB + 4×3.84TB + NVMe | $112,000 |
| Server | PYD20P-Server-IT-16 | 16× D20 | 2,048GB | 5,120 | Intel 6530 | 1TB | 2×960GB + 4×3.84TB NVMe | $132,000 |
* Prices are indicative starting prices. Final pricing depends on configuration, quantity, and shipping destination.
Ready for AI Compute Without Limits?
Get a custom quote for your AI infrastructure needs. Our team will help you size the right configuration for your workloads.
📦 Shipping worldwide · 🔧 Custom configurations available · 🇨🇳 100% domestic supply chain
