Hero Banner

新產品 - DeepX

NPI Body

DEEPX 邊緣 AI:超低功耗、GPU 級 NPU 模組與系統

1-5W 突破性能效,全面賦能下一代邊緣攝影機、機器人與智慧工廠

Main image includes the DEEPX products available for customer's to purchase through AVNET.

DEEPX DX-M1 在僅 1–5W 功耗下提供高達 200 TOPS 的 GPU 級 AI 效能。結合 IQ8 智慧 INT8 量化技術(保持 FP32 級精確度),DEEPX 可實現多模型並發運行,具備卓越散熱效能並降低 90% TCO。

DEEPX 提供行業領先的高效邊緣 AI 軟硬體解決方案,推動高性能 AI 在全球市場的輕鬆部署、彈性擴展與極高性價比應用。

 

特點

  • 1–5W 功耗下的 GPU 級 AI 效能:DX-M1 提供高達 200 TOPS 算力,能效達到 40W GPGPU 的 20 倍。
  • DEEPX IQ8 智慧量化:以低功耗 INT8 運行,保留 FP32 級精確度(準確率損失 <1%)。
  • 一體化 DEEPX SDK:無縫支援 PyTorch/ONNX 模型轉換,支援多模型並發高效運行。
  • 工業級可靠性:支援 -25°C 至 85°C 寬溫運行,非常適合無風扇邊緣裝置。
  • 降低 90% TCO:僅需傳統 GPU 10% 的成本,實現突破性的 AI 部署。

 

應用

  • 邊緣攝影機系統
  • 智慧出行
  • 智慧工廠
  • 智慧城市
  • 機器人技術
  • 無人機
  • 邊緣運算
  • 智慧家居
  • 智慧零售

 

關鍵元件

       AI Modules

  • DX-M1
  • DX-M1M
  • DX-H1 V-NPU
  • DX-H1 Quattro

       AI System 

  • DX-AI-Player 

       AI Software Platform

  • DXNN SDK

 

優勢

       Superior Thermal Management

  • Maintains high performance while staying safely within industrial temperature range (-25°C to 85°C), making it uniquely suitable for real-world applications where competing solutions fail due to overheating.

       GPU-Level AI Accuracy

  • Achieves GPU-level AI accuracy using quantized Int-8 precision instead of power-hungry FP32, ensuring reliable on-device intelligence that meets industry standards of less than 1% accuracy drop.

       Exceptional Performance Efficiency

  • Delivers around 20x power performance efficiency compared to GPGPUs and outperforms competitors by more than 2x, ensuring on-device systems can maintain high-level AI capabilities without compromise.

       Industry-Leading TCO

  • Even if competitors offered their chips for free, the DX-M1 would still be more economical. Lower power consumption translates to dramatically reduced operational costs, making advanced AI economically feasible at scale.

 

框圖

 

Body Content Spots

Supplier Logo

Content Spots