开发者支持

技术特点

  • Performance : 32 TOPs /16 TFLOPs
  • Memory: up to 16GB LPDDR4x
  • PCI-E 3.0 x4 , Up to 32Gb/s
  • Support native CUDA programming
  • Compatible CUDA libraries: cuDNN, cuBLAS.
  • Framework: TensorFlow/Pytorch,/Caffe/MxNet/ONNX
  • Decoder :32 CHs H.264/265 1080P@30
  • RT engine: RPP OpenRT
  • MJPEG/JPEG ENCODER:430M pixels/s
  • Power Consume :<14 Watt
  • 技术文档