Nvidia arch 型号对照

• -arch=sm_xx
• Fermi (CUDA 3.2 until CUDA 8) (deprecated from CUDA 9):
  • SM20 or SM_20, compute_30 – Older cards such as GeForce 400, 500, 600, GT-630
• Kepler (CUDA 5 and later):
  • SM30 or SM_30, compute_30 – Kepler architecture (generic – Tesla K40/K80, GeForce 700, GT-730)
    Adds support for unified memory programming
  • SM35 or SM_35, compute_35 – More specific Tesla K40
    Adds support for dynamic parallelism. Shows no real benefit over SM30 in my experience.
  • SM37 or SM_37, compute_37 – More specific Tesla K80
    Adds a few more registers. Shows no real benefit over SM30 in my experience
• Maxwell (CUDA 6 and later):
  • SM50 or SM_50, compute_50 – Tesla/Quadro M series
  • SM52 or SM_52, compute_52 – Quadro M6000 , GeForce