OpenCL performance on intel UHD Graphics 730 by Clpeak

Platform: Intel(R) OpenCL HD Graphics
  Device: Intel(R) UHD Graphics 730 [0x4c8b]
    Driver version  : 1.0.0 (Linux x64)
    Compute units   : 24
    Clock frequency : 1300 MHz

    Global memory bandwidth (GBPS)
      float   : 16.22
      float2  : 16.66
      float4  : 17.17
      float8  : 17.24
      float16 : 17.01

    Single-precision compute (GFLOPS)
      float   : 491.72
      float2  : 424.29
      float4  : 489.42
      float8  : 309.67
      float16 : 308.31

    Half-precision compute (GFLOPS)
      half   : 968.81
      half2  : 957.68
      half4  : 964.45
      half8  : 958.57
      half16 : 947.16

    No double precision support! Skipped

    Integer compute (GIOPS)
      int   : 164.62
      int2  : 110.87
      int4  : 105.41
      int8  : 101.77
      int16 : 122.87

    Integer compute Fast 24bit (GIOPS)
      int   : 164.61
      int2  : 110.88
      int4  : 105.41
      int8  : 101.76
      int16 : 122.86

    Integer char (8bit) compute (GIOPS)
      char   : 323.85
      char2  : 318.47
      char4  : 304.82
      char8  : 290.10
      char16 : 303.78

    Integer short (16bit) compute (GIOPS)
      short   : 956.70
      short2  : 947.31
      short4  : 953.54
      short8  : 941.22
      short16 : 918.27

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer              : 8.48
      enqueueReadBuffer               : 8.53
      enqueueWriteBuffer non-blocking : 7.64
      enqueueReadBuffer non-blocking  : 7.62
      enqueueMapBuffer(for read)      : 5368699.00
        memcpy from mapped ptr        : 8.51
      enqueueUnmap(after write)       : 42949592.00
        memcpy to mapped ptr          : 8.50

    Kernel launch latency : 22.03 us