Personal View site logo
Graphic cards peak theoretical performance
  • Vitaliy_KiselevVitaliy_Kiselev

    Made simple pointed lists that specify only memory bandwidth and peak possible single precision performance.

    In editing and playback the only other things that matter are fixed encoder and decoder, see previous blog posts for it.

    AMD

    • Radeon R9 270X, 179.2GB/s, 2560(2688) GFlops
    • Radeon R9 280X, 288GB/s, 3481.6 (4096) GFlops
    • Radeon R9 290X, 320GB/s, 5632 Gflops
    • Radeon R9 380X, 182.4GB/s, 3973.1 Gflops
    • Radeon R9 390X, 384GB/s, 5913.6 Gflops
    • Radeon RX 460, 112GB/s, 1953(2150) GFlops
    • Radeon RX 470, 211GB/s, 3793(4940) GFlops
    • Radeon RX 480, 224GB/s, 5161 (5834) GFlops
    • Radeon RX 580, 256GB/s, 6100 GFlops

    Nvidia

    • GeForce GTX 670, 192GB/s, 2460 GFlops
    • GeForce GTX 680, 192GB.s, 3090 GFlops
    • GeForce GTX 770, 224GB/sec, 3213 GFlops
    • GeForce GTX 780, 288GB/sec, 3977 GFlops
    • GeForce GTX 780Ti, 336GB/Sec, 5046 GFlops
    • GeForce GTX 970, 196GB/s, 3494 GFlops
    • GeForce GTX 980, 224GB/s, 4612 GFlops
    • GeForce GTX 980 Ti, 336GB/s, 6144 GFlops
    • GeForce GTX 1050 Ti, 112GB/s, 1981(2138) GFlops
    • GeForce GTX 1060 6GB, 224GB/s, 3855 (4372) GFlops
    • GeForce GTX 1070, 256GB/s, 5783 (6463)
    • GeForce GTX 1080, 352GB/s, 8228 (8873)
    • GeForce GTX 1080 Ti, 484GB/s, 10609 (11340)
  • 6 Replies sorted by
  • Vitaliy_KiselevVitaliy_Kiselev

    OpenCL

    https://browser.primatelabs.com/opencl-benchmarks

    CUDA

    https://browser.primatelabs.com/cuda-benchmarks

  • Vitaliy_KiselevVitaliy_Kiselev

    Addition

    AMD Radeon RX Vega 2017

    • Radeon RX Vega Nova, 410Gb/s. 12800 Gflops (25600 half precision)
    • Radeon RX Vega Eclipse, 410Gb/s. 12000 Gflops (24000 half precision)
    • Radeon RX Vega Core, 410Gb/s. 10500 Gflops (21000 half precision)
  • Vitaliy_KiselevVitaliy_Kiselev

    Moved to category.

    Make sure to check

    http://www.personal-view.com/talks/discussion/16921/hevc-decoding-and-nles#Item_2

    http://www.personal-view.com/talks/discussion/16925/h.264-decoding-and-nles#Item_1

  • Vitaliy_KiselevVitaliy_Kiselev

    DaVinci Resolve Is Coded For FP32 Only

    DaVinci Resolve utilises only FP32 operations and makes no use of FP64 operations at all. So any argument that a FP64 workstation class GPU is better suited to DaVinci Resolve due to it’s “better” level of precision is completely unfounded.

  • Vitaliy_KiselevVitaliy_Kiselev

    More Vega 2018 info

    • Vega 20, 720GB/sec, 14000 GFlops
    • Vega 10, 480 GB/sec, 13000 GFlops
    • Vega 11, 240 GB/sec, 9000 GFlops
  • Vitaliy_KiselevVitaliy_Kiselev

    Nvidia 2018 prospects

    • GV102, 670GB/sec, 18000 GFlops
    • GV104, 450GB/sec, 13000 GFlops
    • GV106, 240GB/sec, 7000 GFlops
Add a Comment