wgblas
    Preparing search index...

    Module benchmarks/nvidia-geforce-gtx-1650/dswap

    Benchmark results for dswap on Nvidia Geforce Gtx 1650.

    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0072 0.1422 0.0034 0.3048 46.7%
    64 0.0073 0.2801 0.0037 0.5541 50.5%
    128 0.0073 0.5639 0.0037 1.1082 50.9%
    512 0.0071 2.3063 0.0036 4.6126 50.0%
    1024 0.0074 4.4425 0.0036 9.1429 48.6%
    4096 0.0075 17.4298 0.0039 33.9917 51.3%
    16384 0.0091 57.7919 0.0050 105.0256 55.0%
    65536 0.0156 134.0204 0.0104 201.6492 66.5%
    262144 0.0545 153.7952 0.0514 163.3296 94.2%
    1048576 0.1996 168.0949 0.1954 171.7569 97.9%
    4194304 0.7762 172.9075 0.7717 173.9328 99.4%
    16777216 3.1903 168.2837 3.1234 171.8854 97.9%

    Efficiency = wgblas GB/s ÷ cuBLAS GB/s × 100. 100% means parity with cuBLAS; values above 100% mean wgblas achieved greater throughput.

    dswap-default GB/s chart

    dswap-default ms chart

    • dswap.js — WebGPU benchmark script
    • dswap.c — CUDA / cuBLAS reference script

    Unless noted otherwise, every result above uses unit stride (incx = incy = 1) — the normal case, and the coalesced, best-case GPU access pattern. Real usage sometimes passes a non-unit stride (e.g. operating on a row or column of a larger matrix, where incx = lda), which breaks memory coalescing and costs measurably more. This section sweeps a few representative strides to characterize that cost separately, collapsed below by default — expand a stride to see its table and chart.

    Nvidia Geforce Gtx 1650 — stride = 4
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0060 0.1711 0.0029 0.3497 48.9%
    64 0.0061 0.3333 0.0032 0.6465 51.6%
    128 0.0061 0.6667 0.0031 1.3333 50.0%
    512 0.0061 2.6736 0.0036 4.5310 59.0%
    1024 0.0061 5.3333 0.0041 8.0630 66.1%
    4096 0.0072 18.0839 0.0041 32.0000 56.5%
    16384 0.0129 40.7562 0.0095 55.1650 73.9%
    65536 0.0573 36.5714 0.0550 38.1134 96.0%
    262144 0.2283 36.7380 0.2092 40.1016 91.6%
    1048576 0.9134 36.7354 0.8257 40.6362 90.4%
    4194304 3.6478 36.7944 3.2923 40.7677 90.3%

    dswap-stride4 GB/s chart

    dswap-stride4 ms chart

    Nvidia Geforce Gtx 1650 — stride = 5
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0061 0.1667 0.0029 0.3556 46.9%
    64 0.0061 0.3333 0.0030 0.6809 49.0%
    128 0.0061 0.6667 0.0029 1.3913 47.9%
    512 0.0061 2.6667 0.0035 4.6758 57.0%
    1024 0.0063 5.2245 0.0036 8.9825 58.2%
    4096 0.0077 17.0667 0.0042 31.5077 54.2%
    16384 0.0182 28.7439 0.0184 28.4444 101.1%
    65536 0.0770 27.2273 0.0686 30.5672 89.1%
    262144 0.3012 27.8491 0.2657 31.5684 88.2%
    1048576 1.1973 28.0259 1.0547 31.8136 88.1%

    dswap-stride5 GB/s chart

    dswap-stride5 ms chart

    Nvidia Geforce Gtx 1650 — stride = 32
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0061 0.1667 0.0029 0.3478 47.9%
    64 0.0062 0.3299 0.0031 0.6667 49.5%
    128 0.0063 0.6497 0.0033 1.2249 53.0%
    512 0.0068 2.4151 0.0044 3.7101 65.1%
    1024 0.0077 4.2490 0.0044 7.5018 56.6%
    4096 0.0179 7.3339 0.0064 20.3781 36.0%
    16384 0.0688 7.6205 0.0430 12.1905 62.5%
    65536 0.2765 7.5852 0.1864 11.2527 67.4%
    262144 1.1201 7.4890 0.7566 11.0876 67.5%

    dswap-stride32 GB/s chart

    dswap-stride32 ms chart

    Nvidia Geforce Gtx 1650 — stride = 33
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    64 0.0061 0.3333 0.0033 0.6275 53.1%
    128 0.0062 0.6598 0.0033 1.2427 53.1%
    512 0.0065 2.5037 0.0050 3.2821 76.3%
    1024 0.0072 4.5714 0.0041 8.0000 57.1%
    4096 0.0185 7.0926 0.0102 12.8000 55.4%
    16384 0.0758 6.9189 0.0491 10.6701 64.8%
    65536 0.3297 6.3612 0.2153 9.7415 65.3%
    262144 1.3640 6.1502 0.8911 9.4136 65.3%

    dswap-stride33 GB/s chart

    dswap-stride33 ms chart

    Nvidia Geforce Gtx 1650 — stride = 255
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    512 0.0066 2.4976 0.0041 3.9537 63.2%
    1024 0.0073 4.4912 0.0041 8.0000 56.1%
    4096 0.0301 4.3598 0.0123 10.6667 40.9%
    16384 0.1434 3.6563 0.0737 7.1111 51.4%
    65536 0.6287 3.3355 0.3108 6.7469 49.4%

    dswap-stride255 GB/s chart

    dswap-stride255 ms chart

    Nvidia Geforce Gtx 1650 — stride = 256
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    512 0.0068 2.4265 0.0048 3.3907 71.6%
    1024 0.0076 4.2845 0.0048 6.8725 62.3%
    4096 0.0267 4.9025 0.0095 13.8378 35.4%
    16384 0.1327 3.9499 0.0692 7.5782 52.1%
    65536 0.6427 3.2630 0.2992 7.0096 46.6%

    dswap-stride256 GB/s chart

    dswap-stride256 ms chart

    See also:

    Settings
    Member Visibility

    On This Page
    Nvidia Geforce Gtx 1650 — wgblas vs cuBLASSee alsoStride sweep