wgblas
    Preparing search index...

    Module benchmarks/nvidia-geforce-gtx-1650/snrm2

    Benchmark results for snrm2 on Nvidia Geforce Gtx 1650.

    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0189 0.0068 0.0181 0.0071 95.6%
    64 0.0193 0.0133 0.0178 0.0143 92.8%
    128 0.0189 0.0271 0.0178 0.0287 94.6%
    512 0.0192 0.1066 0.0179 0.1144 93.2%
    1024 0.0192 0.2128 0.0181 0.2267 93.9%
    4096 0.0190 0.8605 0.0182 0.8982 95.8%
    16384 0.0196 3.3437 0.0182 3.5930 93.1%
    65536 0.0198 13.2664 0.0179 14.6810 90.4%
    262144 0.0256 40.9089 0.0185 56.6430 72.2%
    1048576 0.0454 92.3042 0.0448 93.5560 98.7%
    4194304 0.1160 144.6711 0.1433 117.1070 123.5%
    16777216 0.4066 165.0456 0.6264 107.1342 154.1%

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

    snrm2-default GB/s chart

    snrm2-default ms chart

    • snrm2.js — WebGPU benchmark script
    • snrm2.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 0.0190 0.0067 0.0188 0.0068 98.9%
    64 0.0192 0.0133 0.0188 0.0136 98.0%
    128 0.0190 0.0269 0.0188 0.0273 98.6%
    512 0.0192 0.1067 0.0187 0.1096 97.3%
    1024 0.0190 0.2153 0.0188 0.2184 98.6%
    4096 0.0189 0.8649 0.0187 0.8752 98.8%
    16384 0.0194 3.3795 0.0181 3.6248 93.2%
    65536 0.0246 10.6667 0.0190 13.8145 77.2%
    262144 0.0426 24.6283 0.0442 23.7277 103.8%
    1048576 0.1104 37.9919 0.1074 39.0677 97.2%
    4194304 0.3783 44.3523 0.3766 44.5501 99.6%

    snrm2-stride4 GB/s chart

    snrm2-stride4 ms chart

    Nvidia Geforce Gtx 1650 — stride = 32
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0192 0.0067 0.0184 0.0070 95.1%
    64 0.0188 0.0136 0.0182 0.0141 96.4%
    128 0.0189 0.0271 0.0180 0.0284 95.5%
    512 0.0186 0.1101 0.0180 0.1137 96.8%
    1024 0.0191 0.2142 0.0184 0.2220 96.5%
    4096 0.0196 0.8373 0.0180 0.9078 92.2%
    16384 0.0276 2.3786 0.0388 1.6898 140.8%
    65536 0.0433 6.0480 0.0572 4.5868 131.9%
    262144 0.1120 9.3609 0.1213 8.6448 108.3%
    1048576 0.3836 10.9336 0.3888 10.7865 101.4%

    snrm2-stride32 GB/s chart

    snrm2-stride32 ms chart

    Nvidia Geforce Gtx 1650 — stride = 256
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    512 0.0193 0.1063 0.0174 0.1179 90.2%
    1024 0.0205 0.2000 0.0181 0.2261 88.5%
    4096 0.0225 0.7273 0.0181 0.9046 80.4%
    16384 0.0329 1.9932 0.0497 1.3187 151.1%
    65536 0.0599 4.3796 0.0738 3.5525 123.3%

    snrm2-stride256 GB/s chart

    snrm2-stride256 ms chart

    See also:

    Settings
    Member Visibility

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