wgblas
    Preparing search index...

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

    Benchmark results for sdot on Nvidia Geforce Gtx 1650.

    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0152 0.0168 0.2054 0.0012 1399.1%
    64 0.0149 0.0344 0.0168 0.0306 112.4%
    128 0.0147 0.0696 0.0103 0.0994 70.1%
    512 0.0149 0.2750 0.0104 0.3938 69.8%
    1024 0.0149 0.5505 0.0103 0.7926 69.5%
    4096 0.0149 2.1974 0.0180 1.8172 120.9%
    16384 0.0200 6.5694 0.0181 7.2560 90.5%
    65536 0.0212 24.6747 0.0182 28.7944 85.7%
    262144 0.0334 62.7439 0.0269 78.0655 80.4%
    1048576 0.0684 122.6118 0.0609 137.6445 89.1%
    4194304 0.2051 163.6227 0.1955 171.6726 95.3%
    16777216 0.7516 178.5722 0.7337 182.9217 97.6%

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

    sdot-default GB/s chart

    sdot-default ms chart

    • sdot.js — WebGPU benchmark script
    • sdot.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 0.0197 0.0130 0.0187 0.0137 95.0%
    64 0.0197 0.0260 0.0188 0.0273 95.4%
    128 0.0198 0.0518 0.0187 0.0547 94.7%
    512 0.0200 0.2053 0.0188 0.2182 94.1%
    1024 0.0197 0.4159 0.0187 0.4384 94.9%
    4096 0.0205 1.6013 0.0186 1.7595 91.0%
    16384 0.0222 5.9020 0.0191 6.8610 86.0%
    65536 0.0346 15.1634 0.0303 17.2827 87.7%
    262144 0.0684 30.6673 0.0636 32.9492 93.1%
    1048576 0.2081 40.3112 0.1972 42.5351 94.8%
    4194304 0.7779 43.1344 0.7319 45.8434 94.1%

    sdot-stride4 GB/s chart

    sdot-stride4 ms chart

    Nvidia Geforce Gtx 1650 — stride = 32
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0186 0.0137 0.0102 0.0251 54.8%
    64 0.0178 0.0287 0.0172 0.0297 96.7%
    128 0.0183 0.0558 0.0173 0.0591 94.5%
    512 0.0176 0.2332 0.0180 0.2274 102.5%
    1024 0.0175 0.4693 0.0179 0.4567 102.8%
    4096 0.0214 1.5295 0.0180 1.8156 84.2%
    16384 0.0318 4.1270 0.0287 4.5740 90.2%
    65536 0.0652 8.0452 0.0593 8.8467 90.9%
    262144 0.2012 10.4249 0.1949 10.7604 96.9%
    1048576 0.9130 9.1880 0.7303 11.4862 80.0%

    sdot-stride32 GB/s chart

    sdot-stride32 ms chart

    Nvidia Geforce Gtx 1650 — stride = 256
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    512 0.0286 0.1434 0.0174 0.2357 60.8%
    1024 0.0272 0.3014 0.0180 0.4563 66.0%
    4096 0.0306 1.0723 0.0182 1.8028 59.5%
    16384 0.0449 2.9195 0.0313 4.1860 69.7%
    65536 0.0974 5.3842 0.0842 6.2261 86.5%

    sdot-stride256 GB/s chart

    sdot-stride256 ms chart

    See also:

    Settings
    Member Visibility

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