wgblas
    Preparing search index...

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

    Benchmark results for dasum on Nvidia Geforce Gtx 1650.

    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0241 0.0106 0.1174 0.0022 482.0%
    64 0.0233 0.0219 0.1179 0.0043 510.1%
    128 0.0238 0.0430 0.1191 0.0086 499.8%
    512 0.0236 0.1733 0.1174 0.0349 496.6%
    1024 0.0233 0.3516 0.1230 0.0666 528.0%
    4096 0.0240 1.3635 0.0182 1.7965 75.9%
    16384 0.0249 5.2716 0.1247 1.0511 501.5%
    65536 0.0269 19.4584 0.0181 29.0239 67.0%
    262144 0.0407 51.4815 0.0355 59.0414 87.2%
    1048576 0.0763 109.9828 0.0773 108.5033 101.4%
    4194304 0.2158 155.5059 0.2405 139.5218 111.5%
    16777216 0.7908 169.7240 0.8796 152.5923 111.2%

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

    dasum-default GB/s chart

    dasum-default ms chart

    • dasum.js — WebGPU benchmark script
    • dasum.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 0.0241 0.0106 0.0141 0.0181 58.8%
    64 0.0237 0.0216 0.0142 0.0360 60.0%
    128 0.0238 0.0430 0.0218 0.0469 91.7%
    512 0.0237 0.1727 0.0224 0.1827 94.5%
    1024 0.0241 0.3393 0.0222 0.3689 92.0%
    4096 0.0244 1.3421 0.0218 1.5059 89.1%
    16384 0.0276 4.7407 0.0223 5.8851 80.6%
    65536 0.0392 13.3911 0.0339 15.4639 86.6%
    262144 0.0727 28.8387 0.0669 31.3569 92.0%
    1048576 0.2071 40.5043 0.1980 42.3770 95.6%
    4194304 0.7455 45.0110 0.7355 45.6230 98.7%

    dasum-stride4 GB/s chart

    dasum-stride4 ms chart

    Nvidia Geforce Gtx 1650 — stride = 32
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    32 0.0215 0.0119 0.0182 0.0141 84.5%
    64 0.0214 0.0239 0.0181 0.0282 84.7%
    128 0.0211 0.0485 0.0181 0.0565 85.9%
    512 0.0204 0.2011 0.0182 0.2250 89.4%
    1024 0.0204 0.4016 0.0181 0.4531 88.6%
    4096 0.0253 1.2937 0.0180 1.8172 71.2%
    16384 0.0337 3.8843 0.0212 6.1873 62.8%
    65536 0.0684 7.6686 0.0397 13.2076 58.1%
    262144 0.2046 10.2488 0.1072 19.5601 52.4%

    dasum-stride32 GB/s chart

    dasum-stride32 ms chart

    Nvidia Geforce Gtx 1650 — stride = 256
    n wgblas ms wgblas GB/s cuBLAS ms cuBLAS GB/s efficiency
    512 0.0222 0.1848 0.0173 0.2368 78.1%
    1024 0.0239 0.3434 0.0436 0.1880 182.7%
    4096 0.0287 1.1429 0.0473 0.6928 165.0%
    16384 0.0451 2.9091 0.0570 2.2985 126.6%
    65536 0.1004 5.2245 0.0497 10.5533 49.5%

    dasum-stride256 GB/s chart

    dasum-stride256 ms chart

    See also:

    Settings
    Member Visibility

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