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

Benchmark results for ssyr on Nvidia Geforce Gtx 1650.

Nvidia Geforce Gtx 1650 — wgblas vs cuBLAS

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0076	0.5751	0.0041	1.0625	54.1%
64	0.0076	2.2185	0.0041	4.1250	53.8%
128	0.0080	8.3702	0.0041	16.2500	51.5%
256	0.0086	30.6914	0.0044	60.7059	50.6%
512	0.0118	89.5129	0.0077	135.9339	65.9%
1024	0.0572	73.4086	0.0342	122.7364	59.8%
1280	0.0938	69.9710	0.0715	91.8174	76.2%
2048	0.2246	74.7738	0.1247	134.6504	55.5%
4096	0.8136	82.5241	0.4865	138.0016	59.8%

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

See also

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

Uplo sweep

Unless noted otherwise, every result above uses uplo = "lower". Real workgroups dispatch in increasing index order, so uplo = "upper" front-loads the heaviest rows first (worse — long-running heavy workgroups have nothing to overlap with) while lower back-loads them (better — light rows clear fast, the heavy tail gets full GPU to itself) — collapsed below by default, expand a uplo value to see its table and chart.

Nvidia Geforce Gtx 1650 — uplo = lower

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0078	0.5608	0.0041	1.0667	52.6%
64	0.0076	2.2185	0.0041	4.1250	53.8%
128	0.0081	8.2540	0.0041	16.2500	50.8%
256	0.0092	28.8671	0.0044	59.8261	48.3%
512	0.0119	88.6685	0.0079	133.1822	66.6%
1024	0.0365	115.1495	0.0345	121.9387	94.4%
1280	0.0540	121.4446	0.0716	91.7147	132.4%
2048	0.1208	138.9831	0.1251	134.2886	103.5%
4096	0.4443	151.1164	0.4867	137.9517	109.5%

Nvidia Geforce Gtx 1650 — uplo = upper

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0075	0.5824	0.0040	1.0924	53.3%
64	0.0077	2.1863	0.0041	4.1089	53.2%
128	0.0078	8.4898	0.0045	14.6479	58.0%
256	0.0084	31.3916	0.0044	60.7059	51.7%
512	0.0120	87.6059	0.0079	133.1822	65.8%
1024	0.0558	75.3676	0.0350	119.9342	62.8%
1280	0.0922	71.2222	0.0701	93.6194	76.1%
2048	0.2245	74.8165	0.1199	140.0587	53.4%
4096	0.8049	83.4198	0.4669	143.7944	58.0%

See also:

• uplo.ssyr.js — WebGPU uplo-sweep benchmark script
• uplo.ssyr.c — CUDA / cuBLAS uplo-sweep reference script

Lda sweep

Unless noted otherwise, every result above uses a tight lda (no padding). Padding the row stride changes throughput here — the exact mechanism and shape of that effect is routine-specific — collapsed below by default, expand a pad value to see its table and chart.

Nvidia Geforce Gtx 1650 — pad = 0

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0075	0.5787	0.0041	1.0709	54.0%
64	0.0078	2.1728	0.0041	4.1250	52.7%
128	0.0080	8.2704	0.0041	16.2500	50.9%
256	0.0090	29.3808	0.0044	60.2628	48.8%
512	0.0119	88.7881	0.0076	137.9287	64.4%
1024	0.0367	114.4969	0.0343	122.3933	93.5%
1280	0.0538	122.0226	0.0716	91.6328	133.2%
2048	0.1208	138.9831	0.1249	134.4435	103.4%
4096	0.4428	151.6187	0.4862	138.0879	109.8%

Nvidia Geforce Gtx 1650 — pad = 1

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0076	0.5726	0.0041	1.0709	53.5%
64	0.0076	2.2138	0.0041	4.1250	53.7%
128	0.0082	8.1250	0.0041	16.2500	50.0%
256	0.0091	28.9177	0.0048	55.2241	52.4%
512	0.0123	85.6667	0.0082	128.5000	66.7%
1024	0.0594	70.7586	0.0432	97.1720	72.8%
1280	0.0967	67.8644	0.0800	82.0973	82.7%
2048	0.2260	74.3080	0.1516	110.7757	67.1%
4096	0.8512	78.8832	0.7779	86.3127	91.4%

Nvidia Geforce Gtx 1650 — pad = 8

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0073	0.5939	0.0040	1.0794	55.0%
64	0.0072	2.3363	0.0041	4.1575	56.2%
128	0.0077	8.5950	0.0041	16.2500	52.9%
256	0.0086	30.8060	0.0045	58.9714	52.2%
512	0.0113	92.7955	0.0076	137.6402	67.4%
1024	0.0593	70.8159	0.0470	89.4605	79.2%
1280	0.0942	69.6857	0.0746	87.9400	79.2%
2048	0.2171	77.3414	0.1787	93.9828	82.3%
4096	0.8072	83.1751	0.7619	88.1290	94.4%

Nvidia Geforce Gtx 1650 — pad = 16

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0070	0.6182	0.0035	1.2593	49.1%
64	0.0062	2.7216	0.0035	4.8664	55.9%
128	0.0067	10.0000	0.0035	19.0826	52.4%
256	0.0077	34.4000	0.0040	66.5807	51.7%
512	0.0100	105.7749	0.0061	172.2304	61.4%
1024	0.0554	75.8683	0.0444	94.5486	80.2%
1280	0.0860	76.3095	0.0687	95.5602	79.9%
2048	0.2086	80.4970	0.1409	119.1509	67.6%
4096	0.7577	88.6129	0.7432	90.3471	98.1%

Nvidia Geforce Gtx 1650 — pad = 32

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0061	0.7083	0.0036	1.2197	58.1%
64	0.0061	2.7500	0.0035	4.8440	56.8%
128	0.0065	10.2970	0.0037	17.7778	57.9%
256	0.0076	34.8354	0.0039	67.1219	51.9%
512	0.0096	110.0201	0.0061	172.6824	63.7%
1024	0.0343	122.6791	0.0384	109.5770	112.0%
1280	0.0512	128.2000	0.0580	113.1073	113.3%
2048	0.1151	145.8588	0.1374	122.1886	119.4%
4096	0.4480	149.8590	0.6183	108.5900	138.0%

Nvidia Geforce Gtx 1650 — pad = 48

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0061	0.7083	0.0035	1.2477	56.8%
64	0.0061	2.7500	0.0035	4.8889	56.2%
128	0.0065	10.1961	0.0035	18.8235	54.2%
256	0.0076	34.9091	0.0039	68.2314	51.2%
512	0.0097	108.5677	0.0061	171.3333	63.4%
1024	0.0556	75.6498	0.0451	93.2727	81.1%
1280	0.0853	76.9248	0.0690	95.0950	80.9%
2048	0.2126	78.9945	0.1636	102.6504	77.0%
4096	0.7858	85.4480	0.8350	80.4099	106.3%

Nvidia Geforce Gtx 1650 — pad = 64

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0061	0.7083	0.0034	1.2651	56.0%
64	0.0061	2.7500	0.0035	4.8889	56.2%
128	0.0065	10.2970	0.0035	19.1705	53.7%
256	0.0076	34.7621	0.0038	69.3782	50.1%
512	0.0096	109.2890	0.0061	171.3333	63.8%
1024	0.0348	120.7059	0.0406	103.5302	116.6%
1280	0.0492	133.3680	0.0573	114.5922	116.4%
2048	0.1192	140.8670	0.1469	114.3480	123.2%
4096	0.4432	151.4983	0.5932	113.1855	133.8%

Nvidia Geforce Gtx 1650 — pad = 128

n	wgblas ms	wgblas GB/s	cuBLAS ms	cuBLAS GB/s	efficiency
32	0.0061	0.7083	0.0035	1.2364	57.3%
64	0.0061	2.7500	0.0036	4.7354	58.1%
128	0.0064	10.3226	0.0036	18.6547	55.3%
256	0.0077	34.2573	0.0038	69.9661	49.0%
512	0.0097	108.2105	0.0061	171.3333	63.2%
1024	0.0343	122.6218	0.0410	102.4399	119.7%
1280	0.0512	128.2000	0.0615	106.6667	120.2%
2048	0.1193	140.7537	0.1817	92.4188	152.3%
4096	0.4855	138.2835	0.6323	106.1884	130.2%

See also:

• lda.ssyr.js — WebGPU lda-sweep benchmark script
• lda.ssyr.c — CUDA / cuBLAS lda-sweep reference script