Skip to content

Latest commit

 

History

History
235 lines (198 loc) · 10.3 KB

benchmarks.md

File metadata and controls

235 lines (198 loc) · 10.3 KB
Raw

RTX2080 System: GNU 12.2.1 GPU Driver Version: 525.89.02 CUDA Version: 12.0 Linux 6.12-1

PowerMeasurementValues are averages, using a 20ms period on the RTX2080. On the Stratix10, the period is around 140ms, limited by the power-measurement tool provided by the Paderborn Devcloud Tool.

CFD FP32

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 1.6392ms 3.0584ms 0.55210s
RTX 2080 SYCL 1.6937ms 3.2658ms 0.58866s
A100 SYCL 1.1602ms 1.6271ms 0.20142s
Xeon6128 SYCL 19.0318ms 33.3050ms 9.72068s
Epyc 7763 9.0651ms 14.8829ms 2.17930s
P630 SYCL 11.0341ms 33.6172ms 7.08886s
Stratix10 123.1330ms 254.9750ms 56.64280s
Stratix10 OPT 30.3324ms 60.5630ms 11.94730s
Agilex OPT 17.4967ms 33.2780ms 6.11269s
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 46.5 W 46.7 W 141.1 W
Stratix10 OPT 75.0 W 75.0 W 78.0 W

CFD FP64

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 3.7124s 7.3204s 36.5047s
RTX 2080 SYCL 2.4742s 4.8795s 24.4830s
A100 SYCL 0.2246s 0.4599s 2.1434s
Xeon6128 SYCL 4.0683s 11.3628s 73.9684s
Epyc 7763 2.2029s 4.2034s 41.5361s
P630 SYCL 8.5755s 18.3472s 92.4479s
Stratix10 25.5076s 53.1037s 258.2010s
Stratix10 OPT 11.9330s 24.0210s 120.0660s
Agilex OPT 10.9704s 21.4182s 109.4120s
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 46.8 W 50.3 W 55.1 W
Stratix10 OPT 83.0 W 84.0 W 85.0 W

FDTD2D

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 1.490ms 63.207ms 0.533686s
RTX 2080 SYCL 4.565ms 68.323ms 0.539139s
A100 SYCL 8.403ms 28.590ms 0.189980s
Xeon6128 SYCL 122.521ms 769.523ms 7.715659s
Epyc 7763 20.776ms 344.787ms 2.394675s
P630 SYCL 97.493ms 947.787ms 8.055991s
Stratix10 107.203ms 3204.410ms 25.570672s
Stratix10 OPT 18.309ms 585.217ms 4.781038s
Agilex OPT 13.144ms 543.770ms 4.984691s
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 46.2 W 121.3 W 166.4 W
Stratix10 OPT 68.0 W 81.0 W 81.0 W

k-Means

Device 2048 Pts 10'000 Pts 65'536 Pts
RTX 2080 CUDA 34.5562µs 0.05515ms 0.2263ms
RTX 2080 SYCL 80.4439µs 0.08164ms 0.2333ms
A100 SYCL 125.2050µs 0.15579ms 0.3333ms
Xeon6128 SYCL 902.5244µs 3.68568ms 23.2675ms
Epyc 7763 499.1150µs 2.06600ms 12.5059ms
P630 SYCL 832.8460µs 2.48346ms 12.7533ms
Stratix10 15587.3000µs 70.83460ms 463.1910ms
Stratix10 OPT 31.8503µs 0.14153ms 0.9078ms
Agilex OPT 31.5282µs 0.13906ms 0.8896ms
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 49.3 W 54.6 W 139.1 W
Stratix10 OPT 69.0 W 68.0 W 74.0 W

LavaMD

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 170.016µs 35.2951ms 19.2689s
RTX 2080 SYCL 214.721µs 34.1074ms 18.2843s
A100 SYCL 70.144µs 13.9520ms 3.9906s
Xeon6128 SYCL 119.109µs 43.7124ms 22.5764s
Epyc 7763 97.986µs 17.7175ms 10.1274s
P630 SYCL 115.958µs 111.6950ms 50.6277s
Stratix10 113.544µs 371.8460ms 252.0760s
Stratix10 OPT 31.205µs 16.0696ms 10.0122s
Agilex OPT 22.365µs 15.1253ms 9.6560s
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 37.7 W 85.3 W 96.4 W
Stratix10 OPT 76.0 W 83.0 W 111.0 W

Mandelbrot

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 0.6922ms 6.5436ms 87.556ms
RTX 2080 SYCL 0.5878ms 6.1205ms 86.247ms
A100 SYCL 0.4869ms 5.0324ms 39.677ms
Xeon6128 SYCL 10.4500ms 73.1730ms 974.425ms
Epyc 7763 3.7154ms 26.5012ms 349.938ms
P630 SYCL 17.4506ms 204.3280ms 2794.370ms
Stratix10 843.4750ms 10587.9000ms 170603.000ms
Stratix10 OPT 3.5146ms 22.5326ms 358.278ms
Agilex OPT 2.9235ms 25.4920ms 494.123ms
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 67.9 W 83.6 W 202.4 W
Stratix10 OPT 83.0 W 90.0 W 105.0 W

NeedleWunsch

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 75.936µs 0.89150ms 9.224ms
RTX 2080 SYCL 77.695µs 0.92410ms 7.646ms
A100 SYCL 178.177µs 2.02500ms 13.938ms
Xeon6128 SYCL 295.128µs 4.03906ms 40.2501ms
Epyc 7763 162.407µs 1.89001ms 19.7037ms
P630 SYCL 462.747µs 5.55169ms 87.5200ms
Stratix10 1205.280µs 105.29600ms 1422.2000ms
Stratix10 OPT 215.620µs 5.80730ms 80.6206ms
Agilex OPT 105.960µs 3.49629ms 51.8175ms
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 46.4 W 45.8 W 46.4 W
Stratix10 OPT 72.0 W 73.0 W 73.0 W

Particlefilter Naive

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 18.2080µs 291.120µs 6.2020s
RTX 2080 SYCL 16.5435µs 251.570µs 5.9500s
A100 SYCL 41.9830µs 416.253µs 1.0152s
Xeon6128 SYCL 7.7475µs 647.591µs 14.1067s
Epyc 7763 14.0500µs 230.814µs 8.0357s
P630 SYCL 25.7915µs 448.291µs 75.2312s
Stratix10 42.5795µs 2921.960µs 5533.3200s
Stratix10 OPT 52.7855µs 200.205µs 20.2994s
Agilex OPT 92.1390µs 421.022µs 34.0956s
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 46.5 W 49.3 W 117.8 W
Stratix10 OPT 66.0 W 73.0 W 73.0 W

Particlefilter Float

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 132.336µs 1.86836ms 6.53996s
RTX 2080 SYCL 143.950µs 1.58866ms 6.74448s
A100 SYCL 238.533µs 1.47505ms 0.96902s
Xeon6128 SYCL 518.178µs 2.73748ms 31.29600s
Epyc 7763 354.464µs 1.51298ms 16.17800s
P630 SYCL 555.238µs 2.82049ms 80.28850s
Stratix10 629.738µs 10.04030ms 7765.62000s
Stratix10 OPT 152.785µs 0.87129ms 21.10000s
Agilex OPT 273.883µs 1.96399ms 39.00970s
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 46.0 W 46.6 W 114.9 W
Stratix10 OPT 65.0 W 73.0 W 77.0 W

Raytracing

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 0.324203s 1.85996s 22.4497s
RTX 2080 SYCL 0.032864s 0.15042s 1.1606s
A100 SYCL 0.033057s 0.08531s 0.5536s
Xeon6128 SYCL 0.241077s 1.86069s 16.7465s
Epyc 7763 0.114243s 0.81676s 6.9161s
P630 SYCL 0.343128s 2.50398s 45.2825s
Stratix10 4.095220s 33.04290s 299.9090s
Stratix10 OPT 0.153362s 0.91916s 7.3717s
Agilex OPT 0.136367s 0.86583s 7.1678s
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 192.2 W 202.8 W 203.3 W
Stratix10 OPT 90.0 W 95.0 W 94.0 W

SRAD

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 0.06920ms 0.07430ms 0.04587s
RTX 2080 SYCL 0.06197ms 0.07143ms 0.04549s
A100 SYCL 0.12185ms 0.04608ms 0.02131s
Xeon6128 SYCL 1.15556ms 3.05370ms 0.78509s
Epyc 7763 0.70934ms 2.51152ms 0.31922s
P630 SYCL 0.33608ms 0.71717ms 0.70944s
Stratix10 1.01890ms 2.91834ms 5.58274s
Stratix10 OPT 0.48662ms 1.13396ms 1.03689s
Agilex OPT 0.31775ms 1.45761ms 1.26734s
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 46.5 W 48.2 W 53.2 W
Stratix10 OPT 70.0 W 70.0 W 70.0 W

Where

Device Size 1 Size 2 Size 3
RTX 2080 CUDA 29.424µs 31.149µs 38.488ms
RTX 2080 SYCL 115.491µs 117.449µs 82.947ms
A100 SYCL 161.755µs 165.424µs 30.876ms
Xeon6128 SYCL 608.439µs 646.987µs 766.429ms
Epyc 7763 283.663µs 386.083µs 254.765ms
P630 SYCL 237.470µs 326.871µs 746.628ms
Stratix10 6373.370µs 7784.890µs 35000.292ms
Stratix10 OPT 70.215µs 92.390µs 1044.170ms
Agilex OPT 46.360µs 68.970µs CRASH
Device Size 1 Size 2 Size 3
RTX 2080 CUDA 46.8 W 46.8 W 58.9 W
Stratix10 OPT 68.0 W 68.0 W 72.0 W