Hi All,
Earlier I posted provisional benchmarks for AMBER 18 on RTX2080. Since then some initial tweaks from Dave Cerutti for the Turin architecture have been suggested. The following are the RTX2080 performance numbers for AMBER 18 with these tweaks for both the original benchmark set and the new benchmark set. As you can see with these tweaks RTX2080 comes out about 15 to 20% above 1080TI in terms of performance. You can expect a patch with these tweaks to come out for AMBER 18 shortly.
I have also run my GPU validation suite on 16 RTX2080 cards now for a total of 72 hours and have seen no failures so it looks like, for once, the newly released design is reliable. :-)
All numbers are for 4 x RTX2080 with the custom cooling solution developed by Exxact. You will also get these numbers with two stock RTX2080 cards as long as they have a space between them.
New Benchmark Suite with thread count tweak for Turing
Report of all timings:
System Class CPU (1) CPU (4) GPU 0 GPU 1 GPU 2 GPU 3
------------------------------- ---------------- -------- -------- -------- -------- -------- --------
JAC_production_NVE_4fs PME (Standard) 748.76 722.91 745.35 750.98
JAC_production_NPT_4fs PME (Standard) 691.42 675.24 689.72 690.97
Cellulose_production_NVE_4fs PME (Standard) 48.22 46.91 47.88 47.97
Cellulose_production_NPT_4fs PME (Standard) 46.05 45.31 45.55 45.71
FactorIX_production_NVE_4fs PME (Standard) 233.29 228.91 229.80 231.10
FactorIX_production_NPT_4fs PME (Standard) 221.23 216.76 216.12 218.17
STMV_production_NVE_4fs PME (Standard) 16.75 16.49 16.63 16.68
STMV_production_NPT_4fs PME (Standard) 15.73 15.50 15.64 15.63
JAC_production_NVE_4fs PME (Optimized) 837.06 809.03 829.88 829.66
JAC_production_NPT_4fs PME (Optimized) 772.66 745.39 771.34 770.48
Cellulose_production_NVE_4fs PME (Optimized) 51.42 50.42 51.06 51.12
Cellulose_production_NPT_4fs PME (Optimized) 48.80 48.09 48.40 48.51
FactorIX_production_NVE_4fs PME (Optimized) 256.83 253.39 253.89 254.53
FactorIX_production_NPT_4fs PME (Optimized) 239.83 236.58 235.94 239.37
STMV_production_NVE_4fs PME (Optimized) 18.13 17.87 18.03 18.07
STMV_production_NPT_4fs PME (Optimized) 16.99 16.72 16.86 16.91
TRPCage GB 2400.31 2406.72 2393.68 2402.95
myoglobin GB 995.92 988.09 1017.04 1005.40
nucleosome GB 21.05 20.35 20.67 20.97
Original Benchmark Suite with thread count tweak for Turing
JAC_PRODUCTION_NVE - 23,558 atoms PME 4fs
-----------------------------------------
[0] 1 x GPU: | ns/day = 831.85 seconds/ns = 103.86
[1] 1 x GPU: | ns/day = 836.49 seconds/ns = 103.29
[2] 1 x GPU: | ns/day = 838.32 seconds/ns = 103.06
[3] 1 x GPU: | ns/day = 833.40 seconds/ns = 103.67
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 829.60 seconds/ns = 104.15
[1] 1 x GPU: | ns/day = 831.42 seconds/ns = 103.92
[2] 1 x GPU: | ns/day = 835.46 seconds/ns = 103.42
[3] 1 x GPU: | ns/day = 816.44 seconds/ns = 105.82
JAC_PRODUCTION_NPT - 23,558 atoms PME 4fs
-----------------------------------------
[0] 1 x GPU: | ns/day = 773.02 seconds/ns = 111.77
[1] 1 x GPU: | ns/day = 771.10 seconds/ns = 112.05
[2] 1 x GPU: | ns/day = 771.98 seconds/ns = 111.92
[3] 1 x GPU: | ns/day = 778.33 seconds/ns = 111.01
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 768.56 seconds/ns = 112.42
[1] 1 x GPU: | ns/day = 770.58 seconds/ns = 112.12
[2] 1 x GPU: | ns/day = 767.93 seconds/ns = 112.51
[3] 1 x GPU: | ns/day = 762.27 seconds/ns = 113.35
JAC_PRODUCTION_NVE - 23,558 atoms PME 2fs
-----------------------------------------
[0] 1 x GPU: | ns/day = 437.66 seconds/ns = 197.42
[1] 1 x GPU: | ns/day = 442.53 seconds/ns = 195.24
[2] 1 x GPU: | ns/day = 444.05 seconds/ns = 194.57
[3] 1 x GPU: | ns/day = 441.92 seconds/ns = 195.51
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 437.12 seconds/ns = 197.66
[1] 1 x GPU: | ns/day = 441.28 seconds/ns = 195.80
[2] 1 x GPU: | ns/day = 440.08 seconds/ns = 196.33
[3] 1 x GPU: | ns/day = 432.34 seconds/ns = 199.84
JAC_PRODUCTION_NPT - 23,558 atoms PME 2fs
-----------------------------------------
[0] 1 x GPU: | ns/day = 397.27 seconds/ns = 217.48
[1] 1 x GPU: | ns/day = 401.54 seconds/ns = 215.17
[2] 1 x GPU: | ns/day = 402.16 seconds/ns = 214.84
[3] 1 x GPU: | ns/day = 400.22 seconds/ns = 215.88
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 396.28 seconds/ns = 218.03
[1] 1 x GPU: | ns/day = 397.27 seconds/ns = 217.48
[2] 1 x GPU: | ns/day = 400.26 seconds/ns = 215.86
[3] 1 x GPU: | ns/day = 391.15 seconds/ns = 220.89
FACTOR_IX_PRODUCTION_NVE - 90,906 atoms PME
-------------------------------------------
[0] 1 x GPU: | ns/day = 146.87 seconds/ns = 588.28
[1] 1 x GPU: | ns/day = 147.97 seconds/ns = 583.91
[2] 1 x GPU: | ns/day = 148.52 seconds/ns = 581.73
[3] 1 x GPU: | ns/day = 143.61 seconds/ns = 601.65
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 147.36 seconds/ns = 586.32
[1] 1 x GPU: | ns/day = 147.67 seconds/ns = 585.09
[2] 1 x GPU: | ns/day = 148.20 seconds/ns = 583.01
[3] 1 x GPU: | ns/day = 146.32 seconds/ns = 590.48
FACTOR_IX_PRODUCTION_NPT - 90,906 atoms PME
-------------------------------------------
[0] 1 x GPU: | ns/day = 132.57 seconds/ns = 651.73
[1] 1 x GPU: | ns/day = 133.23 seconds/ns = 648.52
[2] 1 x GPU: | ns/day = 133.07 seconds/ns = 649.27
[3] 1 x GPU: | ns/day = 135.68 seconds/ns = 636.81
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 135.18 seconds/ns = 639.14
[1] 1 x GPU: | ns/day = 136.64 seconds/ns = 632.34
[2] 1 x GPU: | ns/day = 137.10 seconds/ns = 630.18
[3] 1 x GPU: | ns/day = 135.36 seconds/ns = 638.32
CELLULOSE_PRODUCTION_NVE - 408,609 atoms PME
--------------------------------------------
[0] 1 x GPU: | ns/day = 29.66 seconds/ns = 2912.60
[1] 1 x GPU: | ns/day = 29.85 seconds/ns = 2894.79
[2] 1 x GPU: | ns/day = 29.93 seconds/ns = 2886.58
[3] 1 x GPU: | ns/day = 29.74 seconds/ns = 2904.98
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 29.77 seconds/ns = 2902.52
[1] 1 x GPU: | ns/day = 29.77 seconds/ns = 2902.17
[2] 1 x GPU: | ns/day = 29.77 seconds/ns = 2901.97
[3] 1 x GPU: | ns/day = 29.37 seconds/ns = 2941.30
CELLULOSE_PRODUCTION_NPT - 408,609 atoms PME
--------------------------------------------
[0] 1 x GPU: | ns/day = 27.60 seconds/ns = 3130.86
[1] 1 x GPU: | ns/day = 27.71 seconds/ns = 3117.51
[2] 1 x GPU: | ns/day = 27.89 seconds/ns = 3098.04
[3] 1 x GPU: | ns/day = 27.64 seconds/ns = 3125.70
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 27.59 seconds/ns = 3131.05
[1] 1 x GPU: | ns/day = 27.75 seconds/ns = 3113.85
[2] 1 x GPU: | ns/day = 27.69 seconds/ns = 3120.74
[3] 1 x GPU: | ns/day = 27.42 seconds/ns = 3151.41
STMV_PRODUCTION_NPT - 1,067,095 atoms PME
-----------------------------------------
[0] 1 x GPU: | ns/day = 17.34 seconds/ns = 4981.60
[1] 1 x GPU: | ns/day = 17.42 seconds/ns = 4960.24
[2] 1 x GPU: | ns/day = 17.63 seconds/ns = 4899.74
[3] 1 x GPU: | ns/day = 17.37 seconds/ns = 4973.73
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 17.40 seconds/ns = 4966.91
[1] 1 x GPU: | ns/day = 17.46 seconds/ns = 4949.82
[2] 1 x GPU: | ns/day = 17.47 seconds/ns = 4945.47
[3] 1 x GPU: | ns/day = 17.26 seconds/ns = 5004.35
TRPCAGE_PRODUCTION - 304 atoms GB
---------------------------------
[0] 1 x GPU: | ns/day = 1114.36 seconds/ns = 77.53
[1] 1 x GPU: | ns/day = 1130.30 seconds/ns = 76.44
[2] 1 x GPU: | ns/day = 1117.49 seconds/ns = 77.32
[3] 1 x GPU: | ns/day = 1152.58 seconds/ns = 74.96
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 1104.43 seconds/ns = 78.23
[1] 1 x GPU: | ns/day = 1152.39 seconds/ns = 74.97
[2] 1 x GPU: | ns/day = 1131.04 seconds/ns = 76.39
[3] 1 x GPU: | ns/day = 1133.21 seconds/ns = 76.24
MYOGLOBIN_PRODUCTION - 2,492 atoms GB
-------------------------------------
[0] 1 x GPU: | ns/day = 482.52 seconds/ns = 179.06
[1] 1 x GPU: | ns/day = 469.51 seconds/ns = 184.02
[2] 1 x GPU: | ns/day = 484.78 seconds/ns = 178.23
[3] 1 x GPU: | ns/day = 482.66 seconds/ns = 179.01
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 462.80 seconds/ns = 186.69
[1] 1 x GPU: | ns/day = 473.04 seconds/ns = 182.65
[2] 1 x GPU: | ns/day = 454.35 seconds/ns = 190.16
[3] 1 x GPU: | ns/day = 465.19 seconds/ns = 185.73
NUCLEOSOME_PRODUCTION - 25,095 atoms GB
---------------------------------------
[0] 1 x GPU: | ns/day = 11.13 seconds/ns = 7760.44
[1] 1 x GPU: | ns/day = 11.25 seconds/ns = 7678.77
[2] 1 x GPU: | ns/day = 11.31 seconds/ns = 7636.90
[3] 1 x GPU: | ns/day = 11.14 seconds/ns = 7758.06
Multiple Single GPU Run Performance
[0] 1 x GPU: | ns/day = 11.10 seconds/ns = 7785.77
[1] 1 x GPU: | ns/day = 11.18 seconds/ns = 7726.95
[2] 1 x GPU: | ns/day = 11.29 seconds/ns = 7655.40
[3] 1 x GPU: | ns/day = 10.99 seconds/ns = 7863.05
All the best
Ross
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Oct 10 2018 - 20:30:02 PDT
