Slide 1: Molecular Dynamics (MD) on GPUs
Molecular Dynamics (MD) on GPUs March 2019
Slide 2: Accelerating Discoveries
Accelerating Discoveries Using a supercomputer powered by the Tesla Platform with over 3,000 Tesla accelerators, University of Illinois scientists performed the first all-atom simulation of the HIV virus and discovered the chemical structure of its capsid — “the perfect target for fighting the infec
Slide 3: Overview of Life & Material Accelerated Apps
Overview of Life & Material Accelerated Apps MD QC All key codes are GPU-accelerated All key codes are ported or optimizing Great multi-GPU, multi-node (dense) performance GPU-accelerated math libraries, OpenACC directives GPU-accelerated apps GPU-accelerated apps ACEMD*, AMBER*, BAND, CHARMM, DESMO
Slide 4: MD vs. QC on GPUs
MD vs. QC on GPUs Molecular Dynamics Quantum Chemistry Properties - electronic properties, Simulates atomic positions over time Calculations ground state, excitation, spectra Chemical-biological or chemical-material Examples: MO, PW, DFT, semi-emp Simple empirical formulas Electron wave function For
Slide 5: GPU-Accelerated Molecular Dynamics Apps
GPU-Accelerated Molecular Dynamics Apps Performance Slides Available • DESMOND/FEP • HTMD ACEMD HOOMD-Blue • ESPResSO • mdcore AMBER/GTI LAMMPS • Folding@Home • MELD Chameleon NAMD • Genesis • OpenMM CHARMM • GPUGrid.net • PolyFTS GROMACS • HALMD GPU Perf compared against dual multi-core x86 CPU soc
Slide 6: MD Applications GPU-Accelerated Computing
MD Applications GPU-Accelerated Computing Turbocharge your research! • Speedup of 3X-8X compared to CPU only in all tests (average) • Majority of compute intensive for classical MD ported to GPUs • Large performance boost and improve TCO for compute infrastructure • Tesla GPUs are more energy effici
Slide 7: AmberMD 18.10-AT_18.12
AmberMD 18.10-AT_18.12 March 2019
Slide 8: AmberMD 18.10_AT_18.12 - PME-Cellulose
AmberMD 18.10_AT_18.12- PME-Cellulose AmberMD 18.10-AT_18.12 - Tesla V100-SXM2-32GB 90 30.0X 80 78.55 25.0X 70 71.11 60 63.19 20.0X Cellulose 57.68 58.54 408,609 atoms 50 Running AmberMD 18.10_AT_18.12 ns/day 48.13 15.0X 40 4X V100 2X V100 24.2X The blue node contains Dual Intel Xeon 4X V100 Gold 61
Slide 9: AmberMD 18.10_AT_18.12 - PME-FactorIX
AmberMD 18.10_AT_18.12 - PME-FactorIX AmberMD 18.10-AT_18.12 - Tesla V100-SXM2-32GB 350 25.0X 326.85 300 20.0X 290.8 Factor IX 90,906 atoms 250 268.08 262.56 236.39 15.0X 200 207.66 ns/day Running AmberMD 18.10_AT_18.12 150 4X V100 4X V100 2X V100 21.0X 10.0X The blue node contains Dual Intel Xeon 2
Slide 10: AmberMD 18.10_AT_18.12 - PME-JAC
AmberMD 18.10_AT_18.12 - PME-JAC AmberMD 18.10-AT_18.12 - Tesla V100-SXM2-32GB 800 14.0X 700 12.0X 687.83 600 622.91 591.28 10.0X DHFR 571.21 500 522.88 23,558 atoms 506.36 8.0X Running AmberMD 18.10_AT_18.12 ns/day 400 4X V100 1X V1002X V100 12.3X 6.0X 300 4X V100 The blue node contains Dual Intel
Slide 11: AmberMD 18.10_AT_18.12 - PME-STMV_NPT
AmberMD 18.10_AT_18.12 - PME-STMV_NPT AmberMD 18.10-AT_18.12 - Tesla V100-SXM2-32GB 25 25.0X 20 20.83 20.0X 19.94 Satellite Tobacco Mosaic Virus 17.02 1,067,095 atoms 15 15.0X ns/day Running AmberMD 18.10_AT_18.12 2X V100 4X V100 10 1X V100 21.0X 21.9X 10.0X The blue node contains Dual Intel Xeon 17
Slide 12: AmberMD 18.10_AT_18.12 – P100 vs V100
AmberMD 18.10_AT_18.12 – P100 vs V100 AmberMD 18.10-AT_18.12 70.00 25.0X 60.00 63.19 All benchmarks compared as set 57.68 20.0X Cellulose, FactorIX, JAC, STMV 50.00 Running AmberMD 18.10_AT_18.12 48.13 40.00 15.0X The blue node contains Dual Intel Xeon 39.99 Gold 6140 (Skylake) CPUs ns/day 37.17 4X
Slide 13: AmberMD 18.10_AT_18.12 - PME-Cellulose (T4)
AmberMD 18.10_AT_18.12- PME-Cellulose AmberMD 18.10-AT_18.12 - Tesla T4 30.0 9.0X 8.0X 25.0 24.9 7.0X 23.9 22.7 20.0 21.8 6.0X Cellulose 408,609 atoms 5.0X 17.1 Running AmberMD 18.10_AT_18.12 ns/day 15.0 16.0 2X T4 4X T4 4.0X 2X T4 4X T4 7.3X 7.0X 7.7X 7.4X The blue node contains Dual Intel Xeon 10.
Slide 14: AmberMD 18.10_AT_18.12 - PME-FactorIX (T4)
AmberMD 18.10_AT_18.12 - PME-FactorIX AmberMD 18.10-AT_18.12 - Tesla T4 140.0 9.0X 8.0X 120.0 123.8 112.5 112.6 7.0X Factor IX 100.0 102.3 90,906 atoms 6.0X 80.0 85.0 5.0X Running AmberMD 18.10_AT_18.12 ns/day 79.6 2X T4 60.0 2X T4 8.0X 4X T4 4.0X The blue node contains Dual Intel Xeon 7.5X 4X T4 7.
Slide 15: AmberMD 18.10_AT_18.12 - PME-JAC (T4)
AmberMD 18.10_AT_18.12 - PME-JAC AmberMD 18.10-AT_18.12 - Tesla T4 400.0 8.0X 372.8 350.0 7.0X 331.8 336.3 300.0 6.0X 301.8 DHFR 285.4 250.0 262.2 5.0X 23,558 atoms ns/day Running AmberMD 18.10_AT_18.12 200.0 4.0X 2X T4 2X T4 The blue node contains Dual Intel Xeon 150.0 6.7X 4X T4 3.0X Gold 6140 (Sk
Slide 16: AmberMD 18.10_AT_18.12 - PME-STMV_NPT (T4)
AmberMD 18.10_AT_18.12 - PME-STMV_NPT AmberMD 18.10-AT_18.12 - Tesla T4 16.0 9.0X 15.0 8.0X 14.0 14.3 7.0X 12.0 Satellite Tobacco Mosaic Virus 10.0 10.7 6.0X 1,067,095 atoms 5.0X ns/day 8.0 Running AmberMD 18.10_AT_18.12 2X T4 4X T4 4.0X 8.2X 6.0 7.8X The blue node contains Dual Intel Xeon 1X T4 3.0
Slide 17: AmberMD recommended usage
AmberMD recommended usage Motherboard and CPU Dual-socket with server x86-64 CPU System memory >=16GB GPUs Tesla V100 SXM2 GPUs per socket 1 to 8 GPUs per task 1 – 4 (case dependent) 17 17
Slide 18: GROMACS 2019.1
GROMACS 2019.1 March 2019 | Naas @SINVIDIA.
Slide 19: GROMACS 2019.1 - ADH Dodec
GROMACS 2019.1 - ADH Dodec GROMACS 2019.1 - Tesla V100-SXM2-32GB 250 4.0X 3.5X 200 193.52 3.0X 184.67 ADH 150 160.21 2.5X 134,000 atoms ns/day 4X V100 2.0X Running GROMACS 2019.1 2X V100 100 1X V100 3.4X 3.6X The blue node contains Dual Intel Xeon 1.5X 3.0X Gold 6140 (Skylake) CPUs 1.0X 50 The green
Slide 20: GROMACS 2019.1 - Cellulose
GROMACS 2019.1 - Cellulose GROMACS 2019.1 - Tesla V100-SXM2-32GB 60 4.0X 54.22 3.5X 50 51.94 3.0X 44.49 40 Cellulose 2.5X 408,609 atoms ns/day 30 4X V100 2.0X Running GROMACS 2019.1 2X V100 1X V100 3.4X 3.6X The blue node contains Dual Intel Xeon 1.5X 20 2.9X Gold 6140 (Skylake) CPUs 1.0X 15.13 The
Slide 21: GROMACS 2019.1 - STMV
GROMACS 2019.1 - STMV GROMACS 2019.1 - Tesla V100-SXM2-32GB 18 5.0X 16 4.5X 15.84 15.95 4.0X 14 3.5X Satellite Tobacco Mosaic Virus 12 1,067,095 atoms 3.0X 10 10.24 ns/day 2X V100 4X V100 2.5X Running GROMACS 2019.1 8 4.5X 4.5X 2.0X The blue node contains Dual Intel Xeon 6 1X V100 Gold 6140 (Skylake
Slide 22: GROMACS 2019.1 - ADH Dodec (T4)
GROMACS 2019.1 - ADH Dodec GROMACS 2019.1 - Tesla T4 200.0 3.5X 180.0 3.0X 176.5 160.0 152.8 2.5X ADH 140.0 120.0 128.5 2.0X 134,000 atoms ns/day 100.0 8X T4 Running GROMACS 2019.1 93.3 4X T4 3.3X 1.5X 80.0 2X T4 The blue node contains Dual Intel Xeon 2.8X Gold 6140 (Skylake) CPUs 60.0 2.4X 1X T4 1.
Slide 23: GROMACS 2019.1 - Cellulose (T4)
GROMACS 2019.1 - Cellulose GROMACS 2019.1 - Tesla T4 60.0 3.5X 3.0X 50.0 2.5X 40.0 42.3 Cellulose 2.0X 408,609 atoms ns/day 33.6 30.0 8X T4 Running GROMACS 2019.1 4X T4 3.3X 1.5X The blue node contains Dual Intel Xeon 20.0 23.3 2X T4 2.8X Gold 6140 (Skylake) CPUs 2.2X 1.0X 15.1 1X T4 The green nodes
Slide 24: GROMACS 2019.1 - STMV (T4)
GROMACS 2019.1 - STMV GROMACS 2019.1 - Tesla T4 16.0 5.0X 15.2 4.5X 14.0 4.0X 12.0 12.0 3.5X Satellite Tobacco Mosaic Virus 10.0 1,067,095 atoms 3.0X ns/day 9.3 8.0 2.5X Running GROMACS 2019.1 8X T4 6.0 4X T4 4.3X 2.0X The blue node contains Dual Intel Xeon Gold 6140 (Skylake) CPUs 2X T4 3.4X 1.5X 4
Slide 25: GROMACS recommended usage
GROMACS recommended usage Motherboard and CPU Dual-socket with server x86-64 CPU System memory >=16GB GPUs Tesla V100 SXM2 GPUs per socket 1 to 4 GPUs per task 1 - 2 (case dependent) 25 25
Slide 26: HOOMD-Blue 2.5.0
HOOMD-Blue 2.5.0 March 2019 | Naas @SINVIDIA.
Slide 27: HOOMD-Blue 2.5.0 - dodecahedron
HOOMD-Blue 2.5.0 - dodecahedron HOOMD-Blue 2.5.0 - Tesla V100-SXM2-16GB 300.00 8X V100 250.00 264.95 4X V100 229.04 200.00 Hard particle Monte Carlo 2X V100 131,072 atoms 188.79 150.00 Running HOOMD-Blue 2.5.0 Average TPS 1X V100 100.00 The blue node contains Dual Intel Xeon 131.04 E5-2698 v4 (Broad
Slide 28: HOOMD-Blue 2.5.0 - hexagon
HOOMD-Blue 2.5.0 - hexagon HOOMD-Blue 2.5.0 - Tesla V100-SXM2-16GB 140.00 120.00 8X V100 115.29 100.00 Hard particle Monte Carlo 80.00 1,048,576 atoms Running HOOMD-Blue 2.5.0 Average TPS 60.00 4X V100 64.54 The blue node contains Dual Intel Xeon 40.00 E5-2698 v4 (Broadwell) CPUs 2X V100 The green n
Slide 29: HOOMD-Blue 2.5.0 - lj-liquid
HOOMD-Blue 2.5.0 - lj-liquid HOOMD-Blue 2.5.0 - Tesla V100-SXM2-16GB 4500.00 4000.00 4X V100 2X V100 3975.29 3500.00 1X V100 3745.16 3490.57 8X V100 3000.00 Lennard-Jones pair force 3184.32 64000 atoms 2500.00 2000.00 Running HOOMD-Blue 2.5.0 Average TPS 1500.00 The blue node contains Dual Intel Xeo
Slide 30: HOOMD-Blue 2.5.0 - microsphere
HOOMD-Blue 2.5.0 - microsphere HOOMD-Blue 2.5.0 - Tesla V100-SXM2-16GB 700.00 8X V100 600.00 629.61 500.00 DPD pair force 1,428,364 atoms 4X V100 400.00 430.08 Running HOOMD-Blue 2.5.0 Average TPS 300.00 2X V100 The blue node contains Dual Intel Xeon 277.13 200.00 E5-2698 v4 (Broadwell) CPUs 1X V100
Slide 31: HOOMD-Blue 2.5.0 - quasicrystal
HOOMD-Blue 2.5.0 - quasicrystal HOOMD-Blue 2.5.0 - Tesla V100-SXM2-16GB 3000.00 4X V100 2500.00 2622.76 8X V100 2X V100 2370.03 2227.43 2000.00 Oscillatory pair potential 100000 atoms 1500.00 1X V100 Running HOOMD-Blue 2.5.0 1548.77 Average TPS 1000.00 The blue node contains Dual Intel Xeon E5-2698
Slide 32: HOOMD-Blue 2.5.0 - triblock-copolymer
HOOMD-Blue 2.5.0 - triblock-copolymer HOOMD-Blue 2.5.0 - Tesla V100-SXM2-16GB 3000.00 1X V100 4X V100 2X V100 2779.28 2770.79 2500.00 2712.92 8X V100 2298.27 2000.00 LJ pair force - forms spherical micelles 64017 atoms 1500.00 Running HOOMD-Blue 2.5.0 Average TPS 1000.00 The blue node contains Dual
Slide 33: HOOMD-Blue recommended usage
HOOMD-Blue recommended usage Motherboard and CPU Dual-socket with server x86-64 CPU System memory >=32GB GPUs Tesla V100 SXM2 GPUs per socket 1 to 4 GPUs per task 1, 4, or 8 based on benchmarks 33 33
Slide 34: LAMMPS 12Dec2018_stable
LAMMPS 12Dec2018_stable March 2019
Slide 35: LAMMPS - Atomic-Fluid Lennard-Jones 2.5 Cutoff
LAMMPS 12Dec2018_stable Atomic-Fluid Lennard-Jones 2.5 Cutoff LAMMPS - 12Dec2018_stable - Atomic-Fluid Lennard-Jones 2.5 Cutoff 1.20E+09 1.07E+09 1.00E+09 8.00E+08 AVG Atom-Timesteps/s 6.00E+08 Running LAMMPS 12Dec2018_stable The blue node contains Dual Intel Xeon 4.72E+08 4.00E+08 Gold 6140 (Skylak
Slide 36: LAMMPS 12Dec2018_stable - EAM
LAMMPS 12Dec2018_stable - EAM LAMMPS 12Dec2018_stable - EAM 3.50E+08 3.00E+08 2.88E+08 2.50E+08 Bulk Cu lattice AVG Atom-Timesteps/s 2.00E+08 Running LAMMPS 12Dec2018_stable 1.50E+08 1.54E+08 The blue node contains Dual Intel Xeon Gold 6140 (Skylake) CPUs 1.00E+08 The green nodes contain Dual Intel
Slide 37: LAMMPS 12Dec2018_stable - Tersoff
LAMMPS 12Dec2018_stable - Tersoff LAMMPS - 12Dec2018_stable - Tersoff 8.00E+08 7.00E+08 7.36E+08 6.00E+08 AVG Atom-Timesteps/s 5.00E+08 Si crystallization 4.00E+08 4.07E+08 Running LAMMPS 12Dec2018_stable 3.00E+08 The blue node contains Dual Intel Xeon Gold 6140 (Skylake) CPUs 2.00E+08 2.12E+08 The
Slide 38: LAMMPS recommended usage
LAMMPS recommended usage Motherboard and CPU Dual-socket with server x86-64 CPU System memory >=32GB GPUs Tesla V100 SXM2 GPUs per socket 1 to 4 GPUs per task 4 38 38
Slide 39: NAMD 2.13
NAMD 2.13 March 2019 @SINVIDIA.
Slide 40: NAMD 2.13 – APO1
NAMD 2.13 – APO1 NAMD 2.13 - Tesla V100-SXM2-32GB 90 20.0X 80 18.0X 75.37 16.0X 70 71.61 70.99 67.19 ApoA1 14.0X 60 63.73 61.29 92,224 atoms 60.18 57.7 12.0X 50 54.02 Average ns/day 10.0X Running NAMD 2.13 40 4X V100 4X V100 2X V100 17.3X 2X V100 17.2X 4X V100 2X V100 1X V100 16.2X 1X V100 16.2X 8.0
Slide 41: NAMD 2.13 – STMV
NAMD 2.13 – STMV NAMD 2.13 - Tesla V100-SXM2-32GB 9 25.0X 8 7.87 7.49 20.0X 7 7.04 6.42 6.88 Satellite Tobacco Mosaic Virus 6 6.19 5.98 1,067,095 atoms 5.71 15.0X 5 5.36 Average ns/day Running NAMD 2.13 4 4X V100 4X V100 2X V100 19.7X 2X V100 20.7X 10.0X The blue node contains Dual Intel Xeon 3 2X V
Slide 42: NAMD 2.13 – APO1 (T4)
NAMD 2.13 – APO1 NAMD 2.13 - Tesla T4 90 20.0X 80 18.0X 75.54 16.0X 70 70.67 ApoA1 66.23 14.0X 60 12.0X 92,224 atoms 50 54.43 Average ns/day 50.38 49.37 10.0X Running NAMD 2.13 40 4X T4 4X T4 4X T4 17.3X 16.0X 17.1X 8.0X The blue node contains Dual Intel Xeon 30 2X T4 2X T4 31.53 2X T4 Gold 6140 (Sk
Slide 43: NAMD 2.13 – STMV (T4)
NAMD 2.13 – STMV NAMD 2.13 - Tesla T4 9 25.0X 8 7.71 20.0X 7 7.11 6.9 Satellite Tobacco Mosaic Virus 6 1,067,095 atoms 15.0X 5 Average ns/day 4.73 Running NAMD 2.13 4 4.44 4.47 4X T4 4X T4 4X T4 20.3X 10.0X The blue node contains Dual Intel Xeon 3 18.2X 18.7X Gold 6140 (Skylake) CPUs 2X T4 2X T4 2X
Slide 44: NAMD recommended usage
NAMD recommended usage Motherboard and CPU Dual-socket with server x86-64 CPU System memory >=16GB Tesla V100 GPUs Tesla T4 GPUs per socket 1 to 4 GPUs per task 4 44 44
Slide 45: MD Applications GPU-Accelerated Computing
MD Applications GPU-Accelerated Computing Turbocharge your research! • Speedup of 3X-8X compared to CPU only in all tests (average) • Majority of compute intensive for classical MD ported to GPUs • Large performance boost and improve TCO for compute infrastructure • Tesla GPUs are more energy effici
Slide 46: Molecular Dynamics (MD) on GPUs
Molecular Dynamics (MD) on GPUs March 2019
Slide 47: GPU-Accelerated Quantum Chemistry Apps
GPU-Accelerated Quantum Chemistry Apps Green Lettering Indicates Performance Slides Included Abinit GPAW Octopus Quantum ACES III SuperCharger FHI-AIMS ONETEP Library ADF LATTE Petot RMG BigDFT LSDalton Q-Chem TeraChem CP2K MOLCAS QMCPACK UNM DIRAC Mopac2012 Quantum VASP GAMESS-US Espresso NWChem WL