NVIDIA HPC Application Performance
For Deep Learning performance, please go here.
Modern HPC data centers are key to solving some of the world’s most important scientific and engineering challenges. The NVIDIA A100, V100 and T4 GPUs fundamentally change the economics of the data center, delivering breakthrough performance with dramatically fewer servers, less power consumption, and reduced networking overhead, resulting in total cost savings of 5X-10X.
The number of CPU-only servers replaced by a single GPU-accelerated server is called the node replacement factor (NRF). To arrive at NRF, we measure application performance with up to 8 CPU-only servers. Then we use linear scaling to scale beyond 8 servers to calculate the NRF. The NRF will vary by application.
Engineering
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A100 SXM 80GB |FUN3D Benchmark: dpw_wbt0_crs-3.6Mn_5, CUDA Version: 11.4.2
Geoscience
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A100 SXM 80GB | RTM Benchmark: Isotropic Radius 4, CUDA Version: 11.4.2 | SPECFEM3D Benchmark: four_material_simple_model, CUDA Version: 11.4.2
Molecular Dynamics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A100 SXM 80GB | AMBER Benchmark: DC-STMV_NPT, CUDA Version: 11.4.2 | GROMACS Benchmark: Cellulose, CUDA Version: 11.4.2 | LAMMPS Benchmark: SNAP, CUDA Version: 11.4.2 | NAMD Benchmark: apoa1_nve_cuda, CUDA Version: 11.4.2 | Relion Benchmark: Plasmodium Ribosome (2D), CUDA Version: 11.4.2
Physics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A100 SXM 80GB | GTC Benchmark: moi#proc.in, CUDA Version: 11.3.1 | MILC Benchmark: Apex Medium, CUDA Version: 11.4.2 | Chroma Benchmark: szscl21_24_128, CUDA Version: 11.3.1
Quantum Mechanics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A100 SXM 80GB | Quantum Espresso Benchmark: AUSURF112-jR, CUDA Version: 11.4.2 | ICON Benchmark: SLAM 191 levels 160 km resolution with radiation, CUDA Version: 11.3.1
Detailed A100 application performance data is located below in alphabetical order.
AMBER
Molecular Dynamics
Suite of programs to simulate molecular dynamics on biomolecule
VERSION
20.12-AT_21.10
ACCELERATED FEATURES
- PMEMD Explicit Solvent and GB Implicit Solvent
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AMBER [PME-Cellulose_NPT_4fs] | ns/day | DC-Cellulose_NPT | yes | 4.40 | 151 | 302 | 603 | 1,207 | 147 | 294 | 589 | 1,178 |
| AMBER [PME-Cellulose_NPT_4fs] | NRF | DC-Cellulose_NPT | yes | 1x | 34x | 69x | 137x | 274x | 33x | 67x | 134x | 268x |
| AMBER [PME-Cellulose_NVE_4fs] | ns/day | DC-Cellulose_NVE | yes | 4.41 | 163 | 327 | 653 | 1,306 | 162 | 324 | 648 | 1,295 |
| AMBER [PME-Cellulose_NVE_4fs] | NRF | DC-Cellulose_NVE | yes | 1x | 37x | 74x | 148x | 296x | 37x | 73x | 147x | 294x |
| AMBER [PME-FactorIX_NPT_4fs] | ns/day | DC-FactorIX_NPT | yes | 22.80 | 536 | 1,072 | 2,144 | 4,288 | 532 | 1,064 | 2,128 | 4,255 |
| AMBER [PME-FactorIX_NPT_4fs] | NRF | DC-FactorIX_NPT | yes | 1x | 24x | 47x | 94x | 188x | 23x | 47x | 93x | 187x |
| AMBER [PME-FactorIX_NVE_4fs] | ns/day | DC-FactorIX_NVE | yes | 23.13 | 569 | 1,138 | 2,276 | 4,552 | 576 | 1,152 | 2,304 | 4,609 |
| AMBER [PME-FactorIX_NVE_4fs] | NRF | DC-FactorIX_NVE | yes | 1x | 25x | 49x | 98x | 197x | 25x | 50x | 100x | 199x |
| AMBER [PME-JAC_NPT_4fs] | ns/day | DC-JAC_NPT | yes | 97.51 | 1,200 | 2,399 | 4,799 | 9,598 | 1,209 | 2,417 | 4,835 | 9,670 |
| AMBER [PME-JAC_NPT_4fs] | NRF | DC-JAC_NPT | yes | 1x | 12x | 25x | 49x | 98x | 12x | 25x | 50x | 99x |
| AMBER [PME-JAC_NVE_4fs] | ns/day | DC-JAC_NVE | yes | 99.02 | 1,272 | 2,544 | 5,089 | 10,178 | 1,278 | 2,556 | 5,113 | 10,225 |
| AMBER [PME-JAC_NVE_4fs] | NRF | DC-JAC_NVE | yes | 1x | 13x | 26x | 51x | 103x | 13x | 26x | 52x | 103x |
| AMBER [PME-STMV_NPT_4fs] | ns/day | DC-STMV_NPT | yes | 1.42 | 55 | 111 | 221 | 442 | 53 | 106 | 213 | 426 |
| AMBER [PME-STMV_NPT_4fs] | NRF | DC-STMV_NPT | yes | 1x | 39x | 78x | 156x | 311x | 37x | 75x | 150x | 300x |
Benchmarked in ensemble mode
Chroma
Physics
Lattice Quantum Chromodynamics (LQCD)
VERSION
V 2021.05
ACCELERATED FEATURES
- Wilson-clover fermions, Krylov solvers, Domain-decomposition
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Chroma | Total Time (Sec) | szscl21_24_128 | no | 1,115 | 36 | 20 | 11 | 7 | 44 | 25 | 13 | 9 |
| Chroma | NRF | szscl21_24_128 | yes | 1x | 32x | 55x | 99x | 163x | 26x | 46x | 84x | 129x |
FUN3D
Engineering
Suite of tools actively developed at NASA for Aeronautics and Space Technology by modeling fluid flow.
VERSION
13.7 (update 1)
ACCELERATED FEATURES
- Full range of Mach number regimes for the Reynolds-averaged Navier Stokes (RANS) formulation
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| FUN3D | Loop Time (Sec) | dpw_wbt0_crs-3.6Mn_5 | no | 524 | 52 | 28 | 16 | 11 | 54 | 29 | 16 | 12 |
| FUN3D | NRF | dpw_wbt0_crs-3.6Mn_5 | yes | 1x | 13x | 24x | 41x | 57x | 12x | 23x | 40x | 53x |
GROMACS
Molecular Dynamics
Simulation of biochemical molecules with complicated bond interactions
VERSION
2021.3
ACCELERATED FEATURES
- Implicit (5x), Explicit (2x) Solvent
SCALABILITY
Multi-GPU, Single Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GROMACS [ADH Dodec] | ns/day | ADH Dodec | yes | 56 | 329 | - | 477 | - | 327 | - | 494 | - |
| GROMACS [ADH Dodec] | NRF | ADH Dodec | yes | 1x | 8x | - | 11x | - | 8x | - | 12x | - |
| GROMACS [Cellulose] | ns/day | Cellulose | yes | 16 | 96 | 139 | 226 | 248 | 96 | 130 | 169 | 187 |
| GROMACS [Cellulose] | NRF | Cellulose | yes | 1x | 9x | 13x | 21x | 23x | 9x | 12x | 16x | 18x |
| GROMACS [STMV] | ns/day | STMV | yes | 4 | 22 | 38 | 56 | 112 | 22 | 38 | 54 | 77 |
| GROMACS [STMV] | NRF | STMV | yes | 1x | 6x | 11x | 16x | 31x | 6x | 11x | 15x | 22x |
GTC
Physics
GTC is used for Gyrokinetic Particle Simulation of Turbulent Transport in Burning Plasmas.
VERSION
V 4.5 Updated
ACCELERATED FEATURES
- Push, shift, and collision
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GTC | Mpush/Sec | moi#proc.in | yes | 35 | 489 | 923 | 1,803 | 3,566 | 480 | 893 | 1,748 | 3,436 |
| GTC | NRF | moi#proc.in | yes | 1x | 14x | 27x | 53x | 104x | 14x | 26x | 51x | 100x |
ICON
Weather and Climate
A global unified atmosphere model for numerical weather prediction and climate modeling research.
VERSION
2.6.2+RRTMGP
ACCELERATED FEATURES
- Full model of dynamics and physics
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB |
|---|---|---|---|---|---|---|---|---|---|---|
| ICON [SLAM 191 - 160KM - no radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution without radiation | no | 80 | 13 | 9 | 7 | 6 | 14 | 11 |
| ICON [SLAM 191 - 160KM - no radiation] | NRF | SLAM 191 levels 160 km resolution without radiation | yes | 1x | 6x | 9x | 11x | 13x | 6x | 7x |
| ICON [SLAM 191 - 160KM - with radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution with radiation | no | 146 | 25 | 15 | 10 | 8 | 26 | 17 |
| ICON [SLAM 191 - 160KM - with radiation] | NRF | SLAM 191 levels 160 km resolution with radiation | yes | 1x | 6x | 10x | 15x | 18x | 6x | 9x |
LAMMPS
Molecular Dynamics
Classical molecular dynamics package
VERSION
stable_29Sep2021
ACCELERATED FEATURES
- Lennard-Jones, Gay-Berne, Tersoff, many more potentials
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LAMMPS [LJ 2.5] | ATOM-Time Steps/s | LJ 2.5 | yes | 1.07E+08 | 5.54E+08 | 1.04E+09 | 1.93E+09 | 3.61E+09 | 5.25E+08 | 9.61E+08 | 1.76E+09 | 2.76E+09 |
| LAMMPS [LJ 2.5] | NRF | LJ 2.5 | yes | 1x | 5x | 10x | 19x | 35x | 5x | 9x | 17x | 27x |
| LAMMPS [EAM] | ATOM-Time Steps/s | EAM | yes | 5.09E+07 | 2.82E+08 | 5.10E+08 | 8.86E+08 | 1.50E+09 | 2.70E+08 | 4.81E+08 | 8.13E+08 | 1.26E+09 |
| LAMMPS [EAM] | NRF | EAM | yes | 1x | 6x | 10x | 17x | 29x | 5x | 9x | 16x | 25x |
| LAMMPS [ReaxFF/C] | ATOM-Time Steps/s | ReaxFF/C | yes | 3.41E+05 | 3.06E+06 | 5.32E+06 | 8.73E+06 | 1.23E+07 | 3.04E+06 | 5.30E+06 | 8.55E+06 | 1.14E+07 |
| LAMMPS [ReaxFF/C] | NRF | ReaxFF/C | yes | 1x | 15x | 26x | 43x | 61x | 15x | 26x | 43x | 57x |
| LAMMPS [Rhodopsin] | ATOM-Time Steps/s | Rhodopsin | yes | 7.34E+06 | - | 1.92E+07 | 3.13E+07 | 4.71E+07 | - | 1.86E+07 | 2.80E+07 | 3.20E+07 |
| LAMMPS [Rhodopsin] | NRF | Rhodopsin | yes | 1x | - | 3x | 5x | 7x | - | 2x | 4x | 5x |
| LAMMPS [SNAP] | ATOM-Time Steps/s | SNAP | yes | 1.14E+05 | 2.09E+06 | 4.16E+06 | 8.21E+06 | 1.58E+07 | 2.05E+06 | 4.09E+06 | 8.11E+06 | 1.56E+07 |
| LAMMPS [SNAP] | NRF | SNAP | yes | 1x | 18x | 36x | 72x | 138x | 18x | 36x | 71x | 137x |
| LAMMPS [Tersoff] | ATOM-Time Steps/s | Tersoff | yes | 2.78E+07 | 4.61E+08 | 8.27E+08 | 1.46E+09 | 2.24E+09 | 4.44E+08 | 7.82E+08 | 1.33E+09 | 1.84E+09 |
| LAMMPS [Tersoff] | NRF | Tersoff | yes | 1x | 17x | 30x | 52x | 80x | 16x | 28x | 48x | 66x |
MILC
Physics
Lattice Quantum Chromodynamics (LQCD) codes simulate how elemental particles are formed and bound by the “strong force” to create larger particles like protons and neutrons
VERSION
develop_d98de0c4
ACCELERATED FEATURES
- Staggered fermions, Krylov solvers, Gauge-link fattening
SCALABILITY
Multi-GPU and Multi-Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MILC | Total Time (Sec) | Apex Medium | no | 67,366 | 2,260 | 1,319 | 700 | 394 | 2,374 | 1,373 | 660 | 624 |
| MILC | NRF | Apex Medium | yes | 1x | 33x | 56x | 106x | 188x | 31x | 54x | 112x | 119x |
NAMD
Molecular Dynamics
Designed for high-performance simulation of large molecular systems
VERSION
GPU, AMD CPU V 3.0a9 ; Intel CPU V 2.15a AVX512
ACCELERATED FEATURES
- Full electrostatics with PME and most simulation features
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NAMD [apoa1_npt_cuda] | Ave ns/day | apoa1_npt_cuda | yes | 19.12 | 121 | 243 | 490 | 974 | 121 | 227 | 479 | 961 |
| NAMD [apoa1_npt_cuda] | NRF | apoa1_npt_cuda | yes | 1x | 6x | 13x | 26x | 51x | 6x | 12x | 25x | 50x |
| NAMD [apoa1_nptsr_cuda] | Ave ns/day | apoa1_nptsr_cuda | yes | 19.63 | 126 | 250 | 498 | 995 | 121 | 244 | 483 | 974 |
| NAMD [apoa1_nptsr_cuda] | NRF | apoa1_nptsr_cuda | yes | 1x | 6x | 13x | 25x | 51x | 6x | 12x | 25x | 50x |
| NAMD [apoa1_nve_cuda] | Ave ns/day | apoa1_nve_cuda | yes | 20.64 | 164 | 326 | 649 | 1,310 | 160 | 324 | 647 | 1,275 |
| NAMD [apoa1_nve_cuda] | NRF | apoa1_nve_cuda | yes | 1x | 8x | 16x | 31x | 63x | 8x | 16x | 31x | 62x |
| NAMD [stmv_npt_cuda] | Ave ns/day | stmv_npt_cuda | yes | 1.78 | 12 | 24 | 47 | 95 | 12 | 23 | 46 | 92 |
| NAMD [stmv_npt_cuda] | NRF | stmv_npt_cuda | yes | 1x | 7x | 13x | 27x | 54x | 6x | 13x | 26x | 52x |
| NAMD [stmv_nptsr_cuda] | Ave ns/day | stmv_nptsr_cuda | yes | 1.80 | 12 | 24 | 49 | 97 | 12 | 24 | 47 | 94 |
| NAMD [stmv_nptsr_cuda] | NRF | stmv_nptsr_cuda | yes | 1x | 7x | 13x | 27x | 54x | 7x | 13x | 26x | 52x |
| NAMD [stmv_nve_cuda] | Ave ns/day | stmv_nve_cuda | yes | 1.95 | 14 | 28 | 55 | 110 | 14 | 27 | 54 | 107 |
| NAMD [stmv_nve_cuda] | NRF | stmv_nve_cuda | yes | 1x | 7x | 14x | 28x | 56x | 7x | 14x | 27x | 55x |
Benchmarked in ensemble mode
Quantum Espresso
Material Science (Quantum Chemistry)
An Open-source suite of computer codes for electronic structure calculations and materials modeling at the nanoscale
VERSION
CPU 6.7; GPU 6.8
ACCELERATED FEATURES
- linear algebra (matix multiply)
- explicit computational kernels
- 3D FFTs
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Quantum Espressso | Total CPU Time (Sec) | AUSURF112-jR | no | 765 | 135 | 84 | 58 | 49 | 147 | 92 | 78 |
| Quantum Espressso | NRF | AUSURF112-jR | yes | 1x | 6x | 10x | 15x | 17x | 6x | 9x | 11x |
RELION
Microscopy
Stand-alone computer program that employs an empirical Bayesianapproach to refinement of (multiple) 3D reconstructions or 2D class averages in electron cryo-microscopy (cryo-EM)
VERSION
3.1.3
ACCELERATED FEATURES
- Reduced memory requirements; high-resolution cryo-EM structure determination in a matter of day on a single workstation
SCALABILITY
Multi-GPU and Single Node
MORE INFORMATION
https://www2.mrc-lmb.cam.ac.uk/relion/index.php/Download_%26_install
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Relion [Plasmodium Ribosome] | Total Wall Clock (Sec) | MB numbers Plasmodium Ribosime on Relion-3.0 | no | 14,862 | 3,110 | 1,781 | 1,458 | 1,313 | 3,401 | 1,994 | 1,838 | 1,735 |
| Relion [Plasmodium Ribosome] | NRF | MB numbers Plasmodium Ribosime on Relion-3.0 | yes | 1x | 5x | 8x | 10x | 11x | 4x | 7x | 8x | 9x |
| Relion [Plasmodium Ribosome 2D] | Total Wall Clock (Sec) | Plasmodium Ribosome (2D) | no | 92,404 | 22,960 | 12,416 | 8,587 | 6,299 | 25,275 | 13,414 | 9,583 | 7,266 |
| Relion [Plasmodium Ribosome 2D] | NRF | Plasmodium Ribosome (2D) | yes | 1x | 4x | 9x | 14x | 19x | 4x | 9x | 12x | 16x |
RTM
Geoscience
Reverse time migration (RTM) modeling is a critical component in the seismic processing workflow of oil and gas exploration
VERSION
nvidia_2021_05
ACCELERATED FEATURES
- Batch algorithm
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RTM [Isotropic Radius 4] | Mcells/s | Isotropic Radius 4 | yes | 11,318 | 89,560 | 178,509 | 357,150 | 714,213 | 89,576 | 178,629 | 357,258 | 714,486 |
| RTM [Isotropic Radius 4] | NRF | Isotropic Radius 4 | yes | 1x | 8x | 16x | 32x | 63x | 8x | 16x | 32x | 63x |
| RTM [TTI Radius 8 1-pass] | Mcells/s | TTI Radius 8 1-pass | yes | 3,773 | 12,853 | 25,433 | 50,556 | 101,167 | 12,930 | 25,795 | 51,541 | 102,613 |
| RTM [TTI Radius 8 1-pass] | NRF | TTI Radius 8 1-pass | yes | 1x | 3x | 7x | 13x | 27x | 3x | 7x | 14x | 27x |
| RTM [TTI RX 2Pass mgpu] | Mcells/s | TTI RX 2Pass mgpu | yes | 3,773 | 13,940 | 27,580 | 54,758 | 108,322 | 13,716 | 27,221 | 54,185 | 107,829 |
| RTM [TTI RX 2Pass mgpu] | NRF | TTI RX 2Pass mgpu | yes | 1x | 4x | 7x | 15x | 29x | 4x | 7x | 14x | 29x |
SPECFEM3D
Geoscience
Simulates Seismic wave propagation
VERSION
devel_0a5acff9
ACCELERATED FEATURES
- OpenCL and CUDA hardware accelerators, based on an automatic source-to-source transformation library
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A100 SXM 80GB | 2x A100 SXM 80GB | 4x A100 SXM 80GB | 8x A100 SXM 80GB | 1x A100 PCIe 80GB | 2x A100 PCIe 800GB | 4x A100 PCIe 80GB | 8x A100 PCIe 80GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SPECFEM3D | Total Time (Sec) | four_material_simple_model | no | 1,991 | 77 | 40 | 22 | 14 | 78 | 41 | 22 | 15 |
| SPECFEM3D | NRF | four_material_simple_model | yes | 1x | 30x | 56x | 104x | 159x | 29x | 56x | 103x | 151x |
Engineering
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A30 | FUN3D Benchmark: dpw_wbt0_crs-3.6Mn_5, CUDA Version: 11.4.2
Geoscience
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A30 | RTM Benchmark: Isotropic Radius 4, CUDA Version: 11.4.2 | SPECFEM3D Benchmark: four_material_simple_model, CUDA Version: 11.4.2
Molecular Dynamics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A30 | AMBER Benchmark: DC-STMV_NPT, CUDA Version: 11.4.2 | GROMACS Benchmark: Cellulose, CUDA Version: 11.4.2 | LAMMPS Benchmark: SNAP, CUDA Version: 11.4.2 | NAMD Benchmark: apoa1_nve_cuda, CUDA Version: 11.4.2 | Relion Benchmark: Plasmodium Ribosome (2D), CUDA Version: 11.4.2
Physics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A30 | Chroma Benchmark: szscl21_24_128, CUDA Version: 11.3.1 | GTC Benchmark: moi#proc.in, CUDA Version: 11.3.1 | MILC Benchmark: Apex Medium, CUDA Version: 11.4.2
Quantum Mechanics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A30 | ICON Benchmark: SLAM 191 levels 160 km resolution with radiation, CUDA Version: 11.3.1 | Quantum Espresso Benchmark: AUSURF112-jR, CUDA Version: 11.4.2
Detailed A30 application performance data is located below in alphabetical order.
AMBER
Molecular Dynamics
Suite of programs to simulate molecular dynamics on biomolecule
VERSION
20.12-AT_21.10
ACCELERATED FEATURES
- PMEMD Explicit Solvent and GB Implicit Solvent
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AMBER [PME-Cellulose_NPT_4fs] | ns/day | DC-Cellulose_NPT | yes | 4.40 | 81 | 162 | 324 | 648 | ||||||||
| AMBER [PME-Cellulose_NPT_4fs] | NRF | DC-Cellulose_NPT | yes | 1x | 18x | 37x | 74x | 147x | ||||||||
| AMBER [PME-Cellulose_NVE_4fs] | ns/day | DC-Cellulose_NVE | yes | 4.41 | 86 | 171 | 342 | 685 | ||||||||
| AMBER [PME-Cellulose_NVE_4fs] | NRF | DC-Cellulose_NVE | yes | 1x | 19x | 39x | 78x | 155x | ||||||||
| AMBER [PME-FactorIX_NPT_4fs] | ns/day | DC-FactorIX_NPT | yes | 22.80 | 333 | 666 | 1,333 | 2,665 | ||||||||
| AMBER [PME-FactorIX_NPT_4fs] | NRF | DC-FactorIX_NPT | yes | 1x | 15x | 29x | 58x | 117x | ||||||||
| AMBER [PME-FactorIX_NVE_4fs] | ns/day | DC-FactorIX_NVE | yes | 23.13 | 353 | 707 | 1,414 | 2,827 | ||||||||
| AMBER [PME-FactorIX_NVE_4fs] | NRF | DC-FactorIX_NVE | yes | 1x | 15x | 31x | 61x | 122x | ||||||||
| AMBER [PME-JAC_NPT_4fs] | ns/day | DC-JAC_NPT | yes | 97.51 | 902 | 1,804 | 3,608 | 7,216 | ||||||||
| AMBER [PME-JAC_NPT_4fs] | NRF | DC-JAC_NPT | yes | 1x | 9x | 19x | 37x | 74x | ||||||||
| AMBER [PME-JAC_NVE_4fs] | ns/day | DC-JAC_NVE | yes | 99.02 | 946 | 1,892 | 3,783 | 7,566 | ||||||||
| AMBER [PME-JAC_NVE_4fs] | NRF | DC-JAC_NVE | yes | 1x | 10x | 19x | 38x | 76x | ||||||||
| AMBER [PME-STMV_NPT_4fs] | ns/day | DC-STMV_NPT | yes | 1.42 | 29 | 58 | 117 | 234 | ||||||||
| AMBER [PME-STMV_NPT_4fs] | NRF | DC-STMV_NPT | yes | 1x | 21x | 41x | 82x | 165x |
Benchmarked in ensemble mode
Chroma
Physics
Lattice Quantum Chromodynamics (LQCD)
VERSION
V 2021.05
ACCELERATED FEATURES
- Wilson-clover fermions, Krylov solvers, Domain-decomposition
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Chroma | Total Time (Sec) | szscl21_24_128 | no | 1,115 | 35 | 18 | 11 | ||||||||
| Chroma | NRF | szscl21_24_128 | yes | 1x | 33x | 62x | 103x |
FUN3D
Engineering
Suite of tools actively developed at NASA for Aeronautics and Space Technology by modeling fluid flow.
VERSION
13.7 (update 1)
ACCELERATED FEATURES
- Full range of Mach number regimes for the Reynolds-averaged Navier Stokes (RANS) formulation
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| FUN3D | Loop Time (Sec) | dpw_wbt0_crs-3.6Mn_5 | no | 524 | 111 | 56 | 30 | 18 | ||||||||
| FUN3D | NRF | dpw_wbt0_crs-3.6Mn_5 | yes | 1x | 5x | 12x | 22x | 36x |
GROMACS
Molecular Dynamics
Simulation of biochemical molecules with complicated bond interactions
VERSION
2021.3
ACCELERATED FEATURES
- Implicit (5x), Explicit (2x) Solvent
SCALABILITY
Multi-GPU, Single Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GROMACS [ADH Dodec] | ns/day | ADH Dodec | yes | 56 | 191 | - | 354 | - | ||||||||
| GROMACS [ADH Dodec] | NRF | ADH Dodec | yes | 1x | 5x | - | 8x | - | ||||||||
| GROMACS [Cellulose] | ns/day | Cellulose | yes | 16 | 51 | 75 | 112 | 127 | ||||||||
| GROMACS [Cellulose] | NRF | Cellulose | yes | 1x | 3x | 5x | 10x | 12x | ||||||||
| GROMACS [STMV] | ns/day | STMV | yes | 4 | 12 | 21 | 33 | 51 | ||||||||
| GROMACS [STMV] | NRF | STMV | yes | 1x | 3x | 5x | 9x | 14x |
GTC
Physics
GTC is used for Gyrokinetic Particle Simulation of Turbulent Transport in Burning Plasmas.
VERSION
V 4.5 Updated
ACCELERATED FEATURES
- Push, shift, and collision
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GTC | Mpush/Sec | moi#proc.in | yes | 35 | 276 | 523 | 1,030 | 2,040 | ||||||||
| GTC | NRF | moi#proc.in | yes | 1x | 8x | 15x | 30x | 59x |
ICON
Weather and Climate
A global unified atmosphere model for numerical weather prediction and climate modeling research.
VERSION
2.6.2+RRTMGP
ACCELERATED FEATURES
- Full model of dynamics and physics
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ICON [SLAM 191 - 160KM - no radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution without radiation | no | 80 | 24 | 16 | - | - | ||||||||
| ICON [SLAM 191 - 160KM - no radiation] | NRF | SLAM 191 levels 160 km resolution without radiation | yes | 1x | 3x | 5x | - | - | ||||||||
| ICON [SLAM 191 - 160KM - with radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution with radiation | no | 146 | 47 | 28 | 23 | 19 | ||||||||
| ICON [SLAM 191 - 160KM - with radiation] | NRF | SLAM 191 levels 160 km resolution with radiation | yes | 1x | 3x | 5x | 6x | 8x |
LAMMPS
Molecular Dynamics
Classical molecular dynamics package
VERSION
stable_29Sep2021
ACCELERATED FEATURES
- Lennard-Jones, Gay-Berne, Tersoff, many more potentials
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LAMMPS [LJ 2.5] | ATOM-Time Steps/s | LJ 2.5 | yes | 1.07E+08 | 2.65E+08 | 5.03E+08 | 9.42E+08 | 1.72E+09 | ||||||||
| LAMMPS [LJ 2.5] | NRF | LJ 2.5 | yes | 1x | 3x | 5x | 9x | 17x | ||||||||
| LAMMPS [EAM] | ATOM-Time Steps/s | EAM | yes | 5.09E+07 | 1.29E+08 | 2.43E+08 | 4.44E+08 | 7.59E+08 | ||||||||
| LAMMPS [EAM] | NRF | EAM | yes | 1x | 3x | 5x | 9x | 15x | ||||||||
| LAMMPS [ReaxFF/C] | ATOM-Time Steps/s | ReaxFF/C | yes | 3.41E+05 | 1.68E+06 | 3.11E+06 | 5.45E+06 | 8.11E+06 | ||||||||
| LAMMPS [ReaxFF/C] | NRF | ReaxFF/C | yes | 1x | 8x | 15x | 27x | 40x | ||||||||
| LAMMPS [Rhodopsin] | ATOM-Time Steps/s | Rhodopsin | 7.34E+06 | - | 1.16E+07 | 1.96E+07 | - | |||||||||
| LAMMPS [Rhodopsin] | NRF | Rhodopsin | 1x | - | 2x | 3x | - | |||||||||
| LAMMPS [SNAP] | ATOM-Time Steps/s | SNAP | yes | 1.14E+05 | 1.07E+06 | 2.13E+06 | 4.24E+06 | 8.19E+06 | ||||||||
| LAMMPS [SNAP] | NRF | SNAP | yes | 1x | 9x | 19x | 37x | 72x | ||||||||
| LAMMPS [Tersoff] | ATOM-Time Steps/s | Tersoff | yes | 2.78E+07 | 2.24E+08 | 4.19E+08 | 7.58E+08 | 1.10E+09 | ||||||||
| LAMMPS [Tersoff] | NRF | Tersoff | yes | 1x | 8x | 15x | 27x | 39x |
MILC
Physics
Lattice Quantum Chromodynamics (LQCD) codes simulate how elemental particles are formed and bound by the “strong force” to create larger particles like protons and neutrons
VERSION
develop_d98de0c4
ACCELERATED FEATURES
- Staggered fermions, Krylov solvers, Gauge-link fattening
SCALABILITY
Multi-GPU and Multi-Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MILC | Total Time (Sec) | Apex Medium | no | 67,366 | 5,451 | 2,313 | 1,167 | 691 | ||||||||
| MILC | NRF | Apex Medium | yes | 1x | 14x | 32x | 63x | 107x |
NAMD
Molecular Dynamics
Designed for high-performance simulation of large molecular systems
VERSION
GPU, AMD CPU V 3.0a9 ; Intel CPU V 2.15a AVX512
ACCELERATED FEATURES
- Full electrostatics with PME and most simulation features
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NAMD [apoa1_npt_cuda] | Ave ns/day | apoa1_npt_cuda | yes | 19.12 | 73 | 146 | 291 | 582 | ||||||||
| NAMD [apoa1_npt_cuda] | NRF | apoa1_npt_cuda | yes | 1x | 4x | 8x | 15x | 30x | ||||||||
| NAMD [apoa1_nptsr_cuda] | Ave ns/day | apoa1_nptsr_cuda | yes | 19.63 | 76 | 151 | 302 | 601 | ||||||||
| NAMD [apoa1_nptsr_cuda] | NRF | apoa1_nptsr_cuda | yes | 1x | 4x | 8x | 15x | 31x | ||||||||
| NAMD [apoa1_nve_cuda] | Ave ns/day | apoa1_nve_cuda | yes | 20.64 | 92 | 184 | 369 | 737 | ||||||||
| NAMD [apoa1_nve_cuda] | NRF | apoa1_nve_cuda | yes | 1x | 4x | 9x | 18x | 36x | ||||||||
| NAMD [stmv_npt_cuda] | Ave ns/day | stmv_npt_cuda | yes | 1.78 | 6 | 13 | 25 | 50 | ||||||||
| NAMD [stmv_npt_cuda] | NRF | stmv_npt_cuda | yes | 1x | 4x | 7x | 14x | 28x | ||||||||
| NAMD [stmv_nptsr_cuda] | Ave ns/day | stmv_nptsr_cuda | yes | 1.80 | 6 | 13 | 26 | 51 | ||||||||
| NAMD [stmv_nptsr_cuda] | NRF | stmv_nptsr_cuda | yes | 1x | 4x | 7x | 14x | 28x | ||||||||
| NAMD [stmv_nve_cuda] | Ave ns/day | stmv_nve_cuda | yes | 1.95 | 7 | 14 | 28 | 56 | ||||||||
| NAMD [stmv_nve_cuda] | NRF | stmv_nve_cuda | yes | 1x | 4x | 7x | 15x | 29x |
Benchmarked in ensemble mode
Quantum Espresso
Material Science (Quantum Chemistry)
An Open-source suite of computer codes for electronic structure calculations and materials modeling at the nanoscale
VERSION
CPU 6.7; GPU 6.8
ACCELERATED FEATURES
- linear algebra (matix multiply)
- explicit computational kernels
- 3D FFTs
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Quantum Espressso | Total CPU Time (Sec) | AUSURF112-jR | no | 765 | 287 | 123 | 96 | ||||||||
| Quantum Espressso | NRF | AUSURF112-jR | yes | 1x | 3x | 7x | 9x |
RELION
Microscopy
Stand-alone computer program that employs an empirical Bayesianapproach to refinement of (multiple) 3D reconstructions or 2D class averages in electron cryo-microscopy (cryo-EM)
VERSION
3.1.3
ACCELERATED FEATURES
- Reduced memory requirements; high-resolution cryo-EM structure determination in a matter of day on a single workstation
SCALABILITY
Multi-GPU and Single Node
MORE INFORMATION
https://www2.mrc-lmb.cam.ac.uk/relion/index.php/Download_%26_install
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Relion [Plasmodium Ribosome] | Total Wall Clock (Sec) | MB numbers Plasmodium Ribosime on Relion-3.0 | no | 14,862 | 3,922 | 2,220 | 1,905 | 1,676 | ||||||||
| Relion [Plasmodium Ribosome] | NRF | MB numbers Plasmodium Ribosime on Relion-3.0 | yes | 1x | 4x | 7x | 8x | 9x | ||||||||
| Relion [Plasmodium Ribosome 2D] | Total Wall Clock (Sec) | Plasmodium Ribosome (2D) | no | 92,404 | 35,026 | 18,221 | 11,868 | 8,110 | ||||||||
| Relion [Plasmodium Ribosome 2D] | NRF | Plasmodium Ribosome (2D) | yes | 1x | 3x | 5x | 10x | 14x |
RTM
Geoscience
Reverse time migration (RTM) modeling is a critical component in the seismic processing workflow of oil and gas exploration
VERSION
nvidia_2021_05
ACCELERATED FEATURES
- Batch algorithm
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RTM [Isotropic Radius 4] | Mcells/s | Isotropic Radius 4 | yes | 11,318 | 44,068 | 87,787 | 175,548 | 350,754 | ||||||||
| RTM [Isotropic Radius 4] | NRF | Isotropic Radius 4 | yes | 1x | 4x | 8x | 16x | 31x | ||||||||
| RTM [TTI Radius 8 1-pass] | Mcells/s | TTI Radius 8 1-pass | yes | 3,773 | 6,688 | 13,223 | 26,359 | 52,547 | ||||||||
| RTM [TTI Radius 8 1-pass] | NRF | TTI Radius 8 1-pass | yes | 1x | 2x | 4x | 7x | 14x | ||||||||
| RTM [TTI RX 2Pass mgpu] | Mcells/s | TTI RX 2Pass mgpu | yes | 3,773 | 6,977 | 13,876 | 27,520 | 54,789 | ||||||||
| RTM [TTI RX 2Pass mgpu] | NRF | TTI RX 2Pass mgpu | yes | 1x | 2x | 4x | 7x | 15x |
SPECFEM3D
Geoscience
Simulates Seismic wave propagation
VERSION
devel_0a5acff9
ACCELERATED FEATURES
- OpenCL and CUDA hardware accelerators, based on an automatic source-to-source transformation library
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A30 | 2x A30 | 4x A30 | 8x A30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SPECFEM3D | Total Time (Sec) | four_material_simple_model | no | 1,991 | 156 | 80 | 42 | 23 | ||||||||
| SPECFEM3D | NRF | four_material_simple_model | yes | 1x | 15x | 28x | 54x | 98x |
Engineering
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A40 | FUN3D Benchmark: dpw_wbt0_crs-3.6Mn_5, CUDA Version: 11.4.2
Geoscience
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A40 | SPECFEM3D Benchmark: four_material_simple_model, CUDA Version: 11.4.2
Molecular Dynamics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A40 | AMBER Benchmark: DC-STMV_NPT, CUDA Version: 11.4.2 | GROMACS Benchmark: Cellulose, CUDA Version: 11.4.2 | LAMMPS Benchmark: SNAP, CUDA Version: 11.4.2 | NAMD Benchmark: apoa1_nve_cuda, CUDA Version: 11.4.2 | Relion Benchmark: Plasmodium Ribosome (2D), CUDA Version: 11.4.2
Physics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A40 | Chroma Benchmark: szscl21_24_128, CUDA Version: 11.3.1 | GTC Benchmark: moi#proc.in, CUDA Version: 11.3.1 | MILC Benchmark: Apex Medium, CUDA Version: 11.4.2
Quantum Mechanics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual EPYC 7742@2.25GHz with 4x NVIDIA A40 | ICON Benchmark: SLAM 191 levels 160 km resolution with radiation, CUDA Version: 11.3.1
Detailed A40 application performance data is located below in alphabetical order.
AMBER
Molecular Dynamics
Suite of programs to simulate molecular dynamics on biomolecule
VERSION
20.12-AT_21.10
ACCELERATED FEATURES
- PMEMD Explicit Solvent and GB Implicit Solvent
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AMBER [PME-Cellulose_NPT_4fs] | ns/day | DC-Cellulose_NPT | yes | 4.40 | 90 | 181 | 361 | 722 | ||||||||
| AMBER [PME-Cellulose_NPT_4fs] | NRF | DC-Cellulose_NPT | yes | 1x | 21x | 41x | 82x | 164x | ||||||||
| AMBER [PME-Cellulose_NVE_4fs] | ns/day | DC-Cellulose_NVE | yes | 4.41 | 96 | 192 | 384 | 768 | ||||||||
| AMBER [PME-Cellulose_NVE_4fs] | NRF | DC-Cellulose_NVE | yes | 1x | 22x | 44x | 87x | 174x | ||||||||
| AMBER [PME-FactorIX_NPT_4fs] | ns/day | DC-FactorIX_NPT | yes | 22.80 | 413 | 826 | 1,652 | 3,305 | ||||||||
| AMBER [PME-FactorIX_NPT_4fs] | NRF | DC-FactorIX_NPT | yes | 1x | 18x | 36x | 72x | 145x | ||||||||
| AMBER [PME-FactorIX_NVE_4fs] | ns/day | DC-FactorIX_NVE | yes | 23.13 | 438 | 875 | 1,750 | 3,500 | ||||||||
| AMBER [PME-FactorIX_NVE_4fs] | NRF | DC-FactorIX_NVE | yes | 1x | 19x | 38x | 76x | 151x | ||||||||
| AMBER [PME-JAC_NPT_4fs] | ns/day | DC-JAC_NPT | yes | 97.51 | 1,007 | 2,015 | 4,030 | 8,059 | ||||||||
| AMBER [PME-JAC_NPT_4fs] | NRF | DC-JAC_NPT | yes | 1x | 10x | 21x | 41x | 83x | ||||||||
| AMBER [PME-JAC_NVE_4fs] | ns/day | DC-JAC_NVE | yes | 99.02 | 1,059 | 2,119 | 4,237 | 8,474 | ||||||||
| AMBER [PME-JAC_NVE_4fs] | NRF | DC-JAC_NVE | yes | 1x | 11x | 21x | 43x | 86x | ||||||||
| AMBER [PME-STMV_NPT_4fs] | ns/day | DC-STMV_NPT | yes | 1.42 | 32 | 63 | 126 | 252 | ||||||||
| AMBER [PME-STMV_NPT_4fs] | NRF | DC-STMV_NPT | yes | 1x | 22x | 44x | 89x | 178x |
Benchmarked in ensemble mode
Chroma
Physics
Lattice Quantum Chromodynamics (LQCD)
VERSION
V 2021.05
ACCELERATED FEATURES
- Wilson-clover fermions, Krylov solvers, Domain-decomposition
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Chroma | Total Time (Sec) | szscl21_24_128 | no | 1,115 | 78 | 41 | 22 | 13 | ||||||||
| Chroma | NRF | szscl21_24_128 | yes | 1x | 15x | 28x | 52x | 89x |
FUN3D
Engineering
Suite of tools actively developed at NASA for Aeronautics and Space Technology by modeling fluid flow.
VERSION
13.7 (update 1)
ACCELERATED FEATURES
- Full range of Mach number regimes for the Reynolds-averaged Navier Stokes (RANS) formulation
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| FUN3D | Loop Time (Sec) | dpw_wbt0_crs-3.6Mn_5 | no | 524 | 226 | 115 | 59 | 32 | ||||||||
| FUN3D | NRF | dpw_wbt0_crs-3.6Mn_5 | yes | 1x | 2x | 5x | 11x | 20x |
GROMACS
Molecular Dynamics
Simulation of biochemical molecules with complicated bond interactions
VERSION
2021.3
ACCELERATED FEATURES
- Implicit (5x), Explicit (2x) Solvent
SCALABILITY
Multi-GPU, Single Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GROMACS [ADH Dodec] | ns/day | ADH Dodec | yes | 56 | 303 | 317 | 485 | - | ||||||||
| GROMACS [ADH Dodec] | NRF | ADH Dodec | yes | 1x | 7x | 8x | 12x | - | ||||||||
| GROMACS [Cellulose] | ns/day | Cellulose | yes | 16 | 71 | 91 | 143 | - | ||||||||
| GROMACS [Cellulose] | NRF | Cellulose | yes | 1x | 5x | 9x | 13x | - | ||||||||
| GROMACS [STMV] | ns/day | STMV | yes | 4 | 17 | 35 | 51 | 62 | ||||||||
| GROMACS [STMV] | NRF | STMV | yes | 1x | 5x | 10x | 14x | 17x |
GTC
Physics
GTC is used for Gyrokinetic Particle Simulation of Turbulent Transport in Burning Plasmas.
VERSION
V 4.5 Updated
ACCELERATED FEATURES
- Push, shift, and collision
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GTC | Mpush/Sec | moi#proc.in | yes | 35 | 279 | 523 | 1,033 | 2,032 | ||||||||
| GTC | NRF | moi#proc.in | yes | 1x | 8x | 15x | 30x | 59x |
ICON
Weather and Climate
A global unified atmosphere model for numerical weather prediction and climate modeling research.
VERSION
2.6.2+RRTMGP
ACCELERATED FEATURES
- Full model of dynamics and physics
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ICON [SLAM 191 - 160KM - no radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution without radiation | no | 80 | 29 | 18 | - | 16 | ||||||||
| ICON [SLAM 191 - 160KM - no radiation] | NRF | SLAM 191 levels 160 km resolution without radiation | yes | 1x | 3x | 4x | - | 5x | ||||||||
| ICON [SLAM 191 - 160KM - with radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution with radiation | no | 146 | 66 | 36 | 27 | 21 | ||||||||
| ICON [SLAM 191 - 160KM - with radiation] | NRF | SLAM 191 levels 160 km resolution with radiation | yes | 1x | 2x | 4x | 5x | 7x |
LAMMPS
Molecular Dynamics
Classical molecular dynamics package
VERSION
stable_29Sep2021
ACCELERATED FEATURES
- Lennard-Jones, Gay-Berne, Tersoff, many more potentials
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LAMMPS [LJ 2.5] | ATOM-Time Steps/s | LJ 2.5 | yes | 1.07E+08 | - | 2.15E+08 | 4.10E+08 | 7.72E+08 | ||||||||
| LAMMPS [LJ 2.5] | NRF | LJ 2.5 | yes | 1x | - | 2x | 4x | 7x | ||||||||
| LAMMPS [EAM] | ATOM-Time Steps/s | EAM | yes | 5.09E+07 | - | 9.98E+07 | 1.88E+08 | 3.51E+08 | ||||||||
| LAMMPS [EAM] | NRF | EAM | yes | 1x | - | 2x | 4x | 7x | ||||||||
| LAMMPS [ReaxFF/C] | ATOM-Time Steps/s | ReaxFF/C | yes | 3.41E+05 | - | 7.50E+05 | 1.39E+06 | 2.38E+06 | ||||||||
| LAMMPS [ReaxFF/C] | NRF | ReaxFF/C | yes | 1x | - | 3x | 5x | 12x | ||||||||
| LAMMPS [Rhodopsin] | ATOM-Time Steps/s | Rhodopsin | yes | 7.34E+06 | - | - | 1.26E+07 | 1.92E+07 | ||||||||
| LAMMPS [Rhodopsin] | NRF | Rhodopsin | yes | 1x | - | - | 2x | 3x | ||||||||
| LAMMPS [SNAP] | ATOM-Time Steps/s | SNAP | yes | 1.14E+05 | 2.45E+05 | 4.90E+05 | 9.80E+05 | 1.95E+06 | ||||||||
| LAMMPS [SNAP] | NRF | SNAP | yes | 1x | 2x | 5x | 9x | 17x | ||||||||
| LAMMPS [Tersoff] | ATOM-Time Steps/s | Tersoff | yes | 2.78E+07 | 4.45E+07 | 8.78E+07 | 1.70E+08 | 3.19E+08 | ||||||||
| LAMMPS [Tersoff] | NRF | Tersoff | yes | 1x | 2x | 3x | 6x | 11x |
MILC
Physics
Lattice Quantum Chromodynamics (LQCD) codes simulate how elemental particles are formed and bound by the “strong force” to create larger particles like protons and neutrons
VERSION
develop_d98de0c4
ACCELERATED FEATURES
- Staggered fermions, Krylov solvers, Gauge-link fattening
SCALABILITY
Multi-GPU and Multi-Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MILC | Total Time (Sec) | Apex Medium | no | 67,366 | 6,058 | 3,223 | 1,641 | 933 | ||||||||
| MILC | NRF | Apex Medium | yes | 1x | 12x | 23x | 45x | 79x |
NAMD
Molecular Dynamics
Designed for high-performance simulation of large molecular systems
VERSION
GPU, AMD CPU V 3.0a9 ; Intel CPU V 2.15a AVX512
ACCELERATED FEATURES
- Full electrostatics with PME and most simulation features
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NAMD [apoa1_npt_cuda] | Ave ns/day | apoa1_npt_cuda | yes | 19.12 | 86 | 171 | 344 | 691 | ||||||||
| NAMD [apoa1_npt_cuda] | NRF | apoa1_npt_cuda | yes | 1x | 4x | 9x | 18x | 36x | ||||||||
| NAMD [apoa1_nptsr_cuda] | Ave ns/day | apoa1_nptsr_cuda | yes | 19.63 | 91 | 182 | 365 | 733 | ||||||||
| NAMD [apoa1_nptsr_cuda] | NRF | apoa1_nptsr_cuda | yes | 1x | 5x | 9x | 19x | 37x | ||||||||
| NAMD [apoa1_nve_cuda] | Ave ns/day | apoa1_nve_cuda | yes | 20.64 | 130 | 260 | 525 | 1,058 | ||||||||
| NAMD [apoa1_nve_cuda] | NRF | apoa1_nve_cuda | yes | 1x | 6x | 13x | 25x | 51x | ||||||||
| NAMD [stmv_npt_cuda] | Ave ns/day | stmv_npt_cuda | yes | 1.78 | 7 | 14 | 29 | 58 | ||||||||
| NAMD [stmv_npt_cuda] | NRF | stmv_npt_cuda | yes | 1x | 4x | 8x | 16x | 32x | ||||||||
| NAMD [stmv_nptsr_cuda] | Ave ns/day | stmv_nptsr_cuda | yes | 1.80 | 7 | 15 | 30 | 60 | ||||||||
| NAMD [stmv_nptsr_cuda] | NRF | stmv_nptsr_cuda | yes | 1x | 4x | 8x | 17x | 33x | ||||||||
| NAMD [stmv_nve_cuda] | Ave ns/day | stmv_nve_cuda | yes | 1.95 | 9 | 18 | 37 | 74 | ||||||||
| NAMD [stmv_nve_cuda] | NRF | stmv_nve_cuda | yes | 1x | 5x | 9x | 19x | 38x |
Benchmarked in ensemble mode
RELION
Microscopy
Stand-alone computer program that employs an empirical Bayesianapproach to refinement of (multiple) 3D reconstructions or 2D class averages in electron cryo-microscopy (cryo-EM)
VERSION
3.1.3
ACCELERATED FEATURES
- Reduced memory requirements; high-resolution cryo-EM structure determination in a matter of day on a single workstation
SCALABILITY
Multi-GPU and Single Node
MORE INFORMATION
https://www2.mrc-lmb.cam.ac.uk/relion/index.php/Download_%26_install
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 |
|---|---|---|---|---|---|---|---|---|
| Relion [Plasmodium Ribosome] | Total Wall Clock (Sec) | MB numbers Plasmodium Ribosime on Relion-3.0 | no | 14,862 | 3,810 | 2,165 | 1,818 | 1,678 |
| Relion [Plasmodium Ribosome] | NRF | MB numbers Plasmodium Ribosime on Relion-3.0 | yes | 1x | 4x | 7x | 8x | 9x |
| Relion [Plasmodium Ribosome 2D] | Total Wall Clock (Sec) | Plasmodium Ribosome (2D) | no | 92,404 | 25,429 | 11,993 | 8,385 | 6,005 |
| Relion [Plasmodium Ribosome 2D] | NRF | Plasmodium Ribosome (2D) | yes | 1x | 4x | 10x | 14x | 19x |
RTM
Geoscience
Reverse time migration (RTM) modeling is a critical component in the seismic processing workflow of oil and gas exploration
VERSION
nvidia_2021_05
ACCELERATED FEATURES
- Batch algorithm
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RTM [Isotropic Radius 4] | Mcells/s | Isotropic Radius 4 | yes | 11,318 | 30,995 | ||||||||
| RTM [Isotropic Radius 4] | NRF | Isotropic Radius 4 | yes | 1x | 3x | ||||||||
| RTM [TTI Radius 8 1-pass] | Mcells/s | TTI Radius 8 1-pass | yes | 3,773 | 8,625 | ||||||||
| RTM [TTI Radius 8 1-pass] | NRF | TTI Radius 8 1-pass | yes | 1x | 2x | ||||||||
| RTM [TTI RX 2Pass mgpu] | Mcells/s | TTI RX 2Pass mgpu | yes | 3,773 | 6,406 | ||||||||
| RTM [TTI RX 2Pass mgpu] | NRF | TTI RX 2Pass mgpu | yes | 1x | 2x |
SPECFEM3D
Geoscience
Simulates Seismic wave propagation
VERSION
devel_0a5acff9
ACCELERATED FEATURES
- OpenCL and CUDA hardware accelerators, based on an automatic source-to-source transformation library
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x A40 | 2x A40 | 4x A40 | 8x A40 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SPECFEM3D | Total Time (Sec) | four_material_simple_model | no | 1,991 | 204 | 104 | 54 | 30 | ||||||||
| SPECFEM3D | NRF | four_material_simple_model | yes | 1x | 11x | 22x | 42x | 77x |
Engineering
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual Xeon Gold 6240@2.60GHz with 4x NVIDIA V100 SXM2 | FUN3D Benchmark: dpw_wbt0_crs-3.6Mn_5, CUDA Version: 11.4.2
Geoscience
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual Xeon Gold 6240@2.60GHz 4x NVIDIA V100 SXM2 | RTM Benchmark: Isotropic Radius 4, CUDA Version: 11.4.2 | SPENFEM3D Benchmark: four_material_simple_model, CUDA Version: 11.4.2
Microscopy and Molecular Dynamics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual Xeon Gold 6240@2.60GHz with 4x NVIDIA V100 SXM2 | Relion Benchmark: Plasmodium Ribosome (2D), CUDA Version: 11.4.2 | AMBER Benchmark: DC-Cellulose_NVE, CUDA Version: 11.4.2 | GROMACS Benchmark: Cellulose, CUDA Version: 11.4.2 | LAMMPS Benchmark: SNAP, CUDA Version: 11.4.2 | NAMD Benchmark: apoa1_nve_cuda, CUDA Version: 11.4.2
Physics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual Xeon Gold 6240@2.60GHz with 4x NVIDIA V100 SXM2 | Chroma Benchmark: szscl21_24_128, CUDA Version: 11.3.1 | GTC Benchmark: moi#proc.in, CUDA Version: 11.3.1 | MILC Benchmark: Apex Medium, CUDA Version: 11.4.2
Quantum Mechanics
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: Dual Xeon Gold 6240@2.60GHz with 4x NVIDIA V100 SXM2 | Quantum Espresso Benchmark: AUSURF112-jR, CUDA Version: 11.4.2 | ICON Benchmark: SLAM 191 levels 160 km resolution with radiation, CUDA Version: 11.3.1
Detailed V100 application performance data is located below in alphabetical order.
AMBER
Molecular Dynamics
Suite of programs to simulate molecular dynamics on biomolecule
VERSION
20.12-AT_21.10
ACCELERATED FEATURES
- PMEMD Explicit Solvent and GB Implicit Solvent
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AMBER [PME-Cellulose_NPT_4fs] | ns/day | DC-Cellulose_NPT | yes | 4.40 | 96 | 191 | 383 | 766 | 99 | 197 | 395 | 789 |
| AMBER [PME-Cellulose_NPT_4fs] | NRF | DC-Cellulose_NPT | yes | 1x | 22x | 44x | 87x | 174x | 22x | 45x | 90x | 179x |
| AMBER [PME-Cellulose_NVE_4fs] | ns/day | DC-Cellulose_NVE | yes | 4.41 | 103 | 205 | 410 | 821 | 105 | 211 | 421 | 843 |
| AMBER [PME-Cellulose_NVE_4fs] | NRF | DC-Cellulose_NVE | yes | 1x | 23x | 47x | 93x | 186x | 24x | 48x | 96x | 191x |
| AMBER [PME-FactorIX_NPT_4fs] | ns/day | DC-FactorIX_NPT | yes | 22.80 | 423 | 845 | 1,690 | 3,380 | 433 | 866 | 1,733 | 3,465 |
| AMBER [PME-FactorIX_NPT_4fs] | NRF | DC-FactorIX_NPT | yes | 1x | 19x | 37x | 74x | 148x | 19x | 38x | 76x | 152x |
| AMBER [PME-FactorIX_NVE_4fs] | ns/day | DC-FactorIX_NVE | yes | 23.13 | 451 | 903 | 1,806 | 3,611 | 464 | 928 | 1,856 | 3,711 |
| AMBER [PME-FactorIX_NVE_4fs] | NRF | DC-FactorIX_NVE | yes | 1x | 20x | 39x | 78x | 156x | 20x | 40x | 80x | 160x |
| AMBER [PME-JAC_NPT_4fs] | ns/day | DC-JAC_NPT | yes | 97.51 | 1,136 | 2,272 | 4,544 | 9,088 | 1,170 | 2,340 | 4,680 | 9,360 |
| AMBER [PME-JAC_NPT_4fs] | NRF | DC-JAC_NPT | yes | 1x | 12x | 23x | 47x | 93x | 12x | 24x | 48x | 96x |
| AMBER [PME-JAC_NVE_4fs] | ns/day | DC-JAC_NVE | yes | 99.02 | 1,193 | 2,385 | 4,771 | 9,542 | 1,247 | 2,494 | 4,988 | 9,976 |
| AMBER [PME-JAC_NVE_4fs] | NRF | DC-JAC_NVE | yes | 1x | 12x | 24x | 48x | 96x | 13x | 25x | 50x | 101x |
| AMBER [PME-STMV_NPT_4fs] | ns/day | DC-STMV_NPT | yes | 1.42 | 33 | 65 | 130 | 261 | 33 | 67 | 133 | 267 |
| AMBER [PME-STMV_NPT_4fs] | NRF | DC-STMV_NPT | yes | 1x | 23x | 46x | 92x | 183x | 23x | 47x | 94x | 188x |
Benchmarked in ensemble mode
Chroma
Physics
Lattice Quantum Chromodynamics (LQCD)
VERSION
V 2021.05
ACCELERATED FEATURES
- Wilson-clover fermions, Krylov solvers, Domain-decomposition
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Chroma | Total Time (Sec) | szscl21_24_128 | no | 1,115 | 165 | 31 | 17 | 10 | 142 | 28 | 15 | 13 |
| Chroma | NRF | szscl21_24_128 | yes | 1x | 7x | 37x | 68x | 111x | 8x | 41x | 77x | 85x |
FUN3D
Engineering
Suite of tools actively developed at NASA for Aeronautics and Space Technology by modeling fluid flow.
VERSION
13.7 (update 1)
ACCELERATED FEATURES
- Full range of Mach number regimes for the Reynolds-averaged Navier Stokes (RANS) formulation
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| FUN3D | Loop Time (Sec) | dpw_wbt0_crs-3.6Mn_5 | no | 524 | 99 | 50 | 26 | 15 | 88 | 45 | 24 | 14 |
| FUN3D | NRF | dpw_wbt0_crs-3.6Mn_5 | yes | 1x | 6x | 13x | 25x | 43x | 7x | 15x | 28x | 47x |
GROMACS
Molecular Dynamics
Simulation of biochemical molecules with complicated bond interactions
VERSION
2021.3
ACCELERATED FEATURES
- Implicit (5x), Explicit (2x) Solvent
SCALABILITY
Multi-GPU, Single Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x RTX6000 | 2x RTX6000 | 4x RTX6000 | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GROMACS [ADH Dodec] | ns/day | ADH Dodec | yes | 56 | 224 | 259 | 435 | - | 226 | 250 | 314 | 234 | - | 325 |
| GROMACS [ADH Dodec] | NRF | ADH Dodec | yes | 1x | 5x | 6x | 10x | - | 5x | 6x | 8x | 6x | - | 8x |
| GROMACS [Cellulose] | ns/day | Cellulose | yes | 16 | 61 | 91 | 145 | - | 53 | 75 | 86 | 63 | 85 | 96 |
| GROMACS [Cellulose] | NRF | Cellulose | yes | 1x | 4x | 9x | 14x | - | 3x | 5x | 8x | 5x | 6x | 9x |
| GROMACS [STMV] | ns/day | STMV | yes | 4 | 15 | 27 | 43 | 51 | 12 | 22 | 32 | 14 | 26 | 36 |
| GROMACS [STMV] | NRF | STMV | yes | 1x | 4x | 7x | 12x | 14x | 3x | 6x | 9x | 4x | 7x | 10x |
GTC
Physics
GTC is used for Gyrokinetic Particle Simulation of Turbulent Transport in Burning Plasmas.
VERSION
V 4.5 Updated
ACCELERATED FEATURES
- Push, shift, and collision
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GTC | Mpush/Sec | moi#proc.in | yes | 35 | 268 | 512 | 1,020 | 1,997 | 296 | 556 | 1,098 | 1,813 |
| GTC | NRF | moi#proc.in | yes | 1x | 8x | 15x | 30x | 58x | 9x | 16x | 32x | 53x |
ICON
Weather and Climate
A global unified atmosphere model for numerical weather prediction and climate modeling research.
VERSION
2.6.2+RRTMGP
ACCELERATED FEATURES
- Full model of dynamics and physics
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|
| ICON [SLAM 191 - 160KM - no radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution without radiation | no | 80 | 24 | 15 | 9 | 8 | 21 | 14 | 10 |
| ICON [SLAM 191 - 160KM - no radiation] | NRF | SLAM 191 levels 160 km resolution without radiation | yes | 1x | 3x | 5x | 8x | 10x | 4x | 6x | 8x |
| ICON [SLAM 191 - 160KM - with radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution with radiation | no | 146 | 46 | 26 | 15 | 11 | 39 | 23 | 15 |
| ICON [SLAM 191 - 160KM - with radiation] | NRF | SLAM 191 levels 160 km resolution with radiation | yes | 1x | 3x | 6x | 10x | 13x | 4x | 6x | 10x |
LAMMPS
Molecular Dynamics
Classical molecular dynamics package
VERSION
stable_29Sep2021
ACCELERATED FEATURES
- Lennard-Jones, Gay-Berne, Tersoff, many more potentials
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LAMMPS [LJ 2.5] | ATOM-Time Steps/s | LJ 2.5 | yes | 1.07E+08 | 3.36E+08 | 6.40E+08 | 1.22E+09 | 2.31E+09 | 3.26E+08 | 6.06E+08 | 1.14E+09 | 2.04E+09 |
| LAMMPS [LJ 2.5] | NRF | LJ 2.5 | yes | 1x | 3x | 6x | 12x | 22x | 3x | 6x | 11x | 20x |
| LAMMPS [EAM] | ATOM-Time Steps/s | EAM | yes | 5.09E+07 | 1.19E+08 | 2.61E+08 | 5.34E+08 | 9.66E+08 | 1.21E+08 | 2.56E+08 | 4.99E+08 | 8.80E+08 |
| LAMMPS [EAM] | NRF | EAM | yes | 1x | 2x | 5x | 10x | 19x | 2x | 5x | 10x | 17x |
| LAMMPS [ReaxFF/C] | ATOM-Time Steps/s | ReaxFF/C | yes | 3.41E+05 | 1.86E+06 | 3.45E+06 | 6.14E+06 | 9.78E+06 | 1.91E+06 | 3.54E+06 | 6.38E+06 | 9.88E+06 |
| LAMMPS [ReaxFF/C] | NRF | ReaxFF/C | yes | 1x | 9x | 17x | 31x | 49x | 9x | 18x | 32x | 49x |
| LAMMPS [Rhodopsin] | ATOM-Time Steps/s | Rhodopsin | yes | 7.34E+06 | - | 1.17E+07 | 2.17E+07 | 3.64E+07 | - | 1.21E+07 | 2.10E+07 | - |
| LAMMPS [Rhodopsin] | NRF | Rhodopsin | yes | 1x | - | 2x | 3x | 5x | - | 2x | 3x | - |
| LAMMPS [SNAP] | ATOM-Time Steps/s | SNAP | yes | 1.14E+05 | 1.46E+06 | 2.90E+06 | 5.73E+06 | 1.13E+07 | 1.40E+06 | 2.79E+06 | 5.57E+06 | 1.10E+07 |
| LAMMPS [SNAP] | NRF | SNAP | yes | 1x | 13x | 25x | 50x | 99x | 12x | 24x | 49x | 96x |
| LAMMPS [Tersoff] | ATOM-Time Steps/s | Tersoff | yes | 2.78E+07 | 2.21E+08 | 3.89E+08 | 7.91E+08 | 1.43E+09 | 2.32E+08 | 4.35E+08 | 8.08E+08 | 1.18E+09 |
| LAMMPS [Tersoff] | NRF | Tersoff | yes | 1x | 8x | 14x | 28x | 51x | 8x | 16x | 29x | 43x |
MILC
Physics
Lattice Quantum Chromodynamics (LQCD) codes simulate how elemental particles are formed and bound by the “strong force” to create larger particles like protons and neutrons
VERSION
develop_d98de0c4
ACCELERATED FEATURES
- Staggered fermions, Krylov solvers, Gauge-link fattening
SCALABILITY
Multi-GPU and Multi-Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MILC | Total Time (Sec) | Apex Medium | no | 67,366 | 5,304 | 2,533 | 1,331 | 745 | 4,130 | 2,215 | 1,212 | 1,160 |
| MILC | NRF | Apex Medium | yes | 1x | 14x | 29x | 56x | 99x | 18x | 33x | 61x | 64x |
NAMD
Molecular Dynamics
Designed for high-performance simulation of large molecular systems
VERSION
GPU, AMD CPU V 3.0a9 ; Intel CPU V 2.15a AVX512
ACCELERATED FEATURES
- Full electrostatics with PME and most simulation features
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x RTX6000 | 2x RTX6000 | 4x RTX6000 | 8x RTX6000 | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NAMD [apoa1_npt_cuda] | Ave ns/day | apoa1_npt_cuda | yes | 19.12 | 94 | 186 | 374 | 744 | 60 | 119 | 238 | 474 | 97 | 194 | 384 | 771 |
| NAMD [apoa1_npt_cuda] | NRF | apoa1_npt_cuda | yes | 1x | 5x | 10x | 20x | 39x | 3x | 6x | 12x | 25x | 5x | 10x | 20x | 40x |
| NAMD [apoa1_nptsr_cuda] | Ave ns/day | apoa1_nptsr_cuda | yes | 19.63 | 97 | 196 | 391 | 783 | 63 | 127 | 253 | 507 | 101 | 202 | 402 | 801 |
| NAMD [apoa1_nptsr_cuda] | NRF | apoa1_nptsr_cuda | yes | 1x | 5x | 10x | 20x | 40x | 3x | 6x | 13x | 26x | 5x | 10x | 20x | 41x |
| NAMD [apoa1_nve_cuda] | Ave ns/day | apoa1_nve_cuda | yes | 20.64 | 128 | 256 | 512 | 1,030 | 85 | 170 | 338 | 675 | 131 | 254 | 513 | 1,028 |
| NAMD [apoa1_nve_cuda] | NRF | apoa1_nve_cuda | yes | 1x | 6x | 12x | 25x | 50x | 4x | 8x | 16x | 33x | 6x | 12x | 25x | 50x |
| NAMD [stmv_npt_cuda] | Ave ns/day | stmv_npt_cuda | yes | 1.78 | 8 | 15 | 31 | 61 | 5 | 10 | 20 | 40 | 8 | 16 | 32 | 64 |
| NAMD [stmv_npt_cuda] | NRF | stmv_npt_cuda | yes | 1x | 4x | 9x | 17x | 34x | 3x | 6x | 11x | 23x | 5x | 9x | 18x | 36x |
| NAMD [stmv_nptsr_cuda] | Ave ns/day | stmv_nptsr_cuda | yes | 1.80 | 8 | 16 | 32 | 63 | 5 | 11 | 21 | 42 | 8 | 17 | 33 | 65 |
| NAMD [stmv_nptsr_cuda] | NRF | stmv_nptsr_cuda | yes | 1x | 4x | 9x | 18x | 35x | 3x | 6x | 12x | 23x | 5x | 9x | 18x | 36x |
| NAMD [stmv_nve_cuda] | Ave ns/day | stmv_nve_cuda | yes | 1.95 | 9 | 18 | 36 | 70 | 6 | 12 | 25 | 50 | 9 | 18 | 36 | 73 |
| NAMD [stmv_nve_cuda] | NRF | stmv_nve_cuda | yes | 1x | 4x | 9x | 18x | 36x | 3x | 6x | 13x | 25x | 5x | 9x | 19x | 37x |
Benchmarked in ensemble mode
NV-WRFg
Numerical Weather Prediction
Numerical weather prediction system designed for both atmospheric research and operational forecasting applications
VERSION
3.8.1 NCAR (CPU) / 3.8.1 WRFg 10_28 (GPU)
ACCELERATED FEATURES
- Dynamics modules
- Several Physics modules
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 4x V100 SXM2 32GB | 4x V100S PCIe 32GB |
|---|---|---|---|---|---|---|
| NV-WRFg | Seconds / Timestamps | Conus_2.5k_JA | no | 6 | 0.62 | 0.68 |
| NV-WRFg | NRF | Conus_2.5k_JA | yes | 1x | 10x | 9x |
Quantum Espresso
Material Science (Quantum Chemistry)
An Open-source suite of computer codes for electronic structure calculations and materials modeling at the nanoscale
VERSION
CPU 6.7; GPU 6.8
ACCELERATED FEATURES
- linear algebra (matix multiply)
- explicit computational kernels
- 3D FFTs
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Quantum Espressso | Total CPU Time (Sec) | AUSURF112-jR | no | 765 | 293 | 153 | 99 | 73 | 295 | 163 | 112 | 93 |
| Quantum Espressso | NRF | AUSURF112-jR | yes | 1x | 3x | 6x | 9x | 12x | 3x | 5x | 8x | 9x |
RELION
Microscopy
Stand-alone computer program that employs an empirical Bayesianapproach to refinement of (multiple) 3D reconstructions or 2D class averages in electron cryo-microscopy (cryo-EM)
VERSION
3.1.3
ACCELERATED FEATURES
- Reduced memory requirements; high-resolution cryo-EM structure determination in a matter of day on a single workstation
SCALABILITY
Multi-GPU and Single Node
MORE INFORMATION
https://www2.mrc-lmb.cam.ac.uk/relion/index.php/Download_%26_install
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Relion [Plasmodium Ribosome] | Total Wall Clock (Sec) | MB numbers Plasmodium Ribosime on Relion-3.0 | no | 14,862 | 3,569 | 2,173 | - | 1,673 | 3,559 | 2,190 | - | 1,718 |
| Relion [Plasmodium Ribosome] | NRF | MB numbers Plasmodium Ribosime on Relion-3.0 | yes | 1x | 4x | 7x | - | 9x | 4x | 7x | - | 9x |
| Relion [Plasmodium Ribosome 2D] | Total Wall Clock (Sec) | Plasmodium Ribosome (2D) | no | 92,404 | - | 17,797 | 12,110 | 8,463 | 34,191 | 18,472 | 12,393 | 8,695 |
| Relion [Plasmodium Ribosome 2D] | NRF | Plasmodium Ribosome (2D) | yes | 1x | - | 6x | 10x | 14x | 3x | 5x | 9x | 13x |
RTM
Geoscience
Reverse time migration (RTM) modeling is a critical component in the seismic processing workflow of oil and gas exploration
VERSION
nvidia_2021_05
ACCELERATED FEATURES
- Batch algorithm
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RTM [Isotropic Radius 4] | Mcells/s | Isotropic Radius 4 | yes | 11,318 | 38,158 | 76,043 | 152,116 | 304,085 | 46,046 | 92,037 | 183,565 | 366,995 |
| RTM [Isotropic Radius 4] | NRF | Isotropic Radius 4 | yes | 1x | 3x | 7x | 13x | 27x | 4x | 8x | 16x | 32x |
| RTM [TTI Radius 8 1-pass] | Mcells/s | TTI Radius 8 1-pass | yes | 3,773 | 8,566 | 16,913 | 32,992 | 65,275 | 9,246 | 18,269 | 36,291 | 72,463 |
| RTM [TTI Radius 8 1-pass] | NRF | TTI Radius 8 1-pass | yes | 1x | 2x | 4x | 9x | 17x | 2x | 5x | 10x | 19x |
| RTM [TTI RX 2Pass mgpu] | Mcells/s | TTI RX 2Pass mgpu | yes | 3,773 | 7,182 | 14,245 | 28,261 | 56,416 | 8,512 | 16,881 | 33,548 | 66,986 |
| RTM [TTI RX 2Pass mgpu] | NRF | TTI RX 2Pass mgpu | yes | 1x | 2x | 4x | 7x | 15x | 2x | 4x | 9x | 18x |
SPECFEM3D
Geoscience
Simulates Seismic wave propagation
VERSION
devel_0a5acff9
ACCELERATED FEATURES
- OpenCL and CUDA hardware accelerators, based on an automatic source-to-source transformation library
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 1x V100 SXM2 32GB | 2x V100 SXM2 32GB | 4x V100 SXM2 32GB | 8x V100 SXM2 32GB | 1x V100S PCIe 32GB | 2x V100S PCIe 32GB | 4x V100S PCIe 32GB | 8x V100S PCIe 32GB |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SPECFEM3D | Total Time (Sec) | four_material_simple_model | no | 1,991 | 158 | 82 | 43 | 25 | 132 | 68 | 37 | 22 |
| SPECFEM3D | NRF | four_material_simple_model | yes | 1x | 14x | 28x | 52x | 90x | 17x | 33x | 62x | 104x |
Engineering
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: same CPU server with 4x NVIDIA T4 | FUN3D Benchmark: dpw_wbt0_crs-3.6Mn_5, CUDA Version: 11.4.2
Geoscience
CPU Server: Dual Xeon Gold 6240@2.60GHz | GPU Server: same CPU server with 4x NVIDIA T4 | RTM Benchmark: Isotropic Radius 4, CUDA Version: 11.4.2 | SPECFEM3D Benchmark: four_material_simple_model, CUDA Version 11.4.2
Microscopy and Molecular Dynamics
CPU Server: Dual Xeon Gold 6240@2.60GHz, GPU Server: same CPU server with 4x NVIDIA T4 | AMBER Benchmark: DC-STMV_NPT, CUDA Version: 11.4.2 | Gromacs Benchmark: ADH Dodec, CUDA Version: 11.4.2 | NAMD Benchmark: apoa1_nve_cuda, CUDA Version: 11.4.2 | Relion Benchmark: Plasmodium Ribosome (2D), CUDA Version: 11.4.2
Physics
CPU Server: Dual Xeon Gold 6240@2.60GHz, GPU Server: same CPU server with 4x NVIDIA T4 Chroma Benchmark: szscl21_24_128, CUDA Version: 11.3.1 | GTC Benchmark: moi#proc.in, CUDA Version: 11.3.1 | MILC Benchmark: Apex Medium, CUDA Version: 11.4.2
Detailed T4 application performance data is located below in alphabetical order.
AMBER
Molecular Dynamics
Suite of programs to simulate molecular dynamics on biomolecule
VERSION
20.12-AT_21.10
ACCELERATED FEATURES
- PMEMD Explicit Solvent and GB Implicit Solvent
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| AMBER [PME-Cellulose_NPT_4fs] | ns/day | DC-Cellulose_NPT | yes | 4.40 | 63 | 126 | 251 |
| AMBER [PME-Cellulose_NPT_4fs] | NRF | DC-Cellulose_NPT | yes | 1x | 14x | 29x | 57x |
| AMBER [PME-Cellulose_NVE_4fs] | ns/day | DC-Cellulose_NVE | yes | 4.41 | 64 | 129 | 257 |
| AMBER [PME-Cellulose_NVE_4fs] | NRF | DC-Cellulose_NVE | yes | 1x | 15x | 29x | 58x |
| AMBER [PME-FactorIX_NPT_4fs] | ns/day | DC-FactorIX_NPT | yes | 22.80 | 308 | 616 | 1,232 |
| AMBER [PME-FactorIX_NPT_4fs] | NRF | DC-FactorIX_NPT | yes | 1x | 14x | 27x | 54x |
| AMBER [PME-FactorIX_NVE_4fs] | ns/day | DC-FactorIX_NVE | yes | 23.13 | 317 | 635 | 1,270 |
| AMBER [PME-FactorIX_NVE_4fs] | NRF | DC-FactorIX_NVE | yes | 1x | 14x | 27x | 55x |
| AMBER [PME-JAC_NPT_4fs] | ns/day | DC-JAC_NPT | yes | 97.51 | 1,141 | 2,282 | 4,564 |
| AMBER [PME-JAC_NPT_4fs] | NRF | DC-JAC_NPT | yes | 1x | 12x | 23x | 47x |
| AMBER [PME-JAC_NVE_4fs] | ns/day | DC-JAC_NVE | yes | 99.02 | 1,179 | 2,358 | 4,715 |
| AMBER [PME-JAC_NVE_4fs] | NRF | DC-JAC_NVE | yes | 1x | 12x | 24x | 48x |
| AMBER [PME-STMV_NPT_4fs] | ns/day | DC-STMV_NPT | yes | 1.42 | 22 | 44 | 87 |
| AMBER [PME-STMV_NPT_4fs] | NRF | DC-STMV_NPT | yes | 1x | 15x | 31x | 62x |
Benchmarked in ensemble mode
Chroma
Physics
Lattice Quantum Chromodynamics (LQCD)
VERSION
V 2021.05
ACCELERATED FEATURES
- Wilson-clover fermions, Krylov solvers, Domain-decomposition
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| Chroma | Total Time (Sec) | szscl21_24_128 | no | 1,115 | 117 | 40 | 26 |
| Chroma | NRF | szscl21_24_128 | yes | 1x | 10x | 28x | 44x |
FUN3D
Engineering
Suite of tools actively developed at NASA for Aeronautics and Space Technology by modeling fluid flow.
VERSION
13.7 (update 1)
ACCELERATED FEATURES
- Full range of Mach number regimes for the Reynolds-averaged Navier Stokes (RANS) formulation
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| Fun3D | Loop Time (Sec) | dpw_wbt0_crs-3.6Mn_5 | no | 524 | 285 | 145 | 74 |
| Fun3D | NRF | dpw_wbt0_crs-3.6Mn_5 | yes | 1x | 2x | 4x | 9x |
GROMACS
Molecular Dynamics
Simulation of biochemical molecules with complicated bond interactions
VERSION
2021.3
ACCELERATED FEATURES
- Implicit (5x), Explicit (2x) Solvent
SCALABILITY
Multi-GPU, Single Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| GROMACS [ADH Dodec] | ns/day | ADH Dodec | yes | 56 | 125 | 235 | - |
| GROMACS [ADH Dodec] | NRF | ADH Dodec | yes | 1x | 3x | 6x | - |
| GROMACS [Cellulose] | ns/day | Cellulose | yes | 16 | 39 | 61 | 69 |
| GROMACS [Cellulose] | NRF | Cellulose | yes | 1x | 2x | 4x | 5x |
| GROMACS [STMV] | ns/day | STMV | yes | 4 | 10 | 17 | 25 |
| GROMACS [STMV] | NRF | STMV | yes | 1x | 2x | 4x | 7x |
GTC
Physics
GTC is used for Gyrokinetic Particle Simulation of Turbulent Transport in Burning Plasmas.
VERSION
V 4.5 Updated
ACCELERATED FEATURES
- Push, shift, and collision
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| GTC | Mpush/Sec | moi#proc.in | yes | 35 | 232 | 463 | 853 |
| GTC | NRF | moi#proc.in | yes | 1x | 7x | 13x | 25x |
ICON
Weather and Climate
A global unified atmosphere model for numerical weather prediction and climate modeling research.
VERSION
2.6.2+RRTMGP
ACCELERATED FEATURES
- Full model of dynamics and physics
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| ICON [SLAM 191 - 160KM - no radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution without radiation | no | 80 | 36 | 21 | - |
| ICON [SLAM 191 - 160KM - no radiation] | NRF | SLAM 191 levels 160 km resolution without radiation | yes | 1x | 2x | 4x | - |
| ICON [SLAM 191 - 160KM - with radiation] | Integrate_nh (sec) | SLAM 191 levels 160 km resolution with radiation | no | 146 | 81 | 43 | 31 |
| ICON [SLAM 191 - 160KM - with radiation] | NRF | SLAM 191 levels 160 km resolution with radiation | yes | 1x | 2x | 3x | 5x |
MILC
Physics
Lattice Quantum Chromodynamics (LQCD) codes simulate how elemental particles are formed and bound by the “strong force” to create larger particles like protons and neutrons
VERSION
develop_d98de0c4
ACCELERATED FEATURES
- Staggered fermions, Krylov solvers, Gauge-link fattening
SCALABILITY
Multi-GPU and Multi-Node
MORE INFORMATION
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| MILC | Total Time (Sec) | Apex Medium | no | 67,366 | 7,459 | 3,854 | 2,626 |
| MILC | NRF | Apex Medium | yes | 1x | 10x | 19x | 28x |
NAMD
Molecular Dynamics
Designed for high-performance simulation of large molecular systems
VERSION
GPU, AMD CPU V 3.0a9 ; Intel CPU V 2.15a AVX512
ACCELERATED FEATURES
- Full electrostatics with PME and most simulation features
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| NAMD [apoa1_npt_cuda] | Ave ns/day | apoa1_npt_cuda | yes | 19.12 | 55 | 110 | 218 |
| NAMD [apoa1_npt_cuda] | NRF | apoa1_npt_cuda | yes | 1x | 3x | 6x | 11x |
| NAMD [apoa1_nptsr_cuda] | Ave ns/day | apoa1_nptsr_cuda | yes | 19.63 | 57 | 115 | 227 |
| NAMD [apoa1_nptsr_cuda] | NRF | apoa1_nptsr_cuda | yes | 1x | 3x | 6x | 12x |
| NAMD [apoa1_nve_cuda] | Ave ns/day | apoa1_nve_cuda | yes | 20.64 | 74 | 148 | 292 |
| NAMD [apoa1_nve_cuda] | NRF | apoa1_nve_cuda | yes | 1x | 4x | 7x | 14x |
| NAMD [stmv_npt_cuda] | Ave ns/day | stmv_npt_cuda | yes | 1.78 | 4 | 9 | 17 |
| NAMD [stmv_npt_cuda] | NRF | stmv_npt_cuda | yes | 1x | 2x | 5x | 10x |
| NAMD [stmv_nptsr_cuda] | Ave ns/day | stmv_nptsr_cuda | yes | 1.80 | 4 | 9 | 18 |
| NAMD [stmv_nptsr_cuda] | NRF | stmv_nptsr_cuda | yes | 1x | 2x | 5x | 10x |
| NAMD [stmv_nve_cuda] | Ave ns/day | stmv_nve_cuda | yes | 1.95 | 5 | 10 | 20 |
| NAMD [stmv_nve_cuda] | NRF | stmv_nve_cuda | yes | 1x | 3x | 5x | 10x |
Benchmarked in ensemble mode
NV-WRFg
Numerical Weather Prediction
Numerical weather prediction system designed for both atmospheric research and operational forecasting applications
VERSION
3.8.1 NCAR (CPU) / 3.8.1 WRFg 10_28 (GPU)
ACCELERATED FEATURES
- Dynamics modules
- Several Physics modules
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 8x T4 PCIe |
|---|---|---|---|---|---|
| NV-WRFg | Seconds / Timestamps | Conus_2.5k_JA | no | 5.51 | 0.90 |
| NV-WRFg | NRF | Conus_2.5k_JA | yes | 1x | 7x |
RELION
Microscopy
Stand-alone computer program that employs an empirical Bayesianapproach to refinement of (multiple) 3D reconstructions or 2D class averages in electron cryo-microscopy (cryo-EM)
VERSION
3.1.3
ACCELERATED FEATURES
- Reduced memory requirements; high-resolution cryo-EM structure determination in a matter of day on a single workstation
SCALABILITY
Multi-GPU and Single Node
MORE INFORMATION
https://www2.mrc-lmb.cam.ac.uk/relion/index.php/Download_%26_install
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| Relion [Plasmodium Ribosome] | Total Wall Clock (Sec) | MB numbers Plasmodium Ribosime on Relion-3.0 | no | 14,862 | 3,377 | 2,567 | 1,815 |
| Relion [Plasmodium Ribosome] | NRF | MB numbers Plasmodium Ribosime on Relion-3.0 | yes | 1x | 4x | 6x | 8x |
| Relion [Plasmodium Ribosome 2D] | Total Wall Clock (Sec) | Plasmodium Ribosome (2D) | no | 92,404 | 30,651 | 18,100 | 11,239 |
| Relion [Plasmodium Ribosome 2D] | NRF | Plasmodium Ribosome (2D) | yes | 1x | 3x | 6x | 10x |
RTM
Geoscience
Reverse time migration (RTM) modeling is a critical component in the seismic processing workflow of oil and gas exploration
VERSION
nvidia_2021_05
ACCELERATED FEATURES
- Batch algorithm
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| RTM [Isotropic Radius 4] | Mcells/s | Isotropic Radius 4 | yes | 11,318 | 22,001 | 44,043 | 88,054 |
| RTM [Isotropic Radius 4] | NRF | Isotropic Radius 4 | yes | 1x | 2x | 4x | 8x |
| RTM [TTI Radius 8 1-pass] | Mcells/s | TTI Radius 8 1-pass | yes | 3,773 | 5,906 | 11,541 | 22,948 |
| RTM [TTI Radius 8 1-pass] | NRF | TTI Radius 8 1-pass | yes | 1x | 2x | 3x | 6x |
| RTM [TTI RX 2Pass mgpu] | Mcells/s | TTI RX 2Pass mgpu | yes | 3,773 | - | 9,799 | 19,565 |
| RTM [TTI RX 2Pass mgpu] | NRF | TTI RX 2Pass mgpu | yes | 1x | - | 3x | 5x |
SPECFEM3D
Geoscience
Simulates Seismic wave propagation
VERSION
devel_0a5acff9
ACCELERATED FEATURES
- OpenCL and CUDA hardware accelerators, based on an automatic source-to-source transformation library
| Application | Metric | Test Modules | Bigger is better | Dual Cascade Lake 6240 (CPU-Only) | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| SPECFEM3D | Total Time (Sec) | four_material_simple_model | no | 1,991 | 240 | 122 | 65 |
| SPECFEM3D | NRF | four_material_simple_model | yes | 1x | 9x | 19x | 35x |