NVIDIA HPC Application Performance
For Deep Learning performance, please go here.
This page is designed to provide the latest performance benchmark data for key High Performance Computing applications. Modern HPC data centers are key to solving some of the world’s most important scientific and engineering challenges. NVIDIA® Tesla® accelerated computing platform powers these modern data centers with the industry-leading applications to accelerate HPC and AI workloads. The Tesla 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.
A single GPU-accelerated server can replace over 100 CPU-only servers. 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 use measured benchmark 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 based on the applications running on the server.
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| Abaqus Direct Solver | LS-EPP-Combined-WC-Mkl (RR) | Total Time (Sec) | no | 4376.42 | 1856.00 | 1479.00 | 1856.00 | 1479.00 | ||
| Abaqus Direct Solver | LS-EPP-Combined-WC-Mkl (RR) | NRF | yes | 1.00 | 4.00 | 9.00 | 4.00 | 9.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| ANSYS Fluent | Waterjacket | Total Time (Sec) | no | 1221.77 | 763.89 | 590.54 | 811.54 | |||
| ANSYS Fluent | Waterjacket | NRF | yes | 1.00 | 3.00 | 4.00 | 3.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| Fun3D | dpw_wbt0_crs-3.6Mn_5 | Loop Time (Sec) | no | 612.00 | 50.00 | 26.00 | 18.00 | 49.00 | 25.00 | 18.00 |
| Fun3D | dpw_wbt0_crs-3.6Mn_6 | NRF | yes | 1.00 | 14.00 | 26.00 | 38.00 | 14.00 | 27.00 | 38.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| RTM | Isotropic Radius 4 | Mcells/s | yes | 9432.00 | 84307.00 | 168405.00 | 336862 | 83989.00 | 167950.00 | 336021 |
| RTM | Isotropic Radius 5 | NRF | yes | 1.00 | 9.00 | 18.00 | 36.00 | 9.00 | 18.00 | 36.00 |
| RTM | TTI Radius 8 1-pass | Mcells/s | yes | 3144.00 | 15646.00 | 31167.00 | 62231 | 16635.00 | 33118.00 | 65779 |
| RTM | TTI Radius 8 1-pass | NRF | yes | 1.00 | 5.00 | 10.00 | 20.00 | 5.00 | 11.00 | 21.00 |
| RTM | TTI RX 2Pass mgpu | Mcells/s | yes | 3144.00 | 15023.00 | 29870.00 | 59594 | 15096.00 | 30003.00 | 59790 |
| RTM | TTI RX 2Pass mgpu | NRF | yes | 1.00 | 5.00 | 10.00 | 19.00 | 5.00 | 10.00 | 19.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| SPECFEM3D | four_material_simple_model | Total Time (Sec) | no | 2807.00 | 77.00 | 41.00 | 25.00 | 77.00 | 41.00 | 26.00 |
| SPECFEM3D | four_material_simple_model | NRF | yes | 1.00 | 43.00 | 82.00 | 134.00 | 43.00 | 82.00 | 129.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| Cloverleaf | bm32 | Wall Clock (Sec) | no | 2519.68 | 119.65 | 103.83 | 112.70 | 99.50 | ||
| Cloverleaf | bm33 | NRF | yes | 1.00 | 16.00 | 18.00 | 17.00 | 19.00 | ||
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| HPCG | 256x256x256 local size | GFLOPS | yes | 25.85 | 293.00 | 575.96 | 1056.00 | 293.00 | 575.96 | 1056.00 |
| HPCG | 256x256x256 local size | NRF | yes | 1.00 | 11.00 | 22.00 | 41.00 | 11.00 | 22.00 | 41.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| Linpack | HPL.dat NB=[256] for GPU server NB=[192] for CPU server |
GFLOPS | yes | 1813 | 10090 | 19880 | 24155 | 10090 | 19880 | 24155 |
| Linpack | HPL.dat NB=[256] for GPU server NB=[192] for CPU server |
NRF | yes | 1.00 | 5.00 | 11.00 | 13.00 | 5.00 | 11.00 | 13.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| MiniFE | 350x350x350 | Total CG Time (Sec) | no | 21.35 | 2.94 | 1.44 | 0.82 | 2.94 | 1.44 | 0.82 |
| MiniFE | 350x350x351 | NRF | yes | 1.00 | 7.00 | 15.00 | 26.00 | 7.00 | 15.00 | 26.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| Relion | MB numbers Plasmodium Ribosime on Relion-2.1 |
1/Minutes | yes | 0.00146 | 0.0147 | 0.0161 | 0.0142 | 0.0161 | ||
| Relion | MB numbers Plasmodium Ribosime on Relion-2.2 |
NRF | yes | 1.00 | 11.00 | 12.00 | 10.00 | 12.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| AMBER | DHFR (NVE) (AKA JAC) | ns/day | yes | 103.00 | 2150.00 | 4300.00 | 8600.00 | 2400.00 | 4800.00 | 9600.00 |
| AMBER | DHFR (NVE) (AKA JAC) | NRF | yes | 1.00 | 21.00 | 42.00 | 83.00 | 23.00 | 47.00 | 93.00 |
| AMBER | Factor IX (NPT) | ns/day | yes | 25.00 | 748.00 | 1496.00 | 2992.00 | 806.00 | 1612.00 | 3224.00 |
| AMBER | Factor IX (NPT) | NRF | yes | 1.00 | 30.00 | 60.00 | 120.00 | 32.00 | 64.00 | 129.00 |
| AMBER | PME-Cellulose_NVE | ns/day | yes | 5.20 | 212.00 | 424.00 | 848.00 | 226.00 | 452.00 | 904.00 |
| AMBER | PME-Cellulose_NVE | NRF | yes | 1.00 | 41.00 | 82.00 | 163.00 | 43.00 | 87.00 | 174.00 |
| AMBER | STMV (NPT) | ns/day | yes | 1.80 | 64.00 | 128.00 | 256.00 | 66.00 | 132.00 | 264.00 |
| AMBER | STMV (NPT) | NRF | yes | 1.00 | 36.00 | 71.00 | 142.00 | 37.00 | 73.00 | 147.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| GROMACS | ADH Dodec | ns/day | yes | 54.00 | 176.00 | 193.00 | 175.00 | 201.00 | ||
| GROMACS | ADH Dodec | NRF | yes | 1.00 | 5.00 | 9.00 | 5.00 | 9.00 | ||
| GROMACS | Cellulose | ns/day | yes | 15.00 | 50.00 | 54.00 | ||||
| GROMACS | Cellulose | NRF | yes | 1.00 | 5.00 | 5.00 | ||||
| GROMACS (Projected) | STMV | ns/day | yes | 3.50 | 15 | 34 | ||||
| GROMACS (Projected) | STMV | NRF | yes | 1.00 | 5.00 | 11.00 | ||||
| GROMACS | STMV | ns/day | yes | 3.50 | 16.00 | 15.00 | ||||
| GROMACS | STMV | NRF | yes | 1.00 | 5.00 | 5.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| HOOMD-Blue | microsphere | Ave. TPS | yes | 11.89 | 298.15 | 371.06 | 466.88 | 329.43 | 506.09 | 688.99 |
| HOOMD-Blue | microsphere | NRF | yes | 1.00 | 28.00 | 35.00 | 44.00 | 31.00 | 48.00 | 65.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| LAMMPS | EAM | ATOM-Time Steps/s | yes | 51790411 | 150151168 | 277921792 | 411025408 | 163856384 | 308977664 | 538206208 |
| LAMMPS | EAM | NRF | yes | 1.00 | 3.00 | 6.00 | 10.00 | 4.00 | 7.00 | 12.00 |
| LAMMPS | LJ 2.6 | ATOM-Time Steps/s | yes | 96625195 | 487342080 | 976592896 | 1229783040 | 514015232 | 1122000896 | 2018770944 |
| LAMMPS | LJ 2.6 | NRF | yes | 1.00 | 6.00 | 11.00 | 14.00 | 6.00 | 12.00 | 22.00 |
| LAMMPS | ReaxFF/C | ATOM-Time Steps/s | yes | 332697 | 2253472 | 3771040 | 5947466 | 2378146 | 3871433 | 6071206 |
| LAMMPS | ReaxFF/C | NRF | yes | 1.00 | 15.00 | 25.00 | 39.00 | 15.00 | 25.00 | 39.00 |
| LAMMPS | Tersoff | ATOM-Time Steps/s | yes | 39247062 | 380018688 | 672522240 | 824524800 | 423657472 | 769769472 | 1185472512 |
| LAMMPS | Tersoff | NRF | yes | 1.00 | 10.00 | 18.00 | 22.00 | 11.00 | 20.00 | 31.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| NAMD | apoa1_npt_cuda | Ave ns/day | yes | 4.10 | 67.00 | 77.00 | 70.00 | 81.00 | ||
| NAMD | apoa1_npt_cuda | NRF | yes | 1.00 | 117.00 | 134.00 | 122.00 | 141.00 | ||
| NAMD | apoa1_nptsr_cuda | Ave ns/day | yes | 4.10 | 74.00 | 86.00 | 77.00 | 90.00 | ||
| NAMD | apoa1_nptsr_cuda | NRF | yes | 1.00 | 129.00 | 150.00 | 134.00 | 157.00 | ||
| NAMD | apoa1_nve_cuda | Ave ns/day | yes | 4.40 | 79.00 | 82.00 | 81.00 | 93.00 | ||
| NAMD | apoa1_nve_cuda | NRF | yes | 1.00 | 92.00 | 95.00 | 94.00 | 108.00 | ||
| NAMD | stmv_npt_cuda | Ave ns/day | yes | 0.38 | 7.10 | 7.90 | 6.50 | 7.10 | ||
| NAMD | stmv_npt_cuda | NRF | yes | 1.00 | 32.00 | 35.00 | 29.00 | 32.00 | ||
| NAMD | stmv_nptsr_cuda | Ave ns/day | yes | 0.38 | 7.90 | 8.80 | 6.80 | 7.60 | ||
| NAMD | stmv_nptsr_cuda | NRF | yes | 1.00 | 37.00 | 41.00 | 32.00 | 36.00 | ||
| NAMD | stmv_nve_cuda | Ave ns/day | yes | 0.38 | 8.50 | 9.30 | 7.40 | 8.50 | ||
| NAMD | stmv_nve_cuda | NRF | yes | 1.00 | 34.00 | 37.00 | 30.00 | 34.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| Chroma | szscl21_24_128 | Total Time (Sec) | no | 1140.00 | 75.00 | 19.00 | 18.00 | 81.00 | 19.00 | 14.00 |
| Chroma | szscl21_24_129 | NRF | yes | 1.00 | 28.00 | 109.00 | 116.00 | 26.00 | 109.00 | 149.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| GTC | moi#proc.in | Mpush/Sec | yes | 35.00 | 439.00 | 846.00 | 1430.00 | 457.00 | 877.00 | 1667.00 |
| GTC | moi#proc.in | NRF | yes | 1.00 | 13.00 | 25.00 | 41.00 | 13.00 | 25.00 | 48.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| MILC | Apex Medium | Total Time (Sec) | no | 72194.00 | 3376 | 1637 | 1473 | 3292 | 1623 | 927 |
| MILC | Apex Medium | NRF | yes | 1.00 | 25.00 | 51.00 | 57.00 | 25.00 | 52.00 | 90.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| QUDA | QPhil Dslash Wilson-Clover Precision: Single; Gau. Cmprsn/Recon: 12; Probl. Size 32x32x32x64 |
Dslash GFLOPS | yes | 111.60 | 2672.00 | 4760.58 | 5238.3 | 2663.97 | 5024.36 | 6292.29 |
| QUDA | QPhil Dslash Wilson-Clover Precision: Single; Gau. Cmprsn/Recon: 12; Probl. Size 32x32x32x64 |
NRF | yes | 1.00 | 31.00 | 56.00 | 62.00 | 31.00 | 59.00 | 74.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| QE | AUSURF112-jR | Total CPU Time (Sec) | no | 740.00 | 200.00 | 99.00 | 190.00 | 94.00 | ||
| QE | AUSURF112-jR | NRF | yes | 1.00 | 6.00 | 13.00 | 7.00 | 14.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| VASP | B.hR105 | Elapsed Time (Sec) | no | 661.00 | 123.00 | 80.00 | 125.00 | 84.00 | ||
| VASP | B.hR106 | NRF | yes | 1.00 | 22.00 | 34.00 | 22.00 | 33.00 | ||
| VASP | Si-Huge | Elapsed Time (Sec) | no | 4364.00 | 2094.00 | 1591.00 | 2263.00 | 1771.00 | ||
| VASP | Si-Huge | NRF | yes | 1.00 | 4.00 | 9.00 | 3.00 | 9.00 |
| Application | Test Modules | Metric | Bigger is better |
Dual Skylake | 2x V100 16GB PCIe |
4x V100 16GB PCIe |
8x V100 16GB PCIe |
2x V100 16GB SXM2 |
4x V100 16GB SXM2 |
8x V100 16GB SXM2 |
|---|---|---|---|---|---|---|---|---|---|---|
| WRF | Conus_2.5k_JA | Seconds / Timestamps | no | 5.20 | 0.68 | 0.52 | 0.62 | 0.38 | ||
| WRF | Conus_2.5k_JA | NRF | yes | 1.00 | 9.00 | 11.00 | 9.00 | 15.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| Fun3D | dpw_wbt0_crs-3.6Mn_5 | Loop Time (Sec) | no | 612.00 | 268.00 | 132.00 | 68.00 |
| Fun3D | dpw_wbt0_crs-3.6Mn_6 | NRF | yes | 1.00 | 2.00 | 5.00 | 10.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| RTM | Isotropic Radius 4 | Mcells/s | yes | 9432.00 | 21987.00 | 43948.00 | 87930.00 |
| RTM | Isotropic Radius 5 | NRF | yes | 1.00 | 2.00 | 5.00 | 9.00 |
| RTM | TTI Radius 8 1-pass | Mcells/s | yes | 3144.00 | 5520.00 | 10962.00 | 21749.00 |
| RTM | TTI Radius 8 1-pass | NRF | yes | 1.00 | 2.00 | 3.00 | 7.00 |
| RTM | TTI RX 2Pass mgpu | Mcells/s | yes | 3144.00 | 4945.00 | 9816.00 | 19578.00 |
| RTM | TTI RX 2Pass mgpu | NRF | yes | 1.00 | 2.00 | 3.00 | 6.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| SPECFEM3D | four_material_simple_model | Total Time (Sec) | no | 2807.00 | 207.00 | 105.00 | 57.00 |
| SPECFEM3D | four_material_simple_model | NRF | yes | 1.00 | 16.00 | 32.00 | 59.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| Cloverleaf | bm32 | Wall Clock (Sec) | no | 2519.68 | 1002.00 | ||
| Cloverleaf | bm33 | NRF | yes | 1.00 | 3.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| HPCG | 256x256x256 local size | GFLOPS | yes | 25.85 | 117.20 | 230.38 | 422.40 |
| HPCG | 256x256x256 local size | NRF | yes | 1.00 | 4.00 | 9.00 | 16.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| MiniFE | 350x350x350 | Total CG Time (Sec) | no | 21.35 | 7.18 | 3.62 | 1.94 |
| MiniFE | 350x350x351 | NRF | yes | 1.00 | 3.00 | 6.00 | 11.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| Relion | MB numbers Plasmodium Ribosime on Relion-2.1 | 1/Minutes | yes | 0.00146 | 0.01 | 0.02 | |
| Relion | MB numbers Plasmodium Ribosime on Relion-2.2 | NRF | yes | 1.00 | 9.00 | 11.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| AMBER | DHFR (NVE) (AKA JAC) | ns/day | yes | 103.00 | 1094.00 | 2188.00 | 4376.00 |
| AMBER | DHFR (NVE) (AKA JAC) | NRF | yes | 1.00 | 11.00 | 21.00 | 42.00 |
| AMBER | Factor IX (NPT) | ns/day | yes | 25.00 | 302.00 | 604.00 | 1208.00 |
| AMBER | Factor IX (NPT) | NRF | yes | 1.00 | 12.00 | 24.00 | 48.00 |
| AMBER | PME-Cellulose_NVE | ns/day | yes | 5.20 | 66.00 | 132.00 | 264.00 |
| AMBER | PME-Cellulose_NVE | NRF | yes | 1.00 | 13.00 | 25.00 | 51.00 |
| AMBER | STMV (NPT) | ns/day | yes | 1.80 | 22.00 | 44.00 | 88.00 |
| AMBER | STMV (NPT) | NRF | yes | 1.00 | 12.00 | 24.00 | 49.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| GROMACS | ADH Dodec | ns/day | yes | 54.00 | 129.00 | 153.00 | |
| GROMACS | ADH Dodec | NRF | yes | 1.00 | 3.00 | 5.00 | |
| GROMACS | Cellulose | ns/day | yes | 15.00 | 34.00 | 42.00 | 49.00 |
| GROMACS | Cellulose | NRF | yes | 1.00 | 2.00 | 3.00 | 5.00 |
| GROMACS | STMV | ns/day | yes | 3.50 | 9.30 | ||
| GROMACS | STMV | NRF | yes | 1.00 | 3.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| HOOMD-Blue | microsphere | Ave. TPS | yes | 11.89 | 43.00 | 88.90 | 113.00 |
| HOOMD-Blue | microsphere | NRF | yes | 1.00 | 4.00 | 8.00 | 11.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| LAMMPS | EAM | ATOM-Time Steps/s | yes | 51790411 | 88834048.00 | ||
| LAMMPS | EAM | NRF | yes | 1.00 | 2.00 | ||
| LAMMPS | LJ 2.6 | ATOM-Time Steps/s | yes | 96625195 | 162414592.00 | 318832640.00 | |
| LAMMPS | LJ 2.6 | NRF | yes | 1.00 | 2.00 | 3.00 | |
| LAMMPS | ReaxFF/C | ATOM-Time Steps/s | yes | 332697 | 904471.00 | ||
| LAMMPS | ReaxFF/C | NRF | yes | 1.00 | 4.00 | ||
| LAMMPS | Tersoff | ATOM-Time Steps/s | yes | 39247062 | 76619776.00 | 147668992.00 | |
| LAMMPS | Tersoff | NRF | yes | 1.00 | 2.00 | 4.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| NAMD | apoa1_npt_cuda | Ave ns/day | yes | 4.10 | 49.00 | 66.00 | |
| NAMD | apoa1_npt_cuda | NRF | yes | 1.00 | 85.00 | 115.00 | |
| NAMD | apoa1_nptsr_cuda | Ave ns/day | yes | 4.10 | 50.00 | 71.00 | |
| NAMD | apoa1_nptsr_cuda | NRF | yes | 1.00 | 87.00 | 123.00 | |
| NAMD | apoa1_nve_cuda | Ave ns/day | yes | 4.40 | 54.00 | 76.00 | |
| NAMD | apoa1_nve_cuda | NRF | yes | 1.00 | 63.00 | 88.00 | |
| NAMD | stmv_npt_cuda | Ave ns/day | yes | 0.38 | 4.40 | 6.90 | |
| NAMD | stmv_npt_cuda | NRF | yes | 1.00 | 20.00 | 31.00 | |
| NAMD | stmv_nptsr_cuda | Ave ns/day | yes | 0.38 | 4.50 | 7.10 | |
| NAMD | stmv_nptsr_cuda | NRF | yes | 1.00 | 21.00 | 33.00 | |
| NAMD | stmv_nve_cuda | Ave ns/day | yes | 0.38 | 4.70 | 7.70 | |
| NAMD | stmv_nve_cuda | NRF | yes | 1.00 | 19.00 | 31.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| Chroma | szscl21_24_128 | Total Time (Sec) | no | 1140.00 | 101.00 | 39.00 | 26.00 |
| Chroma | szscl21_24_129 | NRF | yes | 1.00 | 21.00 | 53.00 | 80.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| GTC | moi#proc.in | Mpush/Sec | yes | 35.00 | 264.00 | 523.00 | 928.00 |
| GTC | moi#proc.in | NRF | yes | 1.00 | 8.00 | 15.00 | 27.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| MILC | Apex Medium | Total Time (Sec) | no | 72194.00 | 7978.94 | 3835.19 | 2474.58 |
| MILC | Apex Medium | NRF | yes | 1.00 | 10.00 | 22.00 | 34.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| QUDA | QPhil Dslash Wilson-Clover Precision: Single; Gau. Cmprsn/Recon: 12; Probl. Size 32x32x32x64 |
Dslash GFLOPS | yes | 111.60 | 1984.16 | 2319.25 | 3626.52 |
| QUDA | QPhil Dslash Wilson-Clover Precision: Single; Gau. Cmprsn/Recon: 12; Probl. Size 32x32x32x64 |
NRF | yes | 1.00 | 23.00 | 27.00 | 43.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| VASP | B.hR105 | Elapsed Time (Sec) | no | 661.00 | 515.00 | 288.00 | 182.00 |
| VASP | B.hR106 | NRF | yes | 1.00 | 2.00 | 10.00 | 15.00 |
| VASP | Si-Huge | Elapsed Time (Sec) | no | 4364.00 | 2635.00 | ||
| VASP | Si-Huge | NRF | yes | 1.00 | 2.00 |
| Application | Test Modules | Metric | Bigger is better | Dual Skylake | 2x T4 PCIe | 4x T4 PCIe | 8x T4 PCIe |
|---|---|---|---|---|---|---|---|
| WRF | Conus_2.5k_JA | Seconds / Timestamps | no | 5.20 | 3.60 | 1.80 | 1.00 |
| WRF | Conus_2.5k_JA | NRF | yes | 1.00 | 1.00 | 3.00 | 6.00 |