{ "created_utc": "2026-10-30T09:46:50.155301+00:00", "source_data_sha256": "b46aa6cc4a3928a13c8807cffff48f7c38927d382fabff8d99318b468a3a40d3", "hardware": { "a100": { "CPU": "Intel Xeon Gold 6430 x2", "physical_cores_used_CPU_only": 64, "GPU": "NVIDIA A100 80GB PCIe x4", "driver": "595.99.02", "power_limit_W_per_GPU": 300, "topology": "GPU0-1 and GPU2-3: NV12; between pairs: SYS" }, "h100": { "CPU": "AMD EPYC 9354P x1", "physical_cores_used_CPU_only": 32, "GPU": "NVIDIA H100 PCIe 80GB x2", "driver": "595.91.07", "power_limit_W_per_GPU": 350, "topology": "GPU0-1: PHB; no NVLink edge reported" }, "b200": { "CPU": "AMD EPYC 9575F x2", "physical_cores_used_CPU_only": 128, "GPU": "NVIDIA B200 x8 (HGX)", "driver": "595.91.07", "power_limit_W_per_GPU": 1000, "topology": "All GPU pairs: NV18" } }, "software_sources_and_input_provenance": { "LAMMPS": { "version": "30Sep2026", "url": "https://github.com/lammps/lammps/releases/download/stable_30Sep2026/lammps-src-30Sep2026.tar.gz", "archive_sha256": "a3daf373ff07b8dd98773f2eed2707d0bc5dac96ee2a2599b0d2d98d91c9d20a", "software_license": "GPL v2", "license_url": "https://docs.lammps.org/Intro_opensource.html", "input": "Generated Cu EAM system; Cu_u3.eam supplied in LAMMPS distribution" }, "GROMACS": { "version": "2026.4", "url": "https://ftp.gromacs.org/gromacs/gromacs-2026.4.tar.gz", "archive_sha256": "58eda60979b124fcf9dff8b1b63cedfbc72fdb36d99e921dc1bdc310571098b6", "software_license": "LGPL 2.1 or later (runtime notice)", "license_url": "https://raw.githubusercontent.com/gromacs/gromacs/v2026.4/COPYING", "input": "benchPEP-h, Max Planck Institute benchmark set", "input_url": "https://www.mpinat.mpg.de/benchPEP-h", "input_information_url": "https://www.mpinat.mpg.de/grubmueller/bench", "input_license_note": "Provider describes freely available benchmark inputs; software license is not asserted as the input license. Refer to provider terms; input binary is not redistributed here." }, "Quantum ESPRESSO": { "version": "7.6", "url": "https://github.com/QEF/q-e", "commit": "9f93ddec427d2b9a45bb72d828c6d324f62fcabd", "software_license": "GPL v2", "license_url": "https://raw.githubusercontent.com/QEF/q-e/9f93ddec427d2b9a45bb72d828c6d324f62fcabd/License", "input": "QEF benchmarks PSIWAT", "input_commit": "eff15189e7b1120f5facc6fef202cbd6ff0ed26b", "input_url": "https://github.com/QEF/benchmarks/tree/eff15189e7b1120f5facc6fef202cbd6ff0ed26b/PSIWAT", "input_license": "MIT, repository-level notice", "input_license_url": "https://github.com/QEF/benchmarks/blob/eff15189e7b1120f5facc6fef202cbd6ff0ed26b/LICENSE" } }, "build_and_runtime": { "a100": { "gromacs_version": "2026.4", "GROMACS_gpu_C_compiler": "GNU 13.3.0", "GROMACS_gpu_CXX_compiler": "GNU 13.3.0", "GROMACS_gpu_C_flags": "-fexcess-precision=fast -funroll-all-loops -march=skylake-avx512 -Wno-missing-field-initializers -pthread -O3 -DNDEBUG", "GROMACS_gpu_CXX_flags": "-fexcess-precision=fast -funroll-all-loops -march=skylake-avx512 -Wno-missing-field-initializers -Wno-stringop-truncation -pthread -Wno-old-style-cast -Wno-cast-qual -Wno-suggest-override -Wno-suggest-destructor-override -Wno-zero-as-null-pointer-constant -Wno-cast-function-type-strict SHELL:-fopenmp -O3 -DNDEBUG", "GROMACS_cuda_compiler": "13.2.78", "GROMACS_cuda_flags": "-D_FORCE_INLINES -O3 -DNDEBUG -use_fast_math -static-global-template-stub=false -Xptxas=-warn-double-usage -Xptxas=-Werror -diag-suppress=177 -O3 -DNDEBUG", "GROMACS_cuda_targets": "80", "GROMACS_SIMD": "AVX_512", "GROMACS_CPU_FFT": "fftw-3.3.10-sse2-avx", "GROMACS_MPI_version": "Open MPI v4.1.9a1", "GROMACS_MPI_revision": "v4.1.5-258-gea57aceaf7", "GROMACS_MPI_release_status": "Unreleased developer copy as reported by binary", "HPC_SDK": "26.5", "HPC_X": "2.50", "QE_compilers": "nvc / nvfortran, NVIDIA HPC SDK 26.5", "QE_math_libraries": "Intel oneMKL 2026.0, LP64 sequential BLAS and ScaLAPACK", "QE_release_flags": "-fast -O3" }, "h100": { "gromacs_version": "2026.4", "GROMACS_gpu_C_compiler": "GNU 13.3.0", "GROMACS_gpu_CXX_compiler": "GNU 13.3.0", "GROMACS_gpu_C_flags": "-fexcess-precision=fast -funroll-all-loops -mavx2 -mfma -Wno-missing-field-initializers -pthread -O3 -DNDEBUG", "GROMACS_gpu_CXX_flags": "-fexcess-precision=fast -funroll-all-loops -mavx2 -mfma -Wno-missing-field-initializers -Wno-stringop-truncation -pthread -Wno-old-style-cast -Wno-cast-qual -Wno-suggest-override -Wno-suggest-destructor-override -Wno-zero-as-null-pointer-constant -Wno-cast-function-type-strict SHELL:-fopenmp -O3 -DNDEBUG", "GROMACS_cuda_compiler": "13.2.78", "GROMACS_cuda_flags": "-D_FORCE_INLINES -O3 -DNDEBUG -use_fast_math -static-global-template-stub=false -Xptxas=-warn-double-usage -Xptxas=-Werror -diag-suppress=177 -O3 -DNDEBUG", "GROMACS_cuda_targets": "90", "GROMACS_SIMD": "AVX2_256", "GROMACS_CPU_FFT": "fftw-3.3.10-sse2-avx", "GROMACS_MPI_version": "Open MPI v4.1.9a1", "GROMACS_MPI_revision": "v4.1.5-258-gea57aceaf7", "GROMACS_MPI_release_status": "Unreleased developer copy as reported by binary", "HPC_SDK": "26.5", "HPC_X": "2.50", "QE_compilers": "nvc / nvfortran, NVIDIA HPC SDK 26.5", "QE_math_libraries": "OpenBLAS 0.3.34 + ScaLAPACK 2.2.3", "QE_release_flags": "-fast -O3" }, "b200": { "gromacs_version": "2026.4", "GROMACS_gpu_C_compiler": "GNU 14.2.0", "GROMACS_gpu_CXX_compiler": "GNU 14.2.0", "GROMACS_gpu_C_flags": "-fexcess-precision=fast -funroll-all-loops -march=skylake-avx512 -Wno-missing-field-initializers -pthread -O3 -DNDEBUG", "GROMACS_gpu_CXX_flags": "-fexcess-precision=fast -funroll-all-loops -march=skylake-avx512 -Wno-missing-field-initializers -Wno-stringop-truncation -pthread -Wno-old-style-cast -Wno-cast-qual -Wno-suggest-override -Wno-suggest-destructor-override -Wno-zero-as-null-pointer-constant -Wno-cast-function-type-strict SHELL:-fopenmp -O3 -DNDEBUG", "GROMACS_cuda_compiler": "13.1.80", "GROMACS_cuda_flags": "-D_FORCE_INLINES -O3 -DNDEBUG -use_fast_math -static-global-template-stub=false -Xptxas=-warn-double-usage -Xptxas=-Werror -diag-suppress=177 -O3 -DNDEBUG", "GROMACS_cuda_targets": "100", "GROMACS_SIMD": "AVX_512", "GROMACS_CPU_FFT": "fftw-3.3.10-sse2-avx", "GROMACS_MPI_version": "Open MPI v4.1.9a1", "GROMACS_MPI_revision": "v4.1.5-232-gad48c462ff", "GROMACS_MPI_release_status": "Unreleased developer copy as reported by binary", "HPC_SDK": "26.1", "HPC_X": "2.25.1", "QE_compilers": "nvc / nvfortran, NVIDIA HPC SDK 26.1", "QE_math_libraries": "OpenBLAS 0.3.34 + ScaLAPACK 2.2.3", "QE_release_flags": "-fast -O3" } }, "cases": { "LAMMPS": { "name": "Cu EAM", "atoms": 2048000, "warmup_steps": 1000, "timed_steps": 30000, "timestep_ps": 0.005, "neighbor": "every 1 delay 0 check yes", "precision": "FP64", "input_sha256": "f2564466d5366e9e7859650496dfc727ecbcc48a9ff669c48d10ecc7d2d31b5a", "potential_sha256": "6d37c8e85214154c9d899fd71a95163baf8cc518c9c256007dbe37f3eab9e3fa" }, "GROMACS": { "name": "benchPEP-h", "atoms": 12495503, "input_sha256": "6d7d38191b649132e905c4827a3ec9a1bce89bd6d792886f8cee4f53ea0c127a", "total_steps": 11000, "counter_reset_at_step": 1000, "nominal_timed_steps": 10000, "timestep_ps": 0.002, "nominal_timed_interval_ps": 20, "repeats_per_setting": 3 }, "Quantum ESPRESSO": { "name": "PSIWAT", "atoms": 586, "input_sha256": "43d70eca9989ac9106d0e8b39f7647bd468728878064d069f095af0a71e28dec", "ecutwfc_Ry": 25, "ecutrho_Ry": 200, "kpoints": 4, "bands": 2040, "precision": "FP64", "initialization": "from_scratch each repeat" } }, "conditions": [ { "application": "LAMMPS", "host": "a100", "gpus": 0, "configuration": { "MPI_ranks": 64, "threads_per_rank": 1, "backend": "cpu" } }, { "application": "LAMMPS", "host": "a100", "gpus": 1, "configuration": { "MPI_ranks": 1, "threads_per_rank": 1, "backend": "kokkos" } }, { "application": "LAMMPS", "host": "a100", "gpus": 2, "configuration": { "MPI_ranks": 2, "threads_per_rank": 1, "backend": "kokkos" } }, { "application": "LAMMPS", "host": "a100", "gpus": 4, "configuration": { "MPI_ranks": 16, "threads_per_rank": 1, "backend": "gpu" } }, { "application": "LAMMPS", "host": "h100", "gpus": 0, "configuration": { "MPI_ranks": 32, "threads_per_rank": 1, "backend": "cpu" } }, { "application": "LAMMPS", "host": "h100", "gpus": 1, "configuration": { "MPI_ranks": 1, "threads_per_rank": 1, "backend": "kokkos-serial" } }, { "application": "LAMMPS", "host": "h100", "gpus": 2, "configuration": { "MPI_ranks": 2, "threads_per_rank": 1, "backend": "kokkos" } }, { "application": "LAMMPS", "host": "b200", "gpus": 0, "configuration": { "MPI_ranks": 128, "threads_per_rank": 1, "backend": "cpu" } }, { "application": "LAMMPS", "host": "b200", "gpus": 1, "configuration": { "MPI_ranks": 1, "threads_per_rank": 1, "backend": "kokkos" } }, { "application": "LAMMPS", "host": "b200", "gpus": 2, "configuration": { "MPI_ranks": 2, "threads_per_rank": 1, "backend": "kokkos" } }, { "application": "LAMMPS", "host": "b200", "gpus": 4, "configuration": { "MPI_ranks": 4, "threads_per_rank": 1, "backend": "kokkos" } }, { "application": "LAMMPS", "host": "b200", "gpus": 8, "configuration": { "MPI_ranks": 8, "threads_per_rank": 1, "backend": "kokkos" } }, { "application": "GROMACS", "host": "a100", "gpus": 0, "configuration": { "precision": "double", "MPI_ranks": 16, "threads_per_rank": 4, "PME_ranks": -1, "MD_update": "cpu", "host_staged_GPU_communication": false } }, { "application": "GROMACS", "host": "a100", "gpus": 1, "configuration": { "precision": "mixed", "MPI_ranks": 1, "threads_per_rank": 8, "PME_ranks": 0, "MD_update": "gpu", "host_staged_GPU_communication": false } }, { "application": "GROMACS", "host": "a100", "gpus": 2, "configuration": { "precision": "mixed", "MPI_ranks": 2, "threads_per_rank": 8, "PME_ranks": 1, "MD_update": "gpu", "host_staged_GPU_communication": false } }, { "application": "GROMACS", "host": "a100", "gpus": 4, "configuration": { "precision": "mixed", "MPI_ranks": 4, "threads_per_rank": 8, "PME_ranks": 1, "MD_update": "cpu", "host_staged_GPU_communication": true } }, { "application": "GROMACS", "host": "h100", "gpus": 0, "configuration": { "precision": "double", "MPI_ranks": 16, "threads_per_rank": 2, "PME_ranks": -1, "MD_update": "cpu", "host_staged_GPU_communication": false } }, { "application": "GROMACS", "host": "h100", "gpus": 1, "configuration": { "precision": "mixed", "MPI_ranks": 1, "threads_per_rank": 8, "PME_ranks": 0, "MD_update": "gpu", "host_staged_GPU_communication": false } }, { "application": "GROMACS", "host": "h100", "gpus": 2, "configuration": { "precision": "mixed", "MPI_ranks": 2, "threads_per_rank": 8, "PME_ranks": 1, "MD_update": "cpu", "host_staged_GPU_communication": true } }, { "application": "GROMACS", "host": "b200", "gpus": 0, "configuration": { "precision": "double", "MPI_ranks": 128, "threads_per_rank": 1, "PME_ranks": -1, "MD_update": "cpu", "host_staged_GPU_communication": false } }, { "application": "GROMACS", "host": "b200", "gpus": 1, "configuration": { "precision": "mixed", "MPI_ranks": 1, "threads_per_rank": 8, "PME_ranks": 0, "MD_update": "gpu", "host_staged_GPU_communication": false } }, { "application": "GROMACS", "host": "b200", "gpus": 2, "configuration": { "precision": "mixed", "MPI_ranks": 2, "threads_per_rank": 8, "PME_ranks": 1, "MD_update": "gpu", "host_staged_GPU_communication": false } }, { "application": "GROMACS", "host": "b200", "gpus": 4, "configuration": { "precision": "mixed", "MPI_ranks": 4, "threads_per_rank": 4, "PME_ranks": 1, "MD_update": "gpu", "host_staged_GPU_communication": false } }, { "application": "GROMACS", "host": "b200", "gpus": 8, "configuration": { "precision": "mixed", "MPI_ranks": 8, "threads_per_rank": 4, "PME_ranks": 1, "MD_update": "gpu", "host_staged_GPU_communication": false } }, { "application": "Quantum ESPRESSO", "host": "a100", "gpus": 0, "configuration": { "MPI_ranks": 64, "threads_per_rank": 1, "kpoint_pools": 4, "job_local_UCC_disabled": false } }, { "application": "Quantum ESPRESSO", "host": "a100", "gpus": 1, "configuration": { "MPI_ranks": 1, "threads_per_rank": 8, "kpoint_pools": 1, "job_local_UCC_disabled": false } }, { "application": "Quantum ESPRESSO", "host": "a100", "gpus": 2, "configuration": { "MPI_ranks": 2, "threads_per_rank": 1, "kpoint_pools": 2, "job_local_UCC_disabled": false } }, { "application": "Quantum ESPRESSO", "host": "a100", "gpus": 4, "configuration": { "MPI_ranks": 4, "threads_per_rank": 1, "kpoint_pools": 4, "job_local_UCC_disabled": true } }, { "application": "Quantum ESPRESSO", "host": "h100", "gpus": 0, "configuration": { "ranks": 32, "threads_per_rank": 1, "kpoint_pools": 2, "ndiag": 4 } }, { "application": "Quantum ESPRESSO", "host": "h100", "gpus": 1, "configuration": { "MPI_ranks": 1, "threads_per_rank": 8, "kpoint_pools": 1, "GPU_aware_MPI_build": true, "binding": "core, slot:PE=8", "OMP_PROC_BIND": "close" } }, { "application": "Quantum ESPRESSO", "host": "h100", "gpus": 2, "configuration": { "ranks": 2, "threads_per_rank": 1, "kpoint_pools": 1, "GPU_aware_MPI": false, "UCC_enabled": false, "binary_sha256": "74dab1b18c007a5c62c64e3f0cb9cc8a726fd00293c95b441ea946eb08420894" } }, { "application": "Quantum ESPRESSO", "host": "b200", "gpus": 0, "configuration": { "mpi_ranks": 128, "omp_threads_per_rank": 1, "kpoint_pools": 4 } }, { "application": "Quantum ESPRESSO", "host": "b200", "gpus": 1, "configuration": { "mpi_ranks": 1, "omp_threads_per_rank": 8, "kpoint_pools": 1 } }, { "application": "Quantum ESPRESSO", "host": "b200", "gpus": 2, "configuration": { "mpi_ranks": 2, "omp_threads_per_rank": 1, "kpoint_pools": 2 } }, { "application": "Quantum ESPRESSO", "host": "b200", "gpus": 4, "configuration": { "mpi_ranks": 4, "omp_threads_per_rank": 1, "kpoint_pools": 4 } }, { "application": "Quantum ESPRESSO", "host": "b200", "gpus": 8, "configuration": { "mpi_ranks": 8, "omp_threads_per_rank": 1, "kpoint_pools": 4 } } ], "notes": [ "Power limits are preflight observations, not measured mean power or energy efficiency.", "Physical topology labels do not prove that each application used direct GPU communication.", "Open MPI reports a vendor-bundled developer version; HPC-X package versions differ.", "Compiler flags above are taken from the GROMACS GPU binary; do not assume all applications share them.", "LAMMPS backend and MPI layouts differ across settings; GROMACS CPU double vs GPU mixed precision.", "H100 QE 2GPU uses a separate GPU-aware-MPI-disabled build, unlike 1GPU.", "No private connection details, full environment values, raw inspect, or input binaries are distributed in this JSON." ] }