Maximizando Ciência por Joule com Precisão Numérica e DVFS em GPUs
Resumo
A eficiência energética em computação científica deve considerar não apenas a potência consumida, mas também a quantidade de trabalho útil concluído por unidade de energia. Neste trabalho, investigamos como a precisão numérica afeta essa relação na dinâmica molecular acelerada por GPU. Avaliamos três cargas do LAMMPS, com diferentes participações de cálculos de longo alcance, em três GPUs NVIDIA Ada Lovelace, considerando FP64, FP32/64 e FP32, além de diferentes frequências de operação. Os resultados mostram que o benefício da precisão reduzida depende fortemente da carga e da plataforma: em FP32, a redução de EDP varia de 8,1% a 88,6%. Quando o maior throughput também reduz o tempo de execução, o ganho energético é substancial, alcançando até 3,76× mais atom-steps por joule. Em cargas pouco sensíveis à precisão, esse benefício permanece próximo de 1,1×. Mostramos ainda que FP32/64 preserva grande parte dos ganhos e que a frequência que minimiza o EDP varia conforme o modo numérico, indicando que maximizar a eficiência por joule requer considerar conjuntamente a precisão, as características da carga e a frequência de operação.Referências
Abraham, M. J., Murtola, T., Schulz, R., Páll, S., Smith, J. C., Hess, B., and Lindahl, E. (2015). GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX, 1–2:19–25.
Afzal, A., Hager, G., and Wellein, G. (2026a). Wattlytics: A web platform for co-optimizing performance, energy, and TCO in HPC clusters. Preprint, arXiv:2604.08182 [cs.DC].
Afzal, A., Kahler, A., Hager, G., and Wellein, G. (2026b). GROMACS unplugged: How power capping and frequency shapes performance on GPUs. In Euro-Par 2025: Parallel Processing Workshops, volume 16294 of Lecture Notes in Computer Science, Cham. Springer. No prelo. Preprint: arXiv:2510.06902.
Colberg, P. H. and Höfling, F. (2011). Highly accelerated simulations of glassy dynamics using GPUs: Caveats on limited floating-point precision. Computer Physics Communications, 182:1120–1129.
Essmann, U., Perera, L., Berkowitz, M. L., Darden, T., Lee, H., and Pedersen, L. G. (1995). A smooth particle mesh Ewald method. The Journal of Chemical Physics, 103(19):8577–8593.
Fox, J. and Weisberg, S. (2019). An R Companion to Applied Regression. SAGE Publications, Thousand Oaks, CA, 3 edition.
Godoy, W. F., Hernandez, O., Kent, P. R. C., Patrou, M., Asifuzzaman, K., Miniskar, N. R., Valero-Lara, P., Vetter, J. S., Sinclair, M. D., Lowe-Power, J., and Bruce, B. R. (2026). Characterizing GPU energy usage in exascale-ready portable science applications. In High Performance Computing. ISC High Performance 2025 International Workshops, Revised Selected Papers, volume 16091 of Lecture Notes in Computer Science, pages 177–190, Cham. Springer.
Hockney, R. W. and Eastwood, J. W. (1988). Computer Simulation Using Particles. Adam Hilger, Bristol.
Hoefler, T. and Belli, R. (2015). Scientific benchmarking of parallel computing systems: twelve ways to tell the masses when reporting performance results. In Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, SC ’15, New York, NY, USA. Association for Computing Machinery.
Johansson, A., Weinberg, E., Trott, C. R., McCarthy, M. J., and Moore, S. G. (2025). LAMMPS-Kokkos: Performance portable molecular dynamics across exascale architectures. In Proceedings of the SC ’25 Workshops of the International Conference for High Performance Computing, Networking, Storage and Analysis, pages 1217–1232, New York, NY, USA. Association for Computing Machinery.
Kashi, A., Lu, H., Brewer, W., Rogers, D., Matheson, M., Shankar, M., and Wang, F. (2026). Mixed-precision numerics in scientific applications: survey and perspectives. The Journal of Supercomputing, 82(5):287.
Le Grand, S., Götz, A. W., and Walker, R. C. (2013). SPFP: Speed without compromise— a mixed precision model for GPU accelerated molecular dynamics simulations. Computer Physics Communications, 184(2):374–380.
Machado, R. R. L., Eitzinger, J., Hager, G., and Wellein, G. (2026). On the challenges of energy-efficiency analysis in hpc systems: Evaluating synthetic benchmarks and gromacs. In Proceedings of the Supercomputing Asia and International Conference on High Performance Computing in Asia Pacific Region Workshops, SCA/HPCAsiaWS ’26, page 50–58, New York, NY, USA. Association for Computing Machinery.
Mei, X., Wang, Q., and Chu, X. (2017). A survey and measurement study of GPU DVFS on energy conservation. Digital Communications and Networks, 3(2):89–100.
NVIDIA Corporation (2023). NVIDIA Ada GPU architecture. Whitepaper, v2.02. Disponível em: [link].
Thompson, A. P., Aktulga, H. M., Berger, R., Bolintineanu, D. S., Brown, W. M., Crozier, P. S., in ’t Veld, P. J., Kohlmeyer, A., Moore, S. G., Nguyen, T. D., Shan, R., Stevens, M. J., Tranchida, J., Trott, C., and Plimpton, S. J. (2022). LAMMPS—a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales. Computer Physics Communications, 271:108171.
Trott, C. R., Winterfeld, L., and Crozier, P. S. (2010). General-purpose molecular dynamics simulations on GPU-based clusters. arXiv:1009.4330.
Verlet, L. (1967). Computer “experiments” on classical fluids. I. thermodynamical properties of Lennard-Jones molecules. Physical Review, 159(1):98–103.
Yang, Z., Adámek, K., and Armour, W. (2024). Accurate and convenient energy measurements for gpus: A detailed study of nvidia gpu’s built-in power sensor. In SC24: International Conference for High Performance Computing, Networking, Storage and Analysis, pages 1–17.
Afzal, A., Hager, G., and Wellein, G. (2026a). Wattlytics: A web platform for co-optimizing performance, energy, and TCO in HPC clusters. Preprint, arXiv:2604.08182 [cs.DC].
Afzal, A., Kahler, A., Hager, G., and Wellein, G. (2026b). GROMACS unplugged: How power capping and frequency shapes performance on GPUs. In Euro-Par 2025: Parallel Processing Workshops, volume 16294 of Lecture Notes in Computer Science, Cham. Springer. No prelo. Preprint: arXiv:2510.06902.
Colberg, P. H. and Höfling, F. (2011). Highly accelerated simulations of glassy dynamics using GPUs: Caveats on limited floating-point precision. Computer Physics Communications, 182:1120–1129.
Essmann, U., Perera, L., Berkowitz, M. L., Darden, T., Lee, H., and Pedersen, L. G. (1995). A smooth particle mesh Ewald method. The Journal of Chemical Physics, 103(19):8577–8593.
Fox, J. and Weisberg, S. (2019). An R Companion to Applied Regression. SAGE Publications, Thousand Oaks, CA, 3 edition.
Godoy, W. F., Hernandez, O., Kent, P. R. C., Patrou, M., Asifuzzaman, K., Miniskar, N. R., Valero-Lara, P., Vetter, J. S., Sinclair, M. D., Lowe-Power, J., and Bruce, B. R. (2026). Characterizing GPU energy usage in exascale-ready portable science applications. In High Performance Computing. ISC High Performance 2025 International Workshops, Revised Selected Papers, volume 16091 of Lecture Notes in Computer Science, pages 177–190, Cham. Springer.
Hockney, R. W. and Eastwood, J. W. (1988). Computer Simulation Using Particles. Adam Hilger, Bristol.
Hoefler, T. and Belli, R. (2015). Scientific benchmarking of parallel computing systems: twelve ways to tell the masses when reporting performance results. In Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, SC ’15, New York, NY, USA. Association for Computing Machinery.
Johansson, A., Weinberg, E., Trott, C. R., McCarthy, M. J., and Moore, S. G. (2025). LAMMPS-Kokkos: Performance portable molecular dynamics across exascale architectures. In Proceedings of the SC ’25 Workshops of the International Conference for High Performance Computing, Networking, Storage and Analysis, pages 1217–1232, New York, NY, USA. Association for Computing Machinery.
Kashi, A., Lu, H., Brewer, W., Rogers, D., Matheson, M., Shankar, M., and Wang, F. (2026). Mixed-precision numerics in scientific applications: survey and perspectives. The Journal of Supercomputing, 82(5):287.
Le Grand, S., Götz, A. W., and Walker, R. C. (2013). SPFP: Speed without compromise— a mixed precision model for GPU accelerated molecular dynamics simulations. Computer Physics Communications, 184(2):374–380.
Machado, R. R. L., Eitzinger, J., Hager, G., and Wellein, G. (2026). On the challenges of energy-efficiency analysis in hpc systems: Evaluating synthetic benchmarks and gromacs. In Proceedings of the Supercomputing Asia and International Conference on High Performance Computing in Asia Pacific Region Workshops, SCA/HPCAsiaWS ’26, page 50–58, New York, NY, USA. Association for Computing Machinery.
Mei, X., Wang, Q., and Chu, X. (2017). A survey and measurement study of GPU DVFS on energy conservation. Digital Communications and Networks, 3(2):89–100.
NVIDIA Corporation (2023). NVIDIA Ada GPU architecture. Whitepaper, v2.02. Disponível em: [link].
Thompson, A. P., Aktulga, H. M., Berger, R., Bolintineanu, D. S., Brown, W. M., Crozier, P. S., in ’t Veld, P. J., Kohlmeyer, A., Moore, S. G., Nguyen, T. D., Shan, R., Stevens, M. J., Tranchida, J., Trott, C., and Plimpton, S. J. (2022). LAMMPS—a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales. Computer Physics Communications, 271:108171.
Trott, C. R., Winterfeld, L., and Crozier, P. S. (2010). General-purpose molecular dynamics simulations on GPU-based clusters. arXiv:1009.4330.
Verlet, L. (1967). Computer “experiments” on classical fluids. I. thermodynamical properties of Lennard-Jones molecules. Physical Review, 159(1):98–103.
Yang, Z., Adámek, K., and Armour, W. (2024). Accurate and convenient energy measurements for gpus: A detailed study of nvidia gpu’s built-in power sensor. In SC24: International Conference for High Performance Computing, Networking, Storage and Analysis, pages 1–17.
Publicado
03/11/2026
Como Citar
UEBEL, Átila; COSTA, Matheus; LORENZON, Arthur F..
Maximizando Ciência por Joule com Precisão Numérica e DVFS em GPUs. In: SIMPÓSIO EM SISTEMAS COMPUTACIONAIS DE ALTO DESEMPENHO (SSCAD), 27. , 2026, Natal/RN.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 178-191.
DOI: https://doi.org/10.5753/sscad.2026.34903.
