DGX-A100 Face to Face DGX-2—Performance, Power and Thermal Behavior Evaluation
Nvidia is a leading producer of GPUs for high-performance computing and artificial intelligence, bringing top performance and energy-efficiency. We present performance, power consumption, and thermal behavior analysis of the new Nvidia DGX-A100 server equipped with eight A100 Ampere microarchitectur...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-01-01
|
| Series: | Energies |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1996-1073/14/2/376 |
| _version_ | 1797412840306376704 |
|---|---|
| author | Matej Špeťko Ondřej Vysocký Branislav Jansík Lubomír Říha |
| author_facet | Matej Špeťko Ondřej Vysocký Branislav Jansík Lubomír Říha |
| author_sort | Matej Špeťko |
| collection | DOAJ |
| description | Nvidia is a leading producer of GPUs for high-performance computing and artificial intelligence, bringing top performance and energy-efficiency. We present performance, power consumption, and thermal behavior analysis of the new Nvidia DGX-A100 server equipped with eight A100 Ampere microarchitecture GPUs. The results are compared against the previous generation of the server, Nvidia DGX-2, based on Tesla V100 GPUs. We developed a synthetic benchmark to measure the raw performance of floating-point computing units including Tensor Cores. Furthermore, thermal stability was investigated. In addition, Dynamic Frequency and Voltage Scaling (DVFS) analysis was performed to determine the best energy-efficient configuration of the GPUs executing workloads of various arithmetical intensities. Under the energy-optimal configuration the A100 GPU reaches efficiency of 51 GFLOPS/W for double-precision workload and 91 GFLOPS/W for tensor core double precision workload, which makes the A100 the most energy-efficient server accelerator for scientific simulations in the market. |
| first_indexed | 2024-03-09T05:09:02Z |
| format | Article |
| id | doaj.art-62f4250600c34107a10a2a13d4d746d7 |
| institution | Directory Open Access Journal |
| issn | 1996-1073 |
| language | English |
| last_indexed | 2024-03-09T05:09:02Z |
| publishDate | 2021-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Energies |
| spelling | doaj.art-62f4250600c34107a10a2a13d4d746d72023-12-03T12:51:12ZengMDPI AGEnergies1996-10732021-01-0114237610.3390/en14020376DGX-A100 Face to Face DGX-2—Performance, Power and Thermal Behavior EvaluationMatej Špeťko0Ondřej Vysocký1Branislav Jansík2Lubomír Říha3IT4Innovations National Supercomputing Center, VŠB—Technical University of Ostrava, 708 00 Ostrava, Czech RepublicIT4Innovations National Supercomputing Center, VŠB—Technical University of Ostrava, 708 00 Ostrava, Czech RepublicIT4Innovations National Supercomputing Center, VŠB—Technical University of Ostrava, 708 00 Ostrava, Czech RepublicIT4Innovations National Supercomputing Center, VŠB—Technical University of Ostrava, 708 00 Ostrava, Czech RepublicNvidia is a leading producer of GPUs for high-performance computing and artificial intelligence, bringing top performance and energy-efficiency. We present performance, power consumption, and thermal behavior analysis of the new Nvidia DGX-A100 server equipped with eight A100 Ampere microarchitecture GPUs. The results are compared against the previous generation of the server, Nvidia DGX-2, based on Tesla V100 GPUs. We developed a synthetic benchmark to measure the raw performance of floating-point computing units including Tensor Cores. Furthermore, thermal stability was investigated. In addition, Dynamic Frequency and Voltage Scaling (DVFS) analysis was performed to determine the best energy-efficient configuration of the GPUs executing workloads of various arithmetical intensities. Under the energy-optimal configuration the A100 GPU reaches efficiency of 51 GFLOPS/W for double-precision workload and 91 GFLOPS/W for tensor core double precision workload, which makes the A100 the most energy-efficient server accelerator for scientific simulations in the market.https://www.mdpi.com/1996-1073/14/2/376DGX-A100DGX-2tensor coresperformance analysisenergy efficient computingDVFS |
| spellingShingle | Matej Špeťko Ondřej Vysocký Branislav Jansík Lubomír Říha DGX-A100 Face to Face DGX-2—Performance, Power and Thermal Behavior Evaluation Energies DGX-A100 DGX-2 tensor cores performance analysis energy efficient computing DVFS |
| title | DGX-A100 Face to Face DGX-2—Performance, Power and Thermal Behavior Evaluation |
| title_full | DGX-A100 Face to Face DGX-2—Performance, Power and Thermal Behavior Evaluation |
| title_fullStr | DGX-A100 Face to Face DGX-2—Performance, Power and Thermal Behavior Evaluation |
| title_full_unstemmed | DGX-A100 Face to Face DGX-2—Performance, Power and Thermal Behavior Evaluation |
| title_short | DGX-A100 Face to Face DGX-2—Performance, Power and Thermal Behavior Evaluation |
| title_sort | dgx a100 face to face dgx 2 performance power and thermal behavior evaluation |
| topic | DGX-A100 DGX-2 tensor cores performance analysis energy efficient computing DVFS |
| url | https://www.mdpi.com/1996-1073/14/2/376 |
| work_keys_str_mv | AT matejspetko dgxa100facetofacedgx2performancepowerandthermalbehaviorevaluation AT ondrejvysocky dgxa100facetofacedgx2performancepowerandthermalbehaviorevaluation AT branislavjansik dgxa100facetofacedgx2performancepowerandthermalbehaviorevaluation AT lubomirriha dgxa100facetofacedgx2performancepowerandthermalbehaviorevaluation |