Journal of Parallel and Distributed Computing
| Hi-index | 0.00 |
Multi-core machines are said to provide high performance at low-energy cost via multi-threading. However, obtaining high performance at low-energy cost depends on the optimal usage of the platform. The experimental results show that executing a multi-threaded program with an optimal configuration can lead to significant gains: up to 38.1% reduction in power and 49.72% reduction in total energy. The paper reports on the importance of isolating the energy used by the idle system and considering just the energy used by the multi-threaded program to be able to evaluate the performance of multi-threading accurately. The experimental results also show that the concept that “power consumption is proportional to the cube of the frequency” is not general and does not apply to the scenario investigated. In the experiments presented, the power consumption was rather linear with Intel's mainstream Core Duo Quad microprocessor.