high performance computing on graphics processing units: hgpu.org

The change in processor architectures and 3D benchmarks makes performance characterization important for every processor and 3D application generation. Recent 3D applications require large amount of data to be processed by the GPU and the CPU. This leads to the importance in analyzing processor performance for different architectures and benchmarks so that benchmarks and processors are fine-tuned to achieve better performance. In this paper, we propose a performance analytical model for the CPU and the GPU. The analytical model takes different architecture as input parameters to estimate performance. We used two different 3D applications, Crysis game and 3DMarkVantage. We characterize both benchmarks for performance bottlenecks and propose an analytical model to estimate performance at different configurations with error deviation <5%.