iSGTW has a great writeup from Jan Zverina on the advantages of CPU’s and GPU’s, correctly seeing that each of them have their own areas of expertise and use and neither of them will die completely. They look at recent advances in the TeraGrid systems and how GPUs are offering huge advances in a few areas. Toward the end they speak with some of the developers of the AMBER computational chemistry code.
“GPUs are, for the first time, giving us the increases in capability we have been desperate for since the beginning of the multicore era,” says Walker. “I’m confident that we will soon be achieving throughput with GPU-enabled AMBER that is at least an order of magnitude better than we could ever hope to achieve with CPU-based clusters.”
This nonsensical GPU/CPU comparisons are fueled by poor understanding of different types of parallelisms; unfortunately, the professionals who really understands it rarely writes the articles. GPU vs. CPU is the same as SIMD vs MIMD. It is better to be healthy and wealthy then poor and sick – MIMD is more complicated device and SIMD is just a compromise to get a top performance for specific data-parallel tasks while MIMD handles equally well both – tasks and data parallelisms.
GPU vs. CPU is the same as SIMD vs MIMD. It is better to be healthy and wealthy then poor and sick – MIMD is more complicated device and SIMD is just a compromise to get a top performance for specific data-parallel tasks while MIMD handles equally well both – tasks and data parallelisms. This nonsensical GPU/CPU debates are fueled by poor understanding of different types of parallelisms.