AMD and Nvidia have made fresh inroads into the supercomputing market, each releasing new graphical processing unit (GPU) chips designed for intensive mathematical and scientific computation.
[Image is of a previous generation Nvidia Tesla GPU]
GPUs are increasingly used for high-performance computing applications as they’re optimized for parallel processing (some chips are composed of thousands of cores working in tandem).
Nvidia has developed two new “Tesla” GPUs, the K20 and K20X; the former offers 1.17 teraflops of double-precision performance while the latter has 6GB of memory and provides 1.31 teraflops of double-precision performance, according to this ComputerWorld article.
Higher precision in the context of supercomputing is an important factor in performance as it allows mathematical calculations to be translated into binary format more efficiently.
AMD, meanwhile, released the FirePro SM10000, which has 6GB of memory and comes it at 1.48 teraflops of peak double-precision performance.
As the article notes, AMD is a newcomer to the high-performance computing market and faces stiff competition from Nvidia.