Nvidia has announced two new Kepler-based processing units for professionals that need to leverage high performance computing in their research — the Telsa K10 and K20 graphics cards. The K10, based on the GK104 core currently used in Nvidia's high end desktop cards like the GTX 680 and 690, will be available immediately. The K20, based on a new GK110 core, will be available in Q4 of this year. Both cards are part of Nvidia's CUDA 5, a programming platform for parallelizing intensive data processing tasks that's slated for release to developers in Q3
These new cards feature improved "streaming multiprocessors" — the micro-cores within the GPU — with better energy efficiency, as well as "dynamic parallelism" which allows for the GPU to route data in the most efficient way possible. The Tesla's Hyper-Q feature allows the cards to work in unison with multiple CPU cores, keeping system bottlenecks to a minimum. Nvidia says both of these cards focus on maximizing performance with smart resource allocation while providing the best energy efficiency possible, which means the technical computing pros using these cards will be able to crunch more numbers with less power.