NVidia finally announced the first of their professional lineup of products, based on their next generation Kepler processor. The Quadro K5000 is a high-end professional graphics card, while they also announced thr Telsa K10 and K20, which are dedicated GPU processing cards. They were announced earlier this month during SIGGRAPH, but won’t be released for another month or two.
Besides having a terrible identifier in K5000 for product differentiation, the new Quadro looks to offer some useful new features. It supports PCIe 3.0 for faster transfer of data between the system and the GPU, which is becoming more important with CUDA accelerated processing needing data to be sent back into the system instead of to the display output. Speaking of display outputs, the new Kepler products support 4 separate monitors from a single card. This will be advantageous for those planning to use it to replace a dedicated video I/O card for fullscreen monitoring, but still wanting two UI displays. This was the one feature that AMD’s cards had over the Quadros up to this point. The K5000 will also be able to output 4K over a single DisplayPort cable, which will become more relevent once 4K displays become more widely available.
On the Tesla front, the K10 is basically a dual chip version of the same processor that is in the K5000, while the K20 will be based on a whole new GPU design. Both of these designs are departures from the previous Telsa products, which were basically Quadro cards with the display outputs removed. The new K10 appears to be about four times as powerful as the existing C2075 cards, at least on paper. They also announced the next generation of their Maximus program, marketing Telsa cards in parallel with Quadro cards, but we will have to wait for software developer to really leverage those capabilities.
Hopefully I will be able to offer a more in-depth look at the Quadro K5000, and its processing performance, once it is released to the public.