Chip maker Nvidia has unveiled a new dual-GPU VGX K2 and quad-GPU VGX K1 graphics card specifically designed to overcome the problems associated with rendering desktops in the cloud.
Based on the company's Kepler 28-nanometer architecture found in 600-series GeForce cards, workstations, and in the cloud, the new VGX cards leverage the VGX platform that was outlined earlier this year. This combination of hardware and software will allow enterprises to provide seamless remote computing and allow the most demanding applications to be streamed to a notebook or mobile device.
The VGX K2 features 3,072 CUDA cores, 8GB of GDDR5 RAM, and has a TDP of 225W, which, while sounding a lot, is about 75W lower than Nvidia's flagship gaming card, the GTX 690. A single VGX K2 is capable of rendering up to 64 single display virtual desktops.
The VGX K1 sees four as yet unnamed "entry" Kepler GPU cores packed onto a single card. The 768 CUDA cores and 16GB of DDR3 memory can power up to 100 virtual desktops with a maximum TDP of only 130W.
|VGX K1||VGX K2|
|GPUs||Quad "entry" Kepler||Dual "high-end" Kepler|
|RAM||16GB DDR3||8GB GDDR5|
Both cards feature a high-performance H.264 encoding, enter-class power efficiency, including the revolutionary new streaming multiprocessor, called SMX, and are designed for 24/7 operation.
The VGX K2 support Citrix's HDX 3D Pro on the Xen Desktop 5.6 FP1 and Xen Server 6, and both the VGX K1 and VGX K2 support VMWare and Microsoft's RemoteFX RDP protocol, and Windows Server 2012.
Both cards will be available via OEMs such as Dell, HP and IBM early 2013.
Image source: Nvidia.