NVIDIA Grid available on Amazon Web Services

Nvidia Grid gets real

By Kathleen Maher at GfxSpeak

Although Nvidia’s Grid was an­nounced over a year ago at GTC 2012 and has seen some customer adoption and important partner­ships with HP and Dell, among oth­ers, Nvidia put more meat on the bone with the Kepler rolled out last spring.

Kepler is Nvidia’s first chip that has integrated support for remote computing.

At Nvidia’s GTC devel­oper conference this year, the company in­troduced the Visual Computing Appliance (VCA), a compact, low-cost 2U appliance that enables small companies to take advantage of Nvid­ia’s Grid technology. And Amazon has put the icing on the cake by providing access to Nvidia’s GPUs via its EC2, Elastic Com­pute Cloud web ser­vice available through Amazon Web Services (AWS).

So far, GPU re­sources are available in two instance types, G2 and CG1; both include Intel Xeon processors and Nvidia GPUs, which can be used for graphics applications or GPGPU com­pute applications. The G2 includes ac­cess to Nvidia Grid GPUs via the Kepler GK104 including 4 GB of video memory and the GPU’s on-board hardware video encoder. The CG1 instances offer Nvid­ia’s Tesla M2050 GPUs (Fermi GF100).

Amazon says the initial driver release for the G2 includes support for OpenGL 4.3, DirectX 11, CUDA 5.5, and Open­CL 1.1. The CG1 drivers offer support for CUDA 5.5, OpenCL 1.1, and Direct­Compute. Both options provide access to the Grid SDK. Amazon says the pop­ular use cases for G2 will be service-side graphics applications including game streaming and 3D application stream­ing. The CG1 offers capabilities on the GPGPU compute side including mas­sively parallel tasks like computational chemistry, rendering, financial model­ing, and engineering design.

- See more at GfxSpeak

Find more on NVIDIA Grid for design and CAD professionals here.