NVIDIA announces GRID 2.0: Desktop virtualization more efficient

At VMworld in San Francisco, the annual conference organized by VMware, NVIDIA GRID 2.0, with the aim of managing multiple users and give more performance to each of them.

NVIDIA announced GRID 2.0, the new platform of Virtual Desktop Infrastructure (VDI) accelerated by the GPU. The goal with version 2.0 is to serve twice as many users, thanks to the expansion of the efficiency of the data center, to twice the performance. The new technology also supports Linux as well as Windows.

With GRID, NVIDIA speaks of performance and scalability in the context of virtualization, translating the concepts applied in the past to calculate local key processing remotely and also applying in this area passing architecture of great success in the desktop world, Maxwell. He has now data centers located in various parts of the world, which is able to provide major corporations to enable them to perform more complex calculations on the basis of their business.

The goal of NVIDIA is to remove barriers to the productivity of each employee, giving him a chance to work where he wants thanks to computing power made available by the cloud. And to make sure this is possible has invested billions of dollars in R&D.

Underlying the transitions from GRID 1.0 to GRID 2.0 are the new cards Tesla M60 and M6, supplanting GRID K1 and K2, at the basis of the first version of the virtualization platform. If the previous cards were based respectively on Kepler GK107 and GK104, the new ones are built around the GPU GM204, which are based on the same video cards GeForce GTX 970 and 980.

The difference between the M60 and M6 consists in the fact that the first configuration provides a dual-GPU and is equipped with 16GB GDDR5 on board, unlike the 8GB GDDR5 present on M6, which is configured for single-GPU. M60 consumes between 225W and 300W depending on the performance requirements and the cooling configuration. According to estimates by NVIDIA also a single card M60 can manage up to 32 remote users connected simultaneously. In the case of M6 consumption stood between 75W and 100W.

Maxwell also allows for more efficient form factor and small size more. The NVIDIA gives way to increase the density of blade servers that are at the basis of the functioning of the VDI. The new blade server GRID 2.0 offer a density hardware unimaginable at the time of GRID 1.0, thus doubling the number of users per server can be managed: the new value is 128 users per server. At the same time, more hardware densities and lower workloads for each individual tab allow to unleash a power per user double compared to the previous configuration.

Also, GRID 2.0 supports CUDA, while this type of support was first absent within the vGPU. CUDA applications will finally be completely virtualized. GRID 2.0 introduce also support monitors 4K, while the maximum resolution was previous to WQXGA (2,560 x 1,600). The new limit for GRID 2.0 is 4 to 4K monitor VM. NVIDIA is currently testing the new technology with 12 companies, and will support the other customers on September 15.

GRID 2.0 is a platform designed for various types of companies, the scope to medical business, up to education. ” Think of an airport consists of 70 or 80 models that need to stay connected with each other and work together in a safe environment. For this structure as a VDI environment is the ideal solution, ” said Will Wade, director of GRID Product Management.