nvidia-gridBefore I get into the titel of this article, let me first give some context. During VMworld in the US NVIDIA announced the 2.0 version of NVIDIA GRID. NVIDIA GRID 2.0 delivers unprecedented performance, efficiency and flexibility improvements for virtualized graphics in enterprise workflows. With GRID 2.0 NVIDIA continues their effort to make it possible for companies to virtualize even the most intensive enterprise workflows.

NVIDIA GRID 2.0 offers the following:

  • Doubled user density: doubled user density over the previous version, allowing up to 128 users per server. This enables enterprises to scale more cost effectively, expanding service to more employees at a lower cost per user.
  • Doubled application performance: Using the latest version of NVIDIA’s award-winning Maxwell™ GPU architecture, NVIDIA GRID 2.0 delivers twice the application performance as before — exceeding the performance of many native clients.
  • Blade server support: Enterprises can now run GRID-enabled virtual desktops on blade servers from leading blade server providers.
  • Linux support: No longer limited to the Windows operating system, NVIDIA GRID 2.0 now enables enterprises in industries that depend on Linux applications and workflows to take advantage of graphics-accelerated virtualization.

GRID 2.0

Beside GRID 2.0 NVIDIA also announced two new graphics card along with it, namely the Tesla M60 and Tesla M6. The M60 is a full size form factor card and the M6 is designed for use in blade systems. The technical specs can be found in the table below:

Tesla M60Tesla M6
CUDA Cores4096 (2x 2048)1536
VRAM16GB GDDR5 (2x 8GB)8GB GDDR5
Concurrent Users2-321-16
H.264 1080p30 Streams3618
Form FactorDual Slot PCIeMXM
TDP300W100W
GPU2x GM204GM204

So now that we have a little context lets get to the core of this article.

The NVIDIA GRID 2.0 solution is made up of three components: NVIDIA GRID enterprise software and support, and an NVIDIA Maxwell® architecture-based GPU. In this model the hardware and support haven’t changed, but now there is also a software component added. For this component you have to own a license, which is new to the NVIDIA GRID solution.

So how does this license model work? Well there are three licensed versions and depending on the lincense you own, features will become available. The table below shows which features will be available in which edition:

Virtual PCVirtual WorkstationVirtual Workstation Extended
Maximum Number of Displays 2 4 4
Maximum Resolution Per Display 2560 x 1600 (WQXGA) 2560 x 1600 (WQXGA) 3840 x 2160 (4K)
Windows Guest OS XXX
Linux Guest OSX X
NVIDIA® Quadro® Software FeaturesX X
CUDA Supported X
Open CL Supported X
GPU Pass-Through Supported X
vGPU Profiles Supported:
512 MB (Up to 16 users per GPU) X X X
1 GB ( Up to 8 users per GPU) X X X
2 GB (Up to 4 users per GPU) X X X
4 GB (Up to 2 users per GPU) X
8 GB (1 user per GPU) X

As with most other software, depending on what your needs are you can match your license to it. At the moment there are no prices available for both the licenses and the hardware.

During VMworld in Barcelona I had a talk with one of the NVIDIA employees asking for some more details, which I will recap below:

  • When booting a host that contains a graphics card the host will contact a license server to claim a license
  • Shutting down a host will release the license that was used by the host
  • A license server can be Windows or Linux based, a appliance is to be expected in the future
  • When no license server can be contacted a 7 day grace period will be in affect
  • After the grace period the number of frames will be reduced to 3 and a resolution of 1280 x 1024 will be applied
  • It is possible to mix different licenses, spread over multiple hosts
  • New features will become available which can be used depending on the support and license you own

Until the prices become available we will have to wait and see if this new model will change the total costs for companies. Hopefully the cost model will lower the bar for companies that want to start using the NVIDIA GRID. The M60 and M6 graphics cards are being shipped to OEMs at the moment and are most likely available in the beginning of next year.