vGPU Capabilities


HmC has implemented virtual graphics processing unit (vGPU) and graphics processing unit (GPU) capabilities to support select users that require additional resources for graphical or computational intensive applications.

Capacity & Availability

At this time, the vGPU resources are in a separate cloudspace. If an organization needs access, they should submit a support request. Connectivity between existing cloudspaces and the vGPU cloudspace is limited and will be opened on a case by case basis.

Hanscom Cloud (x86)

Timeline Model Total RAM Total Cores Total CUDA Cores
Current NVIDIA K1 GRID 80 GB 20 3,840
Nov 2019 (est) NVIDIA K1 GRID 160 GB 40 7,680

SciFi Cloud (PowerPC)

Timeline Model Total RAM Total CUDA Cores Total Tensor Cores
Current NVIDIA TESLA P100-SXM2 (NVLINK) 64 GB 14,336 -
DEC 2019 NVIDIA TESLA V100-SXM2 (NVLINK) 32 GB 10,240 1,280

Commercial Providers

HmC users can also access the GPU resources available from commercial providers like Amazon Web Services and Azure.


The Hanscom Cloud (x86) includes NVIDA GRID technology. There are two approaches for enabling vGPU to systems in the cloud - virtual Shared Graphics Acceleration (vSGA) and virtual Direct Graphics Acceleration (vDGA).

vSGA allows all deployed systems on a hypervisor to access the vGPU resource, much like they share the CPU. This implementation should support most use cases for graphic acceleration. The application must support up to DirectX 9.0c and OpenGL 2.1. Applications that require a later version of DirectX may not perform correctly when using vSGA. Each system is allocated a portion of the available video memory, currently 128MB each in HmC.

vDGA requires discrete mapping of a virtual machine to a core on the vGPU. At this time, there is no cloud level automation to support this; the cloud administrators would have to perform a manual action. If a program does not have success with the vSGA implementation, they can request a vDGA pilot. vDGA generally supports the versions of DirectX and OpenGL that the GPU manufacturer’s graphics driver supports. If approved for a pilot, the HmC admin team will work with the program to "test drive" their application in a vDGA mode.

Work in the SciFI Cloud includes NVIDIA TESLA V100 and P100 technology with the NVLink high-speed, bidirectional interconnect. It requires PowerPC hosts to access the available GPUs.


All hypervisors with vGPU capabilities are configured with the NVIDIA drivers. The operating system templates in the vGPU enabled cloudspaces have had the appropriate video driver installed and tested for vSGA access (e.g., VMware SVGA 3D).

Note: There may be driver version dependencies which may affect your application


  1. Ensure the application supports the correct versions of DirectX and OpenGL
  2. Validating browser support for OpenGL:
  3. Validating performance of OpenGL & vGPU on a system:

The HmC admin team can validate deployment runs and video memory usage in support of troubleshooting. However, at this time, they must be on keyboard while the test in underway, so access to this information is reserved for coordinated troubleshooting events with the user.