Documentation/Labs/GPU Virtualization

From Slicer Wiki
Revision as of 20:00, 8 February 2013 by JChris.FillionR (talk | contribs) (Created page with 'This page list information I found while trying to answer the question: "Can multiple virtual machins share a single GPU?" * The new Nvidia VGX platform is notable for allowing…')
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Home < Documentation < Labs < GPU Virtualization

This page list information I found while trying to answer the question: "Can multiple virtual machins share a single GPU?"


  • The new Nvidia VGX platform is notable for allowing multiple users to share a GPU on their virtual desktops.
[...]
Huang said during the keynote that Nvidia had been working on this GPU virtualization solution for five years. The VGX platform enables every virtual machine to have a GPU
[...]
While he noted that putting a GPU into a virtual machine isn't new, what is new is the ability to share that GPU with multiple virtual machines.
[...]

Source: http://redmondmag.com/articles/2012/05/17/nvidia-unveils-virtualized-gpu-supporting-multiple-users.aspx

  • GPU Virtualization on VMware’s Hosted I/O Architecture
[...]
GPUs pose a unique challenge in the field of virtualiza- tion. Machine virtualization multiplexes physical hardware by presenting each VM with a virtual device and combin- ing their respective operations in the hypervisor platform in a way that utilizes native hardware while preserving the illusion that each guest has a complete stand-alone de- vice. Graphics processors are extremely complicated devices.
[...]
Thus, it is nearly intractable to provide a virtual device corresponding to a real modern GPU. Even starting with a complete implementation, updating it for each new GPU generation would be prohibitively laborious. Thus, rather than modeling a complete modern GPU, VMware’s
primary approach paravirtualizes: it delivers an idealized software-only GPU and our own custom graphics driver for interfacing with the guest operating system
[...]

Source: http://graphics.stanford.edu/~yoel/notes/gpu-osr.pdf