Method and system for cloud based virtualized graphics processing for remote displays

Abstract

An apparatus for providing graphics processing. The apparatus includes a dual CPU socket architecture comprising a first CPU socket and a second CPU socket. The apparatus includes a plurality of GPU boards providing a plurality of GPU processors coupled to the first CPU socket and the second CPU socket, wherein each GPU board comprises two or more of the plurality of GPU processors. The apparatus includes a communication interface coupling the first CPU socket to a first subset of one or more GPU boards and the second CPU socket to a second subset of one or more GPU boards.

What is claimed: 
     
         1 . An apparatus for providing graphics processing, comprising:
 a dual CPU socket architecture comprising a first CPU socket and a second CPU socket;   a plurality of GPU boards providing a plurality of GPU processors coupled to said first CPU socket and said second CPU socket, wherein each GPU board comprises two or more of said plurality of GPU processors; and   a communication bus interface coupling said first CPU socket to a first subset of one or more GPU boards and said second CPU socket to a second subset of one or more GPU boards.   
     
     
         2 . The apparatus of  claim 1 , further comprising:
 a multi-core processor coupled to said first CPU socket.   
     
     
         3 . The apparatus of  claim 2 , wherein said multi-core processor comprises a XEON E 2670 processor coupled to said first CPU socket. 
     
     
         4 . The apparatus of  claim 1 , wherein said communication bus interface comprises:
 a plurality of communication bridges each coupling a corresponding GPU board to said first CPU socket and said second CPU socket.   
     
     
         5 . The apparatus of  claim 4 , wherein at least one of said communication bridges comprises a PCIe bridge. 
     
     
         5 . The apparatus of  claim 5 , wherein each of said plurality of GPU boards comprises:
 a bridge splitter coupled to a corresponding PCIe bridge; and   two or more of said plurality of GPU processors.   
     
     
         6 . The apparatus of  claim 5 , wherein said plurality of GPU boards are distributed across said dual CPU socket architecture symmetrically. 
     
     
         7 . The apparatus of  claim 6 , further comprising:
 a first GPU board and a second GPU board each coupled to said first CPU socket; and   a third GPU board and a fourth GPU board each coupled to said second CPU socket;   wherein each of said first, second, third, and fourth GPU boards comprises a two GPU processor configuration.   
     
     
         8 . The apparatus of  claim 5 , wherein said plurality of GPU boards are distributed across said dual CPU socket architecture asymmetrically. 
     
     
         9 . The apparatus of  claim 1 , further comprising:
 a front side bus network coupling said first CPU socket and said second CPU socket;   
     
     
         10 . The apparatus of  claim 1 , wherein said front side bus network comprises:
 a QPI link communicatively coupling said first CPU socket and said second CPU socket; and   a single operating system managing CPU processors coupled to said first CPU socket and said second CPU socket as supported by said QPI link.   
     
     
         11 . The apparatus of  claim 1 , wherein at least one of said plurality of graphics processors comprises an Nvidia GK107 graphics processor chip. 
     
     
         12 . The apparatus of  claim 1 , wherein a GPU processor supports a virtual machine in a one-to-one relationship. 
     
     
         13 . The apparatus of  claim 1 , further comprising:
 a plurality of virtualized GPU processors supported by said plurality of GPU processors, wherein a virtualized GPU supports a virtual machine.   
     
     
         14 . The apparatus of  claim 1 , wherein said dual CPU socket architecture is configured in a Sandy bridge configuration. 
     
     
         15 . The apparatus of  claim 1 , wherein said plurality of GPU boards are identical. 
     
     
         16 . The apparatus of  claim 1 , wherein said dual CPU socket architecture and said plurality of GPU boards support a plurality of virtual machines each comprising portions of one or more CPU cores and one or more GPU processors. 
     
     
         17 . The apparatus of  claim 1 , wherein said dual CPU socket architecture and said plurality of GPU boards are implemented within a pseudo virtual machine system operating under a single operating system and running multiple applications for one or more end users. 
     
     
         18 . A network attached GPU device, comprising:
 a plurality of processing boards providing a plurality of virtual CPU and GPU processors, wherein each of said processing boards comprises:
 a dual CPU socket architecture comprising a first CPU socket and a second CPU socket; 
 a plurality of GPU boards providing a plurality of GPU processors coupled to said first CPU socket and said second CPU socket, wherein each GPU board comprises two or more of said plurality of GPU processors; 
 a first plurality of communication bridges each coupling a corresponding GPU board to said first CPU socket and said second CPU socket; and 
 a communication interface coupling said first CPU socket to a first subset of one or more GPU boards and said second CPU socket to a second subset of one or more GPU boards. 
   
     
     
         19 . The apparatus of  claim 18 , wherein each of said plurality of GPU boards comprises:
 a bridge splitter coupled to a corresponding communication bridge; and   two or more of said plurality of GPU processors.   
     
     
         20 . The apparatus of  claim 18 , wherein said plurality of GPU boards are distributed across said dual CPU socket architecture symmetrically such that a first GPU board and a second GPU board each coupled to said first CPU socket and a third GPU board and a fourth GPU board each coupled to said second CPU socket, wherein each of said first, second, third, and fourth GPU boards comprises a two GPU processor configuration. 
     
     
         21 . The network attached GPU device of  claim 18 , further comprising:
 a plurality of computing device emulators supported by said plurality of processing boards.