Video management for compute nodes

An method comprising:

• at a plurality of video controllers associated with a plurality of compute nodes, providing first requests associated graphics rendering operations to a first memory address space shared in common by the video controllers;
• receiving, at an address translator, the first requests;
• converting, at the address translator, the first requests into second requests directed to a plurality of separate memory spaces such that each separate memory space is associated with a different video controller of the plurality of controllers;
• receiving, at a memory controller, the second requests;
• generating, by the memory controller, memory cycles with a memory in response thereto;
• receiving, at a refresh multiplexer, a set of time overlapping display refresh requests from a group of the video controllers,
• selecting, by the refresh multiplexer, one or more of the video controllers,
• providing, by the refresh multiplexer, data from a rendered frame video memory to the selected one or more video controllers, and
• providing substitute data to the one or more non-selected video controllers of the group,
• wherein the selected one or more video controllers comprise a first video controller and the non-selected one or more video controllers comprise a second video controller,
• providing a first clock signal by a phase locked loop (PLL);
• providing a second clock signal by a non-PLL based frequency divider;
• and
• in response to the selection of the first video controller, providing, by a clock selector, the first clock signal to the first video controller to clock the first video controller and, in response to the non-selection of the second video controller, provide the second clock signal to the second video controller to clock the second video controller.