on page 11 shows a conceptual block diagram of the motherboard
memory riser board pathways.
• If processors access memory located in another NUMAlinked MC990 X server chassis motherboard
within the system, the memory is referred to as remote memory.
• The total memory within the NUMAlinked Integrity MC990 X system is referred to as global memory.
Physical memory riser
Physical memory nodes are memory risers and each is made up of two board assemblies: memory riser
board and power board. Eight memory riser assemblies plug into each MC990 X server chassis
motherboard. The memory riser power board receives power directly from the motherboard. A maximum
of 12 DDR4 memory DIMMs are supported in each memory riser.
1.
Memory controller
Figure 6: MC990 X server chassis memory riser block diagram
Distributed shared I/O
Like DSM, I/O devices are distributed within the MC990 X server chassis. Each BaseIO riser is accessible
by all compute nodes within the SSI through the NUMAlink interconnect fabric.
Rack management controller
Each Integrity MC990 X system has a rack management controller (RMC) located directly above or below
the MC990 X server chassis in a rack. The RMC supports powering up and down of the system
motherboards and environmental monitoring of all Integrity MC990 X system units within the SSI. In
addition, the RMC provides the top layer of system control for Integrity MC990 X system. Through the use
of an internal 24-port Ethernet switch, a single RMC can provide system control for multiple MC990 X
server chassis in an expanded Integrity MC990 X system.
One GigE port from each MC990 X server chassis motherboard connects to the RMC via Cat-5 cable.
Reliability, availability, and serviceability
HPE Integrity MC990 X system components have the following features to increase the reliability,
availability, and serviceability (RAS) of the systems.
Physical memory riser
11
