Guide

Thermal and Mechanical Design
52 Intel® Xeon® Processor E7 2800/4800/8800 v2 Product Family
Thermal/ Mechanical Specifications and Design Guide
2.4 Design Considerations
2.4.1 System Design Considerations
When designing a thermally capable system, all critical components must be
simultaneously considered. The responsible engineer must determine how each
component will affect another, while ensuring target performance for all components.
The term “target performance” is used because some components (for example,
LRDIMM) have better performance, depending on how well they are cooled. The system
design team must set these target performance goals during the design phase so that
they can be achieved with the selected component layout.
The location of components and their interaction must be considered during the layout
phase. For example, memory that is heated by a processor will have worse
performance than a layout that does not shadow memory behind a processor. Although
the memory components have fixed thermal specifications, the performance
management of RDIMM will limit memory throughput to ensure that the temperature
limits are met. Consequently, a poorly cooled memory subsystem will have worse
performance. The processor is somewhat different in that it enables full performance at
all times, as defined by its specifications. The thermal engineer’s responsibility is to
ensure that each and every component meets its performance goals bounded by
thermal and acoustic specifications but also computing performance such as memory
throughput.
The thermal engineer directly influences the critical thermal parameters affecting
processor cooling capability. For a given heatsink and retention solution, the layout and
air-mover selection must ensure that all thermal specifications are met. It is desirable
to drive chassis air temperature rise as low as reasonably possible while maximizing
flow to each component. However, higher chassis temperature rise can be
accommodated as long as the design implements a countering flow increase. These
trade-offs are essential in designing a thermally capable system.
The number, size, and position of fans, vents, and other heat-generating components
determine the component thermal performance and the resultant local ambient and
airflow to the processor. The size and type (passive or active) of the thermal solution
and the amount of system airflow can be traded off against each other to meet specific
system design constraints.
In choosing the boundary conditions for a passive heatsink, the following methodology
is recommended to ensure that a system can deliver the required boundary conditions.
The Intel reference solution was developed by considering various system
implementations to ensure that the boundary conditions were within reason.
• Conceptualize the layout with the system architect, including approximate
volumetric constraints for the heatsink
• Select air movers that will deliver airflow and local temperatures within reason to
all system components (also account for Trise across the air-movers)
• Create a Computational Fluid Dynamics (CFD) model of the system
• Run the CFD model with varying flow resistance representing the finned section of
the heatsink
• Extract an effective air-mover curve from the CFD results
• Optimize the heatsink (fin thickness, quantity, base thickness, and so on) based on
the effective air-mover curve