Fan Speed Control (FSC) Design Process. Intel Xeon 3500 Series

Sensor Based Thermal Specification Design Guidance

Figure 5-5. Thermal Solution Performance vs. Fan Speed

Note:

5.4

0.50

5.9

0.40

5.4

4.9

0.30

4.4

3.9

0.20

3.4

2.9

0.10

2.4

0.00

600 1100 1600 2100 2600 3100 3600

1.9

RPM

Psi-ca System (BA)

This data is taken from the validation of the RCBF5 reference processor thermal solution. The Ψ

CA

vs. RPM data is available in

Table 5-1 at the end of this chapter.

Fan Speed Control (FSC) Design Process

The next step is to incorporate the thermal solution characterization data into the algorithms for the device controlling the fans.

As a reminder, the requirements are:

• When the DTS value is at or below T

CONTROL with prior processors.

, the fans can be slowed down; just as

• When DTS is above T

CONTROL

, FSC algorithms will use knowledge of T

AMBIENT

and

Ψ

CA

vs. RPM to achieve the necessary level of cooling.

This chapter discusses two implementations. The first is a FSC system that is not provided the T

AMBIENT current T

AMBIENT

information and a FSC system that is provided data on the

. Either method will result in a thermally compliant solution and some acoustic benefit by operating the processor closer to the thermal profile. But only the

T

AMBIENT

aware FSC system can fully use the specification for optimized acoustic performance.

In the development of the FSC algorithm it should be noted that the T

AMBIENT

is expected to change at significantly slower rate than the DTS value. The DTS value will be driven by the workload on the processor and the thermal solution will be required to respond to this much more rapidly than the changes in T

AMBIENT

.

An additional consideration in establishing the fan speed curves is to account for the thermal interface material performance degradation over time.

Thermal/Mechanical Design Guide 33