|
Introduction
In a system environment,
the Pentium® processor's temperature is a function of both the
system and component thermal characteristics. The system level thermal
constraints imposed on the package are local ambient temperature
and thermal conductivity (i.e., airflow over the device). The Pentium®
processor thermal characteristics depend on the package (size and
material), the type of interconnection to the printed circuit board
(PCB), the presence of a heat sink, and the thermal conductivity
and the power density of the PCB.
All of these parameters are
aggravated by the continued push of technology to increase the operating
speeds and the packaging density. As operating frequencies increase
and packaging size decreases the power density increases and the
heat sink size and airflow become more constrained. The result is
an increased importance on system design to ensure that thermal
design requirements are met for each component in the system.
In addition to heat sinks
and fans, there are other solutions for cooling integrated circuit
devices. A few of these solutions are: fan mounted on heat sink,
heat pipes, thermoelectric (peltier) cooling, liquid cooling, etc.
While these alternatives are capable of dissipating additional heat,
they have disadvantages in terms of system cost, complexity, reliability,
and efficiency. These techniques are more expensive than a passive
heat sink and fan. The introduction of active devices can also decrease
reliability. Finally, the power efficiency of some of these techniques
is poor, and gets worse as the amount of power being dissipated
increases. Despite these disadvantages, each of these solutions
may be the right one for particular system implementations.
However, for the purpose
of this application note, Intel has focused its efforts on describing
solutions using passive heat sinks and fans.
Document Goal
The goal of this document
is to provide thermal performance information for the Pentium®
processor and recommendations for meeting the thermal requirements
imposed on systems. This application note attempts to provide an
understanding of the thermal characteristics of the Pentium®
processor and some examples of how the thermal requirements can
be met.
Importance
Of Thermal Management
Thermal management of an
electronic system encompasses all of the thermal processes and technologies
that must be employed to remove and transfer heat from individual
components to the system's thermal sink in a controlled manner.
The objective of thermal
management is to ensure that the temperature of all components is
maintained within functional and absolute maximum limits. The functional
temperature limit is the range within which the electrical circuits
can be expected to meet their specified performance requirements.
Operation outside the functional limit can degrade system performance
or cause logic errors. The absolute maximum temperature limit is
the highest temperature that a portion of the component may be safely
exposed. Temperatures exceeding the limit can cause physical destruction
or may result in irreversible changes in operating characteristics.
Higher temperatures result in earlier failure of the devices in
the system. For every 10°C rise above the operating range means
a halving of the mean time between failures.
Pentium®
Processor Power Specifications
The Pentium®
processor's power dissipation for 60 and 66 MHz is shown in Table
1.
Table 1. Pentium®
Processor Power Dissipation
| |
Package Type |
Total Pins |
Pin Array |
Package Size |
Power
(Typical) |
Power
(Max) |
| Pentium® processor
60 MHz |
PGA |
273 |
21x21 |
2.16"x2.16" |
11.9 W |
14.6 W |
| Pentium® processor 66 MHz |
PGA |
273 |
21x21 |
2.16"x2.16" |
13 W |
16 W |
|
To ensure functionality and
reliability of the Pentium® processor, maximum device
junction temperature must remain below 100°C. Considering the
power dissipation levels and typical ambient environments of 40°C
to 45°C, the Pentium® processor's junction temperatures
cannot be maintained below 100° C without additional thermal
enhancement to dissipate the heat generated by this level of power
consumption.
The thermal characterization
data described in Table 2 illustrates that both a heat sink and
airflow are needed. The size of heat sink and the amount of airflow
are interrelated and can be traded off against each other. For example,
an increase in heat sink size decreases the amount of airflow required.
In a typical system, heat sink size is limited by board layout,
spacing, and component placement. Airflow is limited by the size
and number of fans along with their placement in relation to the
components and the airflow channels. In addition, acoustic noise
constraints may limit the size or types of fans limiting the airflow.
To develop a reliable thermal
solution, all of the above variables must be considered. Thermal
characterization and simulation should be carried out at the entire
system level accounting for the thermal requirements of each component.
Thermal Parameters
Component power dissipation
results in a rise in temperature relative to the temperature of
a reference point. The amount of rise in temperature depends on
the net thermal resistance between the junction and the reference
point. Thermal resistance is the key factor in determining the power
handling capability of any electronic package.
Thermal resistance from junction to case (THETAJC), and
from junction to ambient (THETAJA) are the two most often
specified thermal parameters for integrated circuit packages.
Ambient Temperature
Ambient temperature is the
temperature of the undistributed ambient air surrounding the package.
Denoted TA, ambient temperature is usually measured at a specified
distance away from the package. In the laboratory test environment,
ambient temperature is measured 12 inches upstream from the package
under investigation. In a system environment, ambient temperature
is the temperature of the air upstream to the package and in its
close vicinity.
Case Temperature
Case temperature, denoted
TC, is measured at the center of the top surface of the package,
typically the hottest point on the package case. Special care is
required when measuring the case temperature to ensure an accurate
temperature measurement. Thermocouples are often used to measure
TC. Before any temperature measurements, the thermocouples have
to be calibrated. When measuring the temperature of a surface which
is at a different temperature from the surrounding ambient air,
errors could be introduced in the measurements. The measurement
errors could be due to having a poor thermal contact between the
thermocouple junction and the surface, heat loss by radiation or
by conduction through thermocouple leads. To minimize the measurement
errors, it is recommended to use the following approach:
- Use 36 gauge or finer
diameter K, T, or J type thermocouples. The laboratory testing
was done using a thermocouple made by Omega (part number: 5TC-TTK-36-36).
- Attach the thermocouple
bead or junction to the center of the package top surface using
high thermal conductivity cements. The laboratory testing was
done by using Omega Bond (part number: OB-100).
- The thermocouple should
be attached at a 90° angle as shown in Figure 1. When a heat
sink is attached a hole (no larger than 0.15'') should be drilled
through the heat sink to allow probing the center of the package
as shown in Figure 1.
- If the case temperature
is measured with a heat sink attached to the package, drill a
hole through the heat sink to route the thermocouple wire out.

Figure 1. Thermocouple Attachment
Junction
Temperature Junction temperature, denoted TJ, is the average
temperature of the die within the package.
The junction temperature for a given junction-to-ambient thermal
resistance, power dissipation, and ambient temperature is given
by the following formula:
If a heat sink with thermal resistance of QSA (sink-to-ambient) is
used, then the thermal resistance from the junction-to-case, QJC,
is given by the following formula:
TJ = PD
* (THETAJC + THETACS + THETASA)
+ TA
where:
THETACS is the
thermal resistance from the component (case) to the heat sink.
Thermal
Resistance
Thermal resistance (Figure
2) values for junction-to-ambient, THETAJA, and junction-to-case,
THETAJC, are used as measures of IC package thermal performance.
QJC is a measure of the package's internal thermal resistance along
the major heat flow path from silicon die to package exterior. This
value is strongly dependent on the material, thermal conductivity,
and geometry of the package. THETAJC values also depend
on the location of the reference point (in this case center of the
package top surface), the external cooling configurations and the
heat flow paths from the package to the ambient. For example, if
a heat sink is attached to the package top surface or more heat
is pulled into the board through the pins, the THETAJC
values measured with reference to the center of the package top
surface will change. THETAJA values include not only
internal thermal resistance, but also the radiative and convective
thermal resistance from the package exterior to ambient air. THETAJA
values depend on the material, thermal conductivity, and geometry
of the package and also on ambient conditions such as airflow rates
and coolant physical properties.
In order to obtain thermal
resistance values, junction temperature is measured using the temperature
sensitive parameter (TSP) method. With this method, special design
thermal test structures are used which are approximately the same
size as the Pentium® processor die. The test structure consists
of resistors and diodes. Resistors are used to simulate the Pentium®
processor power dissipation and thereby heat up the package. Diodes,
which are located at the center of the thermal test die, are used
to measure the die temperature. The measurements are carried out
in a wind tunnel environment. The air flow rate and the ambient
temperature are measure 12 inches away from the package in the upstream
air.
The parameters are defined
by the following relationships:
THETAJA = (TJ
- TA) / PD
THETAJC = (TJ
- TC) / PD
THETAJA = THETAJC
+ THETACA
where:
THETAJA = junction-to-ambient thermal resistance (°C/W)
THETAJC = junction-to-case thermal resistance (°C/W)
THETACA = case-to-ambient thermal resistance (°C/W)
TJ = average die (junction) temperature (°C)
TC = case temperature at a pre-defined location (°C)
TA = ambient temperature (°C)
PD = device power dissipation (W)

Figure 2. Thermal
Resistance Parameters
Table 2 lists the junction-to-case
and case-to-ambient thermal resistance's for the Pentium® processor
(with and without a heat sink).
Table 2. Thermal
Characterization Data
| |
THETAJC |
THETACA
vs. Airflow (ft./min.) |
| 0 |
200 |
400 |
600 |
800 |
1000 |
| With 0.25" heat sink |
0.6 |
8.3 |
5.8 |
3.9 |
3.0 |
2.5 |
2.2 |
| With 0.35" heat sink |
0.6 |
7.9 |
5.0 |
3.4 |
2.8 |
2.2 |
2.0 |
| With 0.65" heat sink |
0.6 |
6.4 |
3.4 |
2.3 |
1.8 |
1.5 |
1.3 |
| Without a heat sink |
1.2 |
11.6 |
9.4 |
6.7 |
5.4 |
4.6 |
4.2 |
|
NOTE:Heat Sink: 2.05 sq.
in. omni-directional pin, aluminum heat sink with 0.050 in. pin
width, 0.143 in pin-to-pin center spacing and 0.150 in. base thickness.
A thin layer of thermal grease with an average thickness of .003''to
.005'' was used as the interface material between the heat sinks
and the package.
Designing For
Thermal Performance
At this point the application
note turns from describing the characteristics that define thermal
performance to describing how designers should use these characteristics
to assess thermal requirements of PC system designs. The Pentium®
processor specifies a maximum case temperature, TC, of 85°
C. This case temperature limit along with the Pentium® processor's
power and thermal resistance characteristics can be used to determine
the ambient temperature required to keep the Pentium® processor
operating within its specified limits. Using these parameters in
the following equations:
The maximum ambient temperature
required in a Pentium® processor system without any additional
thermal enhancement is -100.6° C at 66 MHz. Obviously, this
ambient temperature is impractical and unachievable in a PC system.
In order to be able to maintain the case temperature at 85°C
in a typical system ambient with air temperature of 40 - 45°
C, the thermal resistance between the case and the ambient must
be reduced.
HEAT SINKS
The most common way to improve
the package thermal performance is to increase the surface area
of the device by attaching a large piece of metal (a heat sink)
to the ceramic package. The heat sink is usually made of Aluminum
and is chosen for its price/thermal-performance ratio. There are
materials that offer higher conductivity such as copper, but cost
becomes prohibitive. To maximize the flow of heat for a given junction
temperature rise over the ambient temperature, the thermal resistance
from heat sink to air can be reduced by a) maximizing the surface
area, and b) maximizing the air flow across the surface area (maximizing
air flow through heat sink fins in most cases).
Intel has used
test data to determine what size of heat sink and airflow is needed
to properly cool a Pentium® processor system. The data
was derived assuming an adhesive attach process that offers thermal
resistance of about 0.2 °C / W.
The testing was done in a
wind tunnel in the configuration (in Figure 3) where the heat sink
was mounted on a real Pentium® processor ceramic package with
a thermal die mounted inside to generate the 16 Watts of power.
The package is then mounted in a socket which is soldered to a 2-layer
PCB that brings power to the die.

Figure 3. Improving
Thermal Performance
Based on these tests, three
specific heat sink and airflow combinations have been identified
that properly dissipate the Pentium® processor's 16 Watts and
maintains a case temperature below 85°C. The three heat sinks
are shown in Figure 4.
Heat sink A: H = 2.3''for
100 LFMs of Air
Heat sink B: H = 1.2''for
200 LFMs of Air
Heat sink C: H = 0.5''for
500 LFMs of Air
Assumption: Air Temp = 45°C

Figure 4. Recommended Combinations
In addition, testing has
been done to provide more general guidelines which allow deviating
from the above conditions. These guidelines allow systems to derive
various combinations of heat sink size and airflow that ensure the
Pentium® processor thermal specifications are met. For example,
by increasing the heat sink x-y dimensions and extending it over
the package footprint, the heat sink height can be reduced while
maintaining the same thermal performance as the taller heat sink
with the same footprint as that of the package. The first three
charts (Figures 5, 6, 7) show the thermal resistance as a function
of heat sink size and airflow. The last three charts (Figures 8,
9, 10) show the power dissipation achievable with a given heat sink
size and airflow. The power dissipation calculations assume TC
= 85°C, TA = 45°C, and THETAJC =
0.6°C/W.
Pmax = (TC-
TA) / (THETAJA- THETAJC) = 40 /
(THETAJA- 0.6)
A key assumption in all
of these calculations is that a perfect thermal connection can be
achieved between the case and the heat sink. One can extrapolate
the heat sink solutions by adding the additional thermal resistance
of any chosen heat sink attach process.

Figure 5. Thermal
Resistance

Figure 6. Thermal Resistance

Figure 7. Thermal Resistance

Figure 8. Power Dissipation

Figure 9. Power Dissipation

Figure 10. Power Dissipation
Airflow
To improve the effectiveness
of heat sinks it is important to manage the air flow so as to maximize
the amount of air that flows over the device or heat sink's surface
area. In the system, the air flow around the processor can be increased
by providing an additional fan or increasing the output of existing
fan. If this is not possible, baffling the airflow to direct it
across the device may help. This means the addition of sheet metal
or objects to guide the air to the target device. Often the addition
of simple baffles can eliminate the need for an extra fan. In addition,
the order in which air passes over devices can impact the amount
of heat dissipated.
Fans
Fans are often needed to
assist in moving the air inside a chassis. A typical variable speed
fan capable of up to 100 CFM of air can be found for approximately
$10.
The airflow rate
is usually directly related to the acoustic noise level of the fan
and system. Therefore maximum acceptable noise levels may limit the
fan output or the number of fans selected for a system.
A fan may be placed at the
top of a heat sink to produce direct air impingement on the heat
sink for efficient heat removal. A key issue with fans is their
reliability. Although many fans are rated for approximately 50,000
hours of operation, operating conditions such as operating temperature,
pressure drop across the fan and the particles in the air can significantly
reduce the fans useful life.
Conclusion
As the complexity of today's
microprocessors continues to increase so do the power dissipation
requirements. Care must be taken to ensure the additional heat resulting
from the power is properly dissipated. As documented, the heat can
be dissipated using passive heat sinks, fans and/or active cooling
devices.
The simplest and probably
most cost effective method is to use a heat sink and a fan. The
size of the heat sink and the output of the fan can be varied to
balance the tradeoffs between size and space constraints versus
noise. For example, if space is available a 1.2'' high heat sink
can be used with only 200 LFM to cool the 66-MHz Pentium® processor
and the 16W it dissipates. Another example in which space is restricted
shows a 0.5'' high heat sink can be used if approximately 500 LFM
of airflow is provided.
These are not the only valid
solutions, but both provide adequate cooling to maintain the Pentium®
processor case temperature at or below the 85° C specified
limit. By maintaining this specification, the system can guarantee
proper functionality and reliability of the Pentium®
processor.
This applies to:
|