Guidelines for Temperature Management of PCIe Devices

Overview

Measurement devices must be kept at stable temperatures to not only prolong the life of the devices but also ensure the most accurate measurements. Although PXI/PXI Express chassis are designed with forced air cooling and thermal monitoring to keep high-powered electronics cool, there is no standard method of cooling for PCs. There is typically a fan and heat sink for the CPU, a fan for the power supply, and maybe local cooling on the graphics card, but little focus on the PCI Express devices. For device longevity and the highest quality measurements, you should concentrate some attention on PCI Express card cooling.  

Contents

Identify Slot Power Contributors

PCI Express devices can dissipate 25 W of power for x1, x4, and x8 slots and up to 75 W for x16 slots. Some of these cards can dissipate much more if they have an onboard power connector. With multiple PCI Express measurement devices plus a graphics card in a PC chassis, you can see how easy it is to generate considerable heat inside a PC chassis. For this reason, it’s important to focus on cooling.

Computer chassis vary greatly in thermal management. The variability includes:

• Number and efficiency of fans
• Placement of heat sinks
• Cord management and congestion inside the chassis
• Heat generated by other components including the CPU and memory

Because of this variability, there is no one-size-fits-all approach to thermal management. You must consider many factors.

First, have an idea of the power dissipated in a system. Hard drives dissipate relatively little power—less than 10 W. The use of solid state drives can further lower this. DVD and Blu-ray drives also contribute little to the total power dissipation—less than 5 W idle and up to 5X that when reading/writing. Computer memory (RAM) modules contribute little as well. Each module generally dissipates less than 10 W, but this is dependent on the clock speed and operating voltage. Newer technology uses lower voltages and therefore dissipates less power. Faster clock rates require higher power.  

The larger contributors vary greatly in power consumption and include graphics cards, processors (CPUs), and PCI Express peripheral cards. Graphics cards can range from 25 W up to 350 W for high-end graphics cards. Look up the power dissipation of your graphics card and be sure to choose the right card for the minimal performance needed. Also, pay attention to its cooling method. Many graphics cards have heat sinks and fans, but the fans do not exhaust the air out of the chassis. This results in an overall warming of surrounding components. Higher end graphics cards exhaust the hot air out of the chassis. Similar to graphics cards, CPUs can vary greatly from 50 W to 150 W dissipation. Recently, newer technology has focused more on reduced power consumption than maximum performance. For that reason, use the latest technology but be aware that a higher number of cores and faster clock speed contribute to higher power dissipated. Like graphics cards, CPUs include heat sinks and fans but generally do not exhaust hot air outside the PC case.  

The final contributors, and focus of this white paper, are the PCI Express peripheral cards. NI products specify the maximum power used by PCI Express modules in their specification sheet. This power spec includes the maximum power of the module when fully loaded but may not include power dissipated in the module in a fault condition. For example, this is the power specified for the PCIe-6376 simultaneous multifunction I/O module:

From this table, you may conclude that this module dissipates 20.35 W total, however, this is the total power consumed from the computer. Some of this power is dissipated outside the PC chassis and, therefore, does not contribute heat to the system. For example, this module provides +5 V to the user as well as digital outputs that can source current. According to the Current Limits section of the specification, the total current from these two outputs is 1.7 A at 5 V. That means that 8.5 W of power is being dissipated outside the PC case. So, the actual power dissipated inside the case is 11.85 W, which is considerably less.  

Compared to other high-power components, 25 W (max) may not sound like a lot for a x1, x4, x8 PCI Express card, however, many motherboards support five or more PCI Express devices. With multiple cards in the system, the total power can add up. Even worse, PCI Express devices generally don’t provide a method for active cooling. Without airflow, pockets of still air create local hot zones around the PCI Express devices. The rest of this paper focuses on keeping these modules cool.  

 

 

Remove Heat From Your System

Now that you’ve identified the power contributors, you can focus on how to remove heat from the system. Many thermal contributors either provide no active cooling or vent hot air back into the PC case to be recirculated. This hot air must be vented with a combination of case fans and vent holes.  

A major consideration should be convection. Hot air rises, and you can take advantage of this. It’s common to see a fan mounted at the top rear of the case, blowing outward. This pulls air from the front of the PC, usually through vent holes near the bottom. This configuration not only works with convection but also prevents hot air from blowing out of the front of the chassis toward the user. You don’t want to try to pull the air down toward the bottom of the case. This method works against convection and has the effect of pulling risen, hot air back over the hot components.  

Much discussion online surrounds inward versus outward airflow, that is positive versus negative pressure. Should a fan be oriented to blow into the case or out of it? Both solutions can achieve the wanted effect if you consider convection and identify hot components. A correctly designed system provides efficient cooling with streamline or laminar airflow. You want this type of airflow because there is little turbulence that recirculates hot air that can cause hot pockets. You can achieve this with positive or negative pressure if there is a clear inlet and output path.

To achieve laminar airflow, you need a clean, unrestricted chassis. If there are any unused peripheral cards, remove them because they only trap heat. For peripherals in use, space them out if possible.  Poor cable management can restrict airflow, so tightly bundle or remove unused power cables and connectors. Most PC chassis have some way to use cable ties to secure cables out of the way. Be sure that you’re not blocking an inlet or output vent. There should be plenty of space around the PC. If the chassis is cleanly organized but there is still insufficient air flow, you may need to upgrade to a larger PC case. Finally, keep your PC clean. Dust is an insulator and makes even a well-designed system inefficient at removing heat.   

 

Cool Your System With Fans

You can choose from many fan options, but few are designed to operate at high temperatures or for extended periods. For this reason, you should consider industrial fans for these applications. When looking at fans from reputable manufacturers like ebm-papst, Noctua, and Sanyo Denki,  consider the following features.

 

Material

The fan’s construction material contributes to not only its max temperature rating but also its durability. This affects how well the fan can handle the accidental foreign object inserted while spinning. Typically, high-quality industrial fans contain fiberglass reinforcement that can take the high temperatures and hazardous environments.  

 

Bearings

Low-cost fans typically contain sleeve bearings that rely on oil/grease for smooth rotation. Over time, the lubricant is lost or changes in viscosity—this is accelerated at high temperatures. Because of the change in lubricant, the fan can fail abruptly and negatively affect the system with little warning. For this reason, you should consider ball bearings. These are sometimes a bit louder because there are moving parts, but they have a much longer lifetime when compared to sleeve bearings. Industrial fan manufacturers have made improvements in both sleeve and ball bearing designs that outperform the standard bearings, making the choice more complicated.   

 

Speed

It may be intuitive to focus on fan speed. You may expect a 2,000 rpm fan to outperform a 1,500 rpm fan but that may not be the case. It’s important to look at the cubic feet per minute, or cfm. This merit describes the amount of air the fan can move. Speed is a factor but there are others as well, such as diameter of the fan, number of blades, angle of the blades, and aerodynamic design. That being said, the importance of a properly designed system with efficient airflow trumps the effect of the fan cfm.  

 

Speed Control

You can control fan speed in a couple of different ways, and it’s important to know what types of fans are compatible with the system you are designing. The two main ways are DC voltage and PWM control.  For a DC-controlled fan, the speed is proportional to the DC voltage that is applied. Some DC fans provide a signal indicating the fan speed back to the system. For PWM fans, the fan speed is proportional to the duty cycle of the signal applied. These fans provide a speed-indication signal back to the system as well.  

 

IP Rating

Some fans include an ingress protection, or IP, rating. This rating system uses two numbers to indicate the level of dust and water protection. As shown in the figure below, the fist number ranges from zero to six and gives the level of dust protection. An IP5x or IP6x fan is recommended to keep out some or all harmful dust that may cause a fan to fail prematurely. The second number indicates the protection from liquids. This is generally not a concern for computer-based systems since other components would fail in wet environments before the fan. Note that a fan that is IP rated doesn’t mean that it is “high quality.” Furthermore, an unrated fan is not necessarily less protected against dust—it just wasn’t tested and certified to an IP rating.  

 

 

Dust Filter

For particularly dusty environments, you should include dust filters on computer fans. Dust filters keep insulating dust out of the system and extend the lifetime of fans by keeping them clean. The filters, however, reduce the total cfm of the fan and that may warrant using a fan with a higher cfm rating. Consider also adding dust filter material to the inlet holes of the system. It is critical that you clean dust filters regularly. A clogged dust filter can hinder a system’s cooling more than it helps.  

 

Warranty

The best quality fans are made by companies that stand by their products. Look for fans that have more than a two-year warranty. Some fans come with a warranty of five years or more. It’s even better if the company provides a mean-time-to-failure number to give an estimated interval before replacement. This value is usually given in hours.  

 

 

Maintain Accuracy Through Monitoring and Fan Control

Up to this point, you’ve focused on maximizing airflow and cooling in a computer system.  That’s great for maximizing the life of the system. But if you’re interested in regulating the temperature, you need to use a different strategy. Temperature regulation is particularly important when trying to maintain the accuracy of a DAQ board. If the ambient temperature of the DAQ board is kept stable, the measurement is less susceptible to outside ambient temperature changes. These effects are quantified in the accuracy table in the DAQ specification document, typically as offset and gain temperature coefficients.  

The general method of regulating temperature is to monitor the temperature near the sensitive devices (DAQ board, for example) and to regulate the speed of fans that are mounted in the computer case.  You can choose from several temperature monitors and controllers. Find one that can measure temperature accurately, can control various type of fans, and has user-friendly software to tune your system. A particularly good temperature regulation system is the aquaero from Aqua Computer. It has eight temperature channels and can control four individual fans. The aquaero also includes a powerful, easy-to-use software environment to set temperature thresholds that are used to control fans.  

The image below shows the aquaero software environment that is set up to measure temperatures at the CPU, RAM, hard drive, and PCI Express DAQ board. This environment is configured to regulate the speed of “Fan 1” to keep the PCI Express DAQ board at 40 °C. Monitoring the DAQ board’s onboard temperature sensor showed that this method kept the board temperature regulated within ±0.25 °C. This is a significant performance boost considering that without intelligent cooling the board’s temperature varied by as much as 5 °C, depending on the intensity of the task being performed. Tighter temperature regulation translates to lower thermal drift and better accuracy of your DAQ device.

 

 

 

Design for the Longest Life and Best Performance

Get the longest life out of your PCI Express DAQ board and computer system by designing in some form of active cooling. Consider the high-power components of your system and choose quality fans to generate the optimal airflow needed for cooling those components. To get the absolute best performance out of your DAQ board, implement a monitoring system to regulate the local ambient temperature of the DAQ device for applications that require low thermal drift.