Controlling thermal issues associated with the CPU is often a top priority for embedded system designers. However, memory modules are not necessarily less important. Thermal management issues present challenging design considerations in embedded environments that require knowledge, precision, and creativity to diagnose and overcome memory subsystem design parameters.

In the past, memory wasn’t as complex and didn’t have the kind of thermal attention that designers paid to CPUs. Since the CPU needs to be cooled, the chipset is equipped with a heatsink as a production standard. In contrast, memory modules only require fine-tuning of airflow to control temperature. But with the increased speed of DDR3 and DDR4 technology in today’s embedded designs, memory module designs have become complex and thermal issues need to be taken care of.

Clock speed is just one reason why memory is hotter than ever. Customer environment, overall design choices (such as memory modules), location on the board, horizontal or vertical module orientation, and airflow on the system can also affect the thermal profile of memory modules.

Embedded system designers often use compact board layouts and require near-perfect engineering to achieve perfect signal integrity and excellent performance. Despite other design concerns, successful system designers view memory thermal management as a higher-level design issue, keeping pace with evolving memory technology and thermal management techniques to reduce heat in memory modules.

Memory designers can use a range of simple but powerful cooling concepts to reduce heat and design better memory subsystems. Likewise, system designers can enhance products by incorporating these concepts when creating designs.

heater on

Memory designers start by selecting memory modules that reduce heat and provide the best overall cooling solution. Combining modules that use the least DRAM into the most module columns can achieve the desired module density and manage power efficiently. The more DRAM in standby mode, the less power the module consumes – usually achieved by using the DRAM with the widest data bus, as shown in Table 1. For example, a 36-chip four-rank x8 DIMM is better than a 36-chip two-rank x4 DIMM.

figure 1

As another example, a 512 MB Error Correcting Code DIMM can use five 64×16 DRAM chips instead of nine 64×8 DRAMs, reducing heat by 44%. Actual reduction may be slightly less due to the different IDD values ​​specified in the datasheet for 64×16 and 64×8 DRAM. Memory designers typically explore whether memory controller chipsets can support wider DRAM data bus widths.

Overall, memory modules that are properly spaced between DRAMs, whether unstacked or without large thermal semiconductors, will have better thermal characteristics. Small form factor memory (such as stacked ultra-thin memory or stacked SODIMM) has higher power density (watts/area) and requires special consideration for cooling. Fully buffered DIMMs also have high power density due to onboard advanced memory buffers and may require additional cooling aids or airflow.

system and memory

Thermal sensors are a critical tool for memory designers. JEDEC’s standard specifies that memory modules have thermal sensors that provide users with monitoring and triggering mechanisms to adjust system performance based on temperature fluctuations.

Depending on the defined parameters, the system can issue an extended mode register set command to double the internal refresh rate on DDR2 DRAM to 32 ms cycles (tREFI = 3.9 μs) at a trigger temperature of +85 ¬∞C to extend the DRAM operating temperature to +95 ¬∞C. If this feature is not available, designers can add special programming on the memory modules to extend temperature operation. Alternatively, the system can use closed-loop dynamic temperature throttling and fan speed control to optimize memory performance.

The key here is that the CPU manages the memory board’s thermal sensor, which suggests that system-level and board-level thermal issues are closely related. The system’s BIOS reads the sensor output and evaluates performance options based on preprogrammed thresholds that identify acceptable temperature ranges. For example, if memory exceeds the limit temperature, the system thermal monitor will alert administrators when the temperature exceeds a defined threshold, prompting them to take necessary steps to reduce the temperature, such as checking the processor and case fans, and addressing any problems with case vents . Possibly blocked, or add another case fan.

Airflow is important

Airflow is a simple but critical issue for memory; the main goal is to avoid blowing preheated air directly over the memory subsystem. Whenever possible, designers should place the memory subsystem on the side of the processor and out of the flow of hot air generated by the processor, heat sink, or other thermal components such as power supplies or chipsets. Ambient intake air should flow evenly through other thermal components such as the memory subsystem and processor.

Air gaps between modules that are too small can create airflow backpressure from physically blocked DIMM modules in the airflow path. This can cause a drop in airflow pressure on the sides of the DIMM, resulting in reduced airflow, or it can divert airflow around or around the entire memory subsystem. DIMM sockets should be 10mm or more on center.

Typically, maximizing airflow extracts heat from the memory. If noise is not an issue, designers should use blowers or dual fans to optimize airflow. Airflow with minimal pressure drop is best achieved by drawing hot air at the exhaust point, but can also be improved by pushing air at the intake point. A plenum, duct, or shroud can be used to direct and control airflow through the memory subsystem, parallel to the longest side and sides of the DIMM. These enclosures may allow for slower fan speeds and less noise without compromising airflow.

Memory modules can be designed to allow airflow over the short sides of the DIMM, thereby eliminating heat dragging from the long sides of the DIMM. This mezzanine connector technology does not expose as much DRAM to the preheated air from the upstream DRAM.

If the motherboard or system board is mounted flat and perpendicular to the line of gravity, the best orientation for the memory will be vertical as the hot air rises along the line of gravity. The vertical DIMM orientation prevents heat from being trapped under the bottom of the memory module. If vertical mounting is not an option, a diagonally mounted DIMM orientation would benefit from a single-sided DIMM with the DRAM components mounted on top. This also applies to memory DIMMs that lie flat on the system board.

Designers should choose a module with a DRAM layout that does not allow all DRAM devices to be active on the same side at the same time. Modules with alternating DRAM placement on each side of each rank of memory modules will spread the heat evenly around the DIMMs. If airflow is restricted on the DIMM side, memory modules that place only the DRAM on the side with the most airflow will perform better at higher temperatures. Figure 1 illustrates how the technique of alternating DRAM columns reduces thermal impact.

figure 2

radiator, etc.

A heat sink is a metal cover placed on the surface of a memory module to distribute heat evenly across the surface and to balance surface temperatures by eliminating localized hot spots. The heatsink is made of a thermally conductive material, such as copper or aluminum, in the shape of a clamshell that wraps around the memory module.

Heat sinks placed on the sides of the memory and/or the top edge of the memory module will maximize heat extraction from the module, if space permits. The extra surface area that the heat sink adds to the memory module without compromising airflow determines its overall efficiency.

Thermally conductive PCBs and PCB cores are also valid options. These metal or carbon composite laminate layers are embedded into the structure of the memory PCB, allowing it to operate at lower temperatures than standard FR-4. These layers also equalize component temperature by eliminating localized hot spots such as phase-locked loops. It is not uncommon for a large number of hot spots to be created through holes under the thermal device to conduct heat into the core. These cores, in turn, conduct heat into the edge fingers of the module and can be brought to the top edge of the PCB, exposing it to a heat sink or heat sink. The top edge of this type of PCB has the DIMM’s internal thermal core, which connects to an integrated heat sink on top of the module, increasing the height of the DIMM.

During the manufacturing process, the memory modules can be tested at high temperatures in the customer’s system running the customer’s diagnostic software. This active aging will screen out potentially weak modules. Passive burn-in (on an unpowered module) has no effect on screening DRAMs with weak cells because DRAM cells are semiconductor-based capacitors that need to be constantly charged or refreshed to preserve binary information. Some memory modules can use DRAM and have an operating temperature range of -40 ¬∞C ≤ Tcase ≤ +95 ¬∞C. This is a special product and not all DRAM suppliers offer industrial temperature DRAM as an option for commercial temperature (0 ¬∞C „Tcase „+85¬∞C).

Comprehensive Thermal Problem

Thermal management issues have evolved with memory technology and become critical to embedded systems, reliability, and performance. Design dynamics between system designers and memory subsystem designers are also evolving and can impact designs built for durability and performance. Trusted system- and board-level partnerships and a better understanding of current thermal concepts related to DRAM memory modules can lead to a successful end product.

Incorporating thermal considerations for DRAM memory modules as part of proven system design principles can give designers a new understanding of ways to improve thermal performance. General design considerations and alternative cooling options can create successful memory subsystem designs that effectively meet system requirements for high memory bandwidth, large memory density, small physical space, and low cost in embedded environments.

Reviewing Editor: Guo Ting

Leave a Reply

Your email address will not be published.