By: Thomas S. Tarter, Principal Member of the Technical Staff, xMEMS Labs, Inc.
MEMS-based miniature-fan technology can provide improved thermal performance, increased reliability, and reduced noise for advanced processors.
Compact mobile devices are desperate for more efficient cooling systems. Already, 5G connectivity, spatial computing, high-resolution video, and gaming are testing the limits of current cooling solutions in thin, mobile, consumer electronics such as tablets and smartphones. Now companies like Apple, Google, and Samsung are in a race to integrate generative AI functionality into their smartphones. This integration includes processing large language models directly on devices rather than in the cloud.
There are many reasons for device makers to engage with edge AI by allowing chatbots, voice assistants, text-to-image generators, and other generative AI applications to run on devices:
- It reduces latency between a device and the cloud.
- It gives users a measure of data privacy, since their information isnโt always traversing networks.
- It allows the use of AI functions when connections are spotty or nonexistent.
- It takes some of the AI processing burden off large data centers and distributes it to the devices.
Temperature management in compact devices has traditionally been limited to the use of high thermal-conductivity materials and heat-spreading through the use of vapor chambers and heat pipes. Due to the relative inefficiency of these methods, throttling various components by lowering clock speeds has become common. This undermines the development of new processing units, along with other components needed to achieve the performance desired by end-users.
One way or another, manufacturers must find new ways of cooling mobile devices that donโt hamper the user experience. In this article, weโll discuss a new solution that aims to solve this problem by using solid-state, silicon-based, miniature fan technology.
The Challenges of Heat Management
When fifth-generation Intel and AMD processors emerged in the early 1990s, they were a sharp departure from their predecessors. They had far faster cores, tighter integration, and higher power. To keep the chips within operating conditions that would allow maximum clock speeds and high performance, additional steps had to be taken.
Initially, passive heat sinks were provided, either with the boxed processors or as a kit. As total power increased, however, it quickly became apparent that passive methods couldnโt provide the necessary level of cooling for these new, high-performance systems.
For almost all modern PC systems, passive solutions have given way to active solutions. These include fans, radiators, and liquid cooling systems. However, the addition of axial fans to the heat sink kit raised concerns related to lifetime, reliability, vibration, noise, and contamination.
For compact mobile systems such as phones, tablets, and ultra-thin computers, the problem of managing heat is exacerbated by the lack of space for active cooling solutions. Add this constraint to the demand for more local computing power, and the problem becomes even more significant.
The Solution: Micro-Cooling Fans on a Chip
Engineers at xMEMS created the first solid-state MEMS speakers for true wireless earbuds and other personal audio devices. Now, theyโve derived a micro-cooling fan on a chip based on the same process and fundamental design. This chip-level approach represents the first new approach to active thermal management of compact systems to emerge in over a century.
At the heart of this development is piezo-MEMS technology, which involves depositing microscopically thin piezoelectric film to silicon as a step in the wafer manufacturing process. This converts electrical energy into mechanical energy. In the case of micro-speakers, piezo-MEMS generates acoustic sound waves. For micro-cooling fans, it produces pulses of air that actively cool high-performance, mobile processors.
Capabilities of the Micro-Cooling Chip
The companyโs first micro-cooling chip, the XMC-2400, is shown below.
The XMC-2400 is just 1 mm thick. Thatโs 90% smaller than non-silicon active-cooling alternatives. Furthermore, as a solid-state fan, it makes no noise.
The airflow can be reversed by a simple I2C command to change the polarity of the MEMS. As a result, it can be integrated in either of the following ways:
- Alongside other microprocessors and electronics.
- Inside advanced 3D stacked system-on-chip (SoC) packages, where multiple chiplets are stacked to save space.
In the latter case, the micro-cooling chips can be placed either inside the package or on top of a stack to reduce temperature local to the silicon. Vents can be fashioned into the package to provide incoming cool air and to expel hot air through tiny ducts. The product is paired with an ASIC that delivers signals to the MEMS. The host system supplies 3V and I2C commands to control the volume and direction of airflow.
The fan-on-a-chip can be surface-mounted to rigid or flexible circuit boards using the same pick-and-place and reflow profile as typical electronic packaging. Flex connectors can be used to route signals to the chip from the end-use PCB, while rigid boards would have pads for electrical connections. In the case of extreme thickness requirements, a very thin harness could be developed.
Multiple micro-cooling fans can be arrayed in parallel to increase airflow volume or stacked together to increase static pressure capability. In this way, they can achieve scalable, micro airflow for meeting different processing demands.
Micro-Cooling Opens Up New Thermal Management Frontiers
Micro-cooling comes at an important time, when managing the heat generated by thin, ultra mobile devices presents a massive challenge for manufacturers and consumers. More devices are beginning to run more heat-generating AI applications, and because devices are so small and thin (and getting thinner), cooling them has only gotten harder.
The primary advantage of a micro-cooling solution is its direct application for thermal management of small spaces combined with the reliability and scalability of silicon manufacturing. Because itโs solid-state, itโs also unconstrained by noise. Micro-cooling chips can even find a home in comparatively larger mobile electronics such as laptops and AR/VR goggles, where traditional cooling fans take up excessive space and run noisily enough for detectable vibrations.
Ultimately, the goal of integrating micro-cooling into mobile and compact devices is to reduce or eliminate the need for throttling and the use of exotic materials. Micro-cooling can be used in places where traditional solutions are limited, creating opportunities for more powerful and compact system designs.