Crafting Cooler Chips

The advent of AI, self-driving cars, the Internet of Things (IoT), and other burgeoning fields are changing the world as we know it. But for potential to become reality, advancements in computing have to keep pace.
 
Meeting this challenge explains why microchips are in a state of perpetual evolution, and why we need new materials and techniques to push the boundaries of what’s possible.

To handle the data explosion, device architectures are moving into the 3rd dimension in some way or form.

This 3D approach to design, where components are stacked on top of each other, offers several benefits. It means more computing resources on a smaller footprint that function more efficiently and use less power.

Our teams are constantly creating new materials and technologies to ensure tomorrow’s demands can be met. We’re developing directed self-assembly techniques to create smaller, faster, more efficient microchips [1]; and we’re tackling obstacles in manufacturing 3D NAND flash memory [2] as data storage needs continue to grow exponentially.

But as components get smaller, taller, and more densely packed, dealing with the heat they produce becomes ever more important, and troublesome. It’s already changing how we think about chip design. Where one chip used to suffice, we’re now using multiple chips purely to address the heat management problem.

Cooling is essential to keeping processors at peak performance, and ensuring memory remains fast and reliable. But it’s also costly and inefficient.

One promising avenue to improving cooling efficiency is to shrink the scale. By using tiny microchannels, you can pump cooling fluid very close to or, in some cases, straight through chips. But these aftermarket ideas present new challenges, whether that be issues with complexity and cost, fluid pressure, or irregular heat management [3].

But a team of researchers have come up with an innovative solution that has none of these problems. They have developed a method of cutting these three-dimensional microchannels into the very silicon wafers that form the backbone of countless electronic components [4]. In this way, they are addressing cooling at the point of manufacture, and what’s more, it can be done with production processes that are already in place [4].

Further work is needed to turn the proof of concept into something mass-market and commercially viable, but it’s a promising route that could revolutionize cooling — making it vastly more efficient and, as a result, significantly reducing the amount of wastewater cooling systems inevitably produce.

Currently, cooling accounts for upward of 30% of the total energy used by data centers. The researchers predict that at that scale, their approach could slash that figure to less than 0.01%.

Very cool.

_References:

^{[1] https://www.emdgroup.com/en/research/science-space/envisioning-tomorrow/smarter-connected-world/dsa.html
[2] https://www.emdgroup.com/en/research/science-space/envisioning-tomorrow/smarter-connected-world/3d.html
[3] https://www.nature.com/articles/d41586-020-02503-1
[4] https://www.nature.com/articles/s41586-020-2666-1}