CHIPLETS Demystified

The semiconductor industry is betting on this emerging technology to hoodwink Moore's Law. Here are the basic details everyone should know about chiplets.

- JANANI G. VIKRAM is a freelance writer based in Chennai, who loves to write on emerging technologies and Indian culture.

What are chiplets?

• Chiplets are small, modular integrated circuits (ICs) with limited functions or features.

• Traditional monolithic design packs all the components into a single large piece of silicon. Chiplets separate the functionality into specialised dies, which are then assembled into a single package.

• Basically, take a system-on-chip (SoC) and break it down into its functional components—general-purpose processing, custom accelerators, data storage, high-speed input and output (I/O) communications. Now, make separate chips specialising in each of these tasks. You now get CPU chiplets, GPU chiplets, memory chiplets, and I/O chiplets, which you can put together like ‘Lego’ building blocks!

• Now what happens is that each component becomes a separate chip - which can scale to the limit of Moore’s Law. Collectively, they can surpass the limitation!

Why chiplets?

• Reusable. The same chiplets can be reused in various devices. This increases flexibility and reduces costs and time-to-market.

• Known good die. Since the chiplets are already tested and guaranteed to be error-free by their manufacturers, the yield and reliability of the final assembly are higher.

• Heterogeneous integration. You can get chiplets from different brands that specialise in specific functions, and package them together—provided they adhere to some basic standards. It is like putting together a team of experts.

•