Tech

What is a chiplet? – Term explained






You may have heard before that your computer’s processor is also known as its “brain”. Similar to the various lobes of the brain, modern processors consist of several chips, the so-called chiplets, and no longer just a single monolithic chip. So what are chiplets and why are they so common now?

A chiplet is part of a processor module that makes up a larger integrated circuit like a computer processor. Instead of building a CPU on a single piece of silicon with the desired number of cores, chiplets allow multiple smaller chips to be used. This allows manufacturers such as AMD and Intel to create a larger integrated circuit.

Multiple chips that work together in a single integrated circuit are known as multi-chip modules (MCMs). AMD’s Ryzen, Threadripper and Epyc CPUs, which are based on the company’s Zen architecture, are examples of marketable products that incorporate chip sets. Furthermore, chiplets only depend on one I / O controller chip, which brings everything together in a single, uniform, integrated circuit.

Why are chiplets necessary at all? Moore’s Law states that the number of transistors in an integrated silicon circuit doubles approximately every two years. This rule of observation was named after the co-founder of Fairchild Semiconductor, Gordon Moore, who would later become CEO of Intel. The forecast dates from 1965 and was valid for around 50 years. However, due to the limited possibilities of silicon, the development of semiconductors slowed down in 2010 and experts now expect Moore’s law to be obsolete by 2025. This has led semiconductor manufacturers to look to materials such as gallium nitride to completely replace silicon.

Thanks to the chiplet design, manufacturers like AMD can develop increasingly complex processors. They combine several interchangeable components in a chip housing that would otherwise have to be placed on a single silicon die.

Enlarge

Thanks to the chiplet design, manufacturers like AMD can develop increasingly complex processors. They combine several interchangeable components in a chip housing that would otherwise have to be placed on a single silicon die.

© AMD

And since it is more and more difficult to fit more transistors on a piece of silicon, the yield drops as the limits of silicon pose further problems for manufacturers.

Chiplets are a solution to this problem. Semiconductors are known to be difficult to manufacture because processors are traditionally built on a single piece of silicon in what is known as a “monolithic” design. Small defects mean that chips are downgraded and sold with fewer cores or even thrown away entirely. In addition, a single defective chiplet can be replaced with another, resulting in less waste than if a much larger chip is discarded or downgraded. This increases yield because manufacturers can build multiple chips into a single processor to achieve the desired number of cores. And a higher yield also means more chips. Manufacturers can better meet their production goals with chiplets because there should be less scrap compared to conventional monolithic designs.

Tip:

CPU terms explained

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button