‘Many innovations were born at Samsung’s India R&D unit’

How has SSIR contributed to the innovations at Samsung?

Globally, Samsung Semiconductor comprises three major business units—memory, foundry, and system large-scale integration. Volatile memories such as static random access memory and dynamic random access memory need power to store data, but non-volatile memory (solid-state drive) does not need power to retain information in the chip. We have been building DRAMs for over 30 years. But we work more on non-volatile memory storage solution space such as SSD drives.

Please give us some examples.

Many innovations at Samsung were born at SSIR. One flagship product, where end-to-end ownership and development were driven out of SSIR, is our Exynos system-on-a-chip (integrated circuit design combining elements of electronic devices on a single chip) which is equipped with a 5G modem and neural network processing unit (for artificial intelligence applications) and powers mobile devices and cars. Our ISOCELL image sensors, too, use advanced pixel technology to power your phone and car cameras. At SSIR, we tune and qualify sensors, memory and storage devices to ensure that they function well under all conditions (severe heat, biting cold, rough weather). Further, sensors in many commercially-available phones globally have been designed at SSIR. Moreover, the world’s first 108MP camera was planned and created by the engineers at SSIR—from designing the specifications, hardware and final software development. SSIR is also among the first to adopt a more advanced transistor architecture, as we (at Samsung Foundry) continue to establish our leadership with the production of the 3nm (nanometre) process node applying Gate-All-Around (for improving performance, reduce power consumption and increase transistor density).

Have you filed any patents?

SSIR has filed more than 1,000 patents globally, and more than 450 have been granted so far.

Do you have quality engineering talent from India?

We have very well-trained talent in our campuses. When we recruit, we gauge their understanding of semiconductor device physics, digital design, and basics of software, along with their problem solving and analytical ability. We spend six months to one year for training them when they come from campuses. But our employees have to continuously improve while we have to upskill our people, especially given the rise in use of AI and ML.

In this context, how is SSIR leveraging AI and ML?

Among other things, we use AI-ML for our design activity—for example on how soon can I debug a problem? We can speed up learnings and techniques with both AI-ML instead of doing things manually. Second, most of our applications have neural processing units that help algorithms to run very fast.

SSIR is partnering Indian Institute of Science (IISc) to promote R&D in the field of on-chip electrostatic discharge (ESD) protection. What does it entail?

Collaborating with some academic institutions is very important for us to push innovation research to the next level. Practically every electronic system comprises integrated circuits and SOCs but they are very sensitive to ESD failure (release of stored static electricity can damage these sensitive equipment). IISc is one of the few institutes in the world leading ESD device research. The collaboration is aimed at helping us develop highly-reliable interfaces and SoCs that operate at low power and high speed. This is a multi-year project, and hopefully, it will open more opportunities for research with them.

Can you elaborate on the IoT opportunity for semiconductor firms with an insistence on smart cities, homes, and smart factories?

I’m no longer talking about just the Internet of Things but also the Intelligence of Things, because artificial general intelligence (AGI) is becoming a very important aspect, which leaves no room for dumb connected devices. You now want a lot of decisions to be made on the edge (closer to the devices rather than on a distant server), say, on your phones or automobiles.

But this also implies that you will need low latency and low power consumption for mission-critical applications like cars and medical equipment. Isn’t it?

We have a lot of innovations on that front, specifically for modems, not only for phones, but your refrigerators will also have a modem in the future. Every device will have a modem and some kind of a neural processing unit or AI-ML processing unit, which will make power management a very important requirement. Further, sensors might not be just for photographs but will also be used in many areas such as light and smell sensors, which may play a very big role.

What more should India do to become a global innovation hub and encourage ‘Make AI in India ?

India already has a big local (semiconductor) design ecosystem. As for AI, there has to be a lot of design models or design use cases. It’s not just about chip design but more on product design. We have to start from the top of the value chain and introduce products systematically for India.