The technology supremacy race between the U.S. and China reflected in the aggressive stance of America to restrict the export of semiconductor technology underscores its increasing importance in determining the new economic and military superpowers. Semiconductors are at the front and center in the U.S.-China tech supremacy race.
Mr. Anurag Bansal, Managing Director of 13D Research & Strategy, an expert on new technologies, their convergence to disrupt corporate paradigms, and how they are becoming essential to contemporary geopolitics, spoke with us in depth. For about 15 years, he has been studying the confluence of exponential technologies such as AI, DNA sequencing, gene editing, automation, and networking.
Anurag emphasized that Chiplet technology is a developing field of study that will assist semiconductors in maintaining their upward trend of getting more powerful. Physical constraints in the traditional semiconductor manufacturing process are driving this modular approach. Furthermore, nations such as China, which does not have the most sophisticated semiconductor manufacturing, may be able to develop comparatively advanced chips and compete with U.S. superiority. The fight for technological superiority is heating up and will shift the balance of power. The winners of the twenty-first century will be determined by who is first to create these technologies and, more significantly, the commercial applications that will use them. Semiconductors are one of the key battlegrounds between the United States and China, with tensions rising regularly. China is determined to become self-sufficient, and the United States appears to do everything necessary to prevent this.
Some excerpts from our conversation –
Semiconductors have become a mainstay of the modern-day economy. However, it is becoming difficult to continue the trend after a near-five-decade robust improvement in semiconductors cost and performance improvement.Â
Moore’s law, the observation that the number of transistors on a microchip double roughly every two years, has been the driving force behind the incredible advancements in computing technology over the past several decades. However, as transistors have shrunk to the nanometer scale, the traditional method of creating monolithic chips, where all components are manufactured on a single piece of silicon, has become increasingly challenging. The physical limitations of shrinking transistors, such as quantum tunneling and leakage current, have made it harder to continue scaling down. Additionally, the economic costs of manufacturing smaller transistors have become prohibitive, besides lower yield rates.
Chiplets are a promising solution to continue Moore’s law and could become a critical component in China’s quest to challenge America’s supremacy in chip-making. Chiplets are a new semiconductor industry paradigm rapidly changing how we think about chip design and manufacturing. Chiplets are small, self-contained modular components designed to be interconnected with other chiplets to create a complete system-on-chip (SoC) on a single piece of silicon. For example, a chiplet might contain a memory controller, while another chiplet might contain a set of processing cores that can be assembled to create a more extensive system. By breaking down complex SoCs into smaller, more manageable components, chiplets offer a range of benefits.
One of the critical advantages of chiplets is that they can be designed and manufactured separately, allowing each chiplet to be optimized for its specific function. For example, a chiplet that controls memory access can be optimized for low latency and high bandwidth. In contrast, a chiplet designed to perform complex computations can be optimized for high throughput and low power consumption.
Other benefits of chiplet technology include the following:
Design Flexibility – chiplets allow engineers to mix and match different components to create custom SoCs tailored to specific applications. This flexibility will enable designers to optimize the performance and power consumption of each chiplet for its particular function, resulting in a more efficient and optimized system.
Improved Yield Rates – chiplets are designed and manufactured separately. Therefore, defects or issues during the manufacturing process can be isolated to a single chiplet rather than affecting the entire SoC. This means defective chiplets can be discarded, while the rest of the SoC can still be salvaged, reducing waste and improving overall yield rates.
Faster Time-to-Market – chiplets modular design and manufacturing enables a quicker time-to-market than traditional SoCs. Engineers can design and test individual components separately and iterate and refine the chip designs without waiting for an entire SoC to be manufactured. This can save valuable time and resources, particularly in fast-moving industries like consumer electronics.
Lower Development Costs – chiplets can also help lower development costs by reducing the need for expensive custom SoC designs. With chiplets, engineers can mix and match pre-existing components to create custom SoCs rather than designing each element from scratch. This can significantly reduce the cost and time required to develop a new electronic device.
Chiplets are already being used in various applications across the semiconductor industry. One of the most prominent examples is the design of high-performance computing (HPC) systems, where chiplets are used to create custom processors that deliver exceptional performance for specific workloads. Chiplets are also being used in designing artificial intelligence (AI) and machine learning (ML) systems, where they can be combined to create custom neural network accelerators and other specialized components.
Chiplet development has increased over the last few years. Sapphire Rapids is a new generation of Intel Xeon processors based on chiplet technology and features enhanced artificial intelligence and cryptography capabilities. Similarly, Ponte Vecchio is a new chiplet-based GPU architecture developed by Intel designed explicitly for high-performance computing workloads. Apple’s M1 chip, which powers the latest generation of Mac computers, is also based on chiplet technology. TSMC has recently introduced a new technology called “system on integrated chips,” which employs 3D-stacking techniques to facilitate the reintegration of chiplets. The initial version of AMD’s EPYC processor utilizes a similar chiplet-based design.
China considers chiplet technology to be crucial in achieving semiconductor self-reliance. Chinese companies such as Huawei and Phytium are developing their chiplet-based processors. Recently, two companies in China demonstrated chiplet-based processors. Biren Technology recently unveiled China’s first general-purpose GPU chip, the BR100, manufactured using chiplet technology. Loongson leveraged chiplet technology to integrate two 16-core processors on a single silicon substrate, resulting in a 32-core processor with eight memory channels. Additionally, Arctic Xiongxin released the “Qiming 930” chip based on chiplet architecture using a 12 nm process and is designed for a range of applications, such as AI reasoning, privacy computing, and industrial intelligence.Â
Investors can position themselves to take advantage of the latest developments in this fast-moving field. The semiconductor industry’s major players driving the development of chiplet-based systems include Intel, AMD, Nvidia, and TSMC. These companies have significant resources and expertise in chip design and manufacturing and are well-positioned to take advantage of the growing demand for chiplet-based products.Â
Author: Anurag Bansal is a Disruption Analysts and serves as Managing Director at 13D Research & Strategy
Follow Anurag on
Instagram – boundlessanurag
Twitter – @boundlessanurag
**Disclaimer: Views expressed here are personal and do not represent any institution or entity I’m associated with.
