Chip Designers: Pushing Moore's Law To The Limit

Moore's Law states that the number of transistors in a chip doubles every two years or so. However, this law is becoming increasingly difficult and expensive to follow, as power and performance benefits diminish. In response, chip designers are finding innovative ways to increase device performance and speed. One method is to stack microchips on top of each other, allowing for more powerful devices. This approach, known as sprawl, is a departure from the traditional flat design of silicon chips. Another strategy is to focus on specialized computing, which involves transforming specific software tasks into physical silicon chips, rather than relying on a single central processing unit. These techniques are being used to drive advancements in artificial intelligence and self-driving cars.

Microprocessors are reaching their physical limitations

With five-nanometre transistors currently being mass-produced, microchips are getting very close to their point of absolute miniaturisation. IBM has developed a two-nanometre chip that fits 50 billion transistors, each the size of roughly five atoms. There are plans to introduce one-nanometre semiconductors in the 2030s, which will allow for unprecedented speed.

To continue improving chip technology, manufacturers are turning to new methods and technologies. Extreme ultraviolet lithography and atomic layer deposition, for example, allow for greater control over the placement and size of components on a microchip, leading to increased efficiency and performance. Another expected game-changer is three-dimensional (3D) chip stacking, where multiple layers of circuitry are stacked on top of each other, allowing for more components to be packed into a smaller space. This results in increased computing power and energy efficiency.

While these ultra-small microchips show exciting potential, the current cost and complexity of manufacturing them may pose challenges for large-scale production. However, chip designers are finding innovative ways to break through the limitations of Moore's Law and continue advancing chip technology.

Chip architecture is driving performance gains

Microprocessors have reached their physical limitations in terms of size, speed, and power efficiency. As a result, chip designers are turning to chip architecture to drive performance gains.

One way they are doing this is by stacking microchips on top of each other, rather than the traditional approach of flat tiles of silicon. This allows for more circuitry to be added for functions such as memory, power management, and graphics. This trend is driven by the need to make chips faster and devices more capable.

Another strategy is "specialized computing," which involves transforming specific software tasks into physical silicon chips instead of relying on a central processing unit (CPU) to perform all tasks. This approach has become the driving force behind many technological advancements, including artificial intelligence and self-driving cars.

Additionally, companies are now dealing with more options and unique challenges as they seek optimal solutions for their specific applications. They are demanding more from the Electronic Design Automation (EDA) ecosystem, which is racing to keep up with advancements in packaging, chiplets, and customized hardware and software.

The end of Dennard scaling and the flattening of Moore's Law have significantly shifted the dynamics for chip design and manufacturing. Architects are now more important than ever, as they need to figure out how to add more features, functions, performance, and cores to chips without increasing the energy footprint.

To address these challenges, chip designers are integrating AI into processor chips to reduce the energy required to access data. They are also exploring various approaches, such as multi-core design, ASIC revolution, and high-level synthesis, to enhance computational performance and specialization.

The industrywide strategy for improving microchips can be summed up as 'sprawl'

The industrywide strategy for improving microchips can be summed up in one word: sprawl. This strategy involves piling microchips on top of one another, akin to building multistory structures instead of the traditional pancake-flat tiles of silicon inside computers. This approach is driven by the relentless pressure to make chips faster and devices more capable, as the chip industry faces technical barriers in shrinking transistors to enhance performance.

The chips inside our most powerful devices are taking up significant real estate, challenging their "micro" designation. Engineers are addressing this challenge by stacking circuitry for functions like memory, power management, and graphics on top of each other. This chip "stacking" approach is a response to the slowing of Moore's Law, which states that the computing power of chips doubles every two years.

Moore's Law is facing headwinds due to factors such as high inflation in the post-COVID era, increasing process complexity, and the physical limitations of CPUs. As a result, chip designers are turning to specialized computing, where specific software tasks are transformed into physical silicon chips, rather than relying solely on CPUs.

The transition from the “Microchip Era” to the “Megachip Age” highlights the increasing scale and complexity of chip design. As chips sprawl and become more akin to multistory buildings, the industry is pushing the boundaries of what "micro" means in the context of microchips.

The race to improve microchip technology is not just a matter of economics and innovation but also has significant political implications. The manufacture and design of semiconductors are driving advancements in artificial intelligence, and governments and tech companies are vying for access to the most cutting-edge chips. This competition has led to restrictions on the sale of chip-making equipment and advanced chips to certain countries, highlighting the strategic importance of microchip technology in the modern world.

The pressure to make chips faster and devices more capable is unrelenting

One strategy employed by chip designers is "sprawl", where microchips inside powerful devices are stacked on top of each other, rather than spread out. This approach, similar to urban infill, allows for more powerful chips without increasing the amount of space they occupy. By piling microchips atop one another, designers can include additional circuitry for functions such as memory, power management, and graphics.

Another approach to breaking Moore's Law is through "specialized computing". Instead of relying solely on a central processing unit (CPU) to perform all tasks, specific software tasks are transformed into physical silicon chips. This strategy has become the driving force behind many recent technological advancements, including artificial intelligence and self-driving cars. By offloading these specialized tasks from the CPU, chip designers can enhance overall system performance.

The pursuit of faster chips and more capable devices is also leading to the integration of AI into processor chips. For example, IBM's Z Systems integrate AI to access and process data more efficiently, reducing the energy footprint of AI computations. This approach not only improves performance but also helps address the rising cost of energy and environmental concerns.

As the dynamics of chip design and manufacturing shift, companies are facing unique challenges and demanding more from the electronic design automation (EDA) ecosystem. They are seeking optimal solutions for their specific applications, including advanced packaging, chiplets, and customized hardware and software. The EDA community is racing to keep up with these changes and provide the necessary tools and technologies to support chip designers in their pursuit of faster, more capable devices.

Companies are demanding more from the EDA ecosystem

Moore's Law states that the achievable density per chip doubles every one to two years. However, as microprocessors reach their physical limitations, chip designers are having to find new ways to drive performance gains.

The pressure to make chips faster and devices more capable is constant, but the chip industry is facing technical barriers. One way in which engineers are overcoming these challenges is by stacking microchips on top of one another. This method of chip design is demanding more from the electronic design automation (EDA) ecosystem.

EDA is a market segment consisting of software, hardware, and services that assist in the definition, planning, design, implementation, verification, and subsequent manufacturing of semiconductor devices, or chips. EDA tools are used to design and validate the semiconductor manufacturing process, as well as to verify that a chip design will meet all the requirements of the manufacturing process.

The EDA ecosystem is critical to the chip design and manufacturing process. Companies are demanding more from EDA, as chips become increasingly complex and the cost of errors in manufactured chips remains high. EDA tools play a crucial role in ensuring that chips are designed and manufactured flawlessly, and in a cost-effective manner.

The EDA market is primarily a software business, with sophisticated software programs that assist in the design and manufacture of chips through simulation, design, and verification tools. EDA solutions are not directly involved in the manufacture of chips, but they are essential to ensuring the performance, density, and functionality of the final product.

Frequently asked questions