TSMC Unveils CoW-SoW: A 3D System-on-Wafer Platform for Stacked Chip Innovation

TSMC is revolutionizing the semiconductor industry with its wafer-scale integration technology, InFO-SoW, and plans to combine it with SoIC to create a new CoW-SoW platform, enabling tighter integration of logic and memory for enhanced computational capabilities and energy efficiency in AI and HPC applications.

Revolutionary wafer-scale processors with low-latency high-bandwidth core-to-core communications
Tighter integration of HBM4 memory with logic for enhanced performance and energy efficiency
Potential for future advancements in AI and HPC applications

TSMC, the leading semiconductor manufacturer, has been offering its System-on-Wafer integration technology, InFO-SoW, since 2020. While only Cerebras and Tesla have ventured into developing wafer scale processor designs using this technology, TSMC believes that the usage of wafer-scale designs will increase in the future. With emerging megatrends like AI and HPC demanding more complex solutions, TSMC envisions vertically stacked system-on-wafer designs as the way forward.

Tesla’s revolutionary wafer-scale processors, known as Dojo, are the first mass-produced solutions based on TSMC’s InFO-SoW technology. These processors offer several advantages over traditional system-in-packages (SiPs), including low-Latency high-bandwidth core-to-core communications, exceptional performance and bandwidth density, efficient power delivery, and redundancy.

However, one limitation of InFO-SoW and other wafer scale integration methods is the reliance on on-chip memory. While this suffices for many applications, it may not meet the demands of next-generation AI workloads. Additionally, InFO-SoW requires the entire wafer to be processed using a single fabrication technology, which may not be cost-effective or optimal for certain designs.

To address these challenges, TSMC plans to combine two of its packaging technologies, InFO-SoW and System on Integrated Chips (SoIC), in its next-generation system-on-wafer platform. This integration will enable the stacking of memory or logic on top of a system-on-wafer using TSMC’s Chip-on-Wafer (CoW) method. The CoW-SoW technology, recently announced at TSMC’s North American Technology Symposium, is expected to enter mass production by 2027.

Currently, TSMC is primarily focusing on integrating wafer scale processors with HBM4 memory. The tighter integration of HBM4 stacks, featuring a 2048-bit interface, with logic is an exciting prospect for the industry.

Kevin Zhang, Vice President of Business Development at TSMC, emphasized the significance of wafer level integrations for future advancements. He stated, “Using wafer level integrations, our customers can integrate even more logic and memory together. SoW is no longer just a concept; we are already working with our customers to produce products using this technology. By leveraging our advanced wafer level integration technology, we can empower our customers to enhance their computational capabilities and achieve greater energy efficiency in their AI clusters or supercomputers.”

As TSMC continues to push the boundaries of semiconductor technology, their upcoming advancements in wafer-scale integration hold immense potential for the future of AI and HPC applications.

(Source)

Background Information

About TSMC: TSMC, or Taiwan Semiconductor Manufacturing Company, is a semiconductor foundry based in Taiwan. Established in 1987, TSMC is a important player in the global semiconductor industry, specializing in the manufacturing of semiconductor wafers for a wide range of clients, including technology companies and chip designers. The company is known for its semiconductor fabrication processes and plays a critical role in advancing semiconductor technology worldwide.

Technology Explained

HPC: HPC, or High Performance Computing, is a type of technology that allows computers to perform complex calculations and process large amounts of data at incredibly high speeds. This is achieved through the use of specialized hardware and software, such as supercomputers and parallel processing techniques. In the computer industry, HPC has a wide range of applications, from weather forecasting and scientific research to financial modeling and artificial intelligence. It enables researchers and businesses to tackle complex problems and analyze vast amounts of data in a fraction of the time it would take with traditional computing methods. HPC has revolutionized the way we approach data analysis and has opened up new possibilities for innovation and discovery in various fields.

Latency: Technology latency is the time it takes for a computer system to respond to a request. It is an important factor in the performance of computer systems, as it affects the speed and efficiency of data processing. In the computer industry, latency is a major factor in the performance of computer networks, storage systems, and other computer systems. Low latency is essential for applications that require fast response times, such as online gaming, streaming media, and real-time data processing. High latency can cause delays in data processing, resulting in slow response times and poor performance. To reduce latency, computer systems use various techniques such as caching, load balancing, and parallel processing. By reducing latency, computer systems can provide faster response times and improved performance.