The chip industry is vigorously exploring the "core" technology - die technology - to improve chip performance from the perspective of advanced packaging. Intel, TSMC and other giants jointly established the UCIe Alliance based on "chip" technology.
Domestic manufacturers have also broken through 5nm "core particle" technology, winning a 5-year contract from American chip giant ADM. So what is the industry's exploration of this core technology? Will pellet technology be the future?
Manufacturing a chip requires many process steps, from wafer manufacturing to chip packaging, each link requires a complete industrial chain.
Based on these industrial chain technologies, the chip process has developed to the level of 4nm and 3nm. Although it has made a big breakthrough in performance, it also faces the physical limitations of Moore's Law. If the lithography machine cannot break through to the next level, does this represent the end of the chip manufacturing process?
GPU giant Nvidia stressed that Moore's Law is outdated and the idea of chips cutting prices is a thing of the past.
As a result, industry giants are always looking for other solutions to improve the performance of chips while keeping chip prices stable. Looking at the large and small chip technology and process path, die technology may be a good direction.
First of all, it is important to understand that so-called die technology is actually an advanced packaging technology that connects two chips together and then combines them into a new system-on-chip.
US chip giant Intel is trying to establish an industry standard for die technology, and formed the UCIe alliance with giants such as TSMC and Samsung to establish an open standard for high-speed interconnection between chip technology.
Intel is launching system-on-chip foundry services to manufacture chips for customers, including core chip foundries. This is only the tip of the iceberg of the industry's exploration of core technologies, and domestic manufacturers are also constantly exploring and making significant progress.
According to industry news, the domestic manufacturer Tongfu Micropower has achieved a technological breakthrough of 5nm in the core technology and completed the practice, and is expected to be shipped in bulk at the end of this year. It has also signed a five-year contract with microchip giant AMD, and the cooperation between the two parties will last at least until 2026.
Tongfu Microelectronics is a semiconductor company engaged in the field of packaging and testing in China, which provides advanced packaging and testing technology services. AMD is one of the customers of Tongfu Microelectric. The chip technology requires advanced packaging technology, so it is in line with the business direction of Tongfu Microelectric.
In terms of core technology, there are some views in the industry that the use of advanced packaging technology can bypass the rules of advanced manufacturing processes, or continue Moore's Law through core technology.
Of course, some people suspect that the core technology is inseparable from the advanced manufacturing process, because the core technology needs to be connected on the basis of two chips, so the production and manufacturing of these two chips also needs the support of traditional semiconductor industry chain technology.
Core technology is the technology of the packaging industry, and the wafer manufacturing step is before packaging, no matter how strong the theoretical performance of the die technology, it is impossible to skip the wafer manufacturing step. Therefore, it is normal to have these doubts, the core technology has not been widely promoted, and the relevant standards are still being formulated.
Perhaps only when the pellet technology becomes a mature technology in the industry can the answer that the outside world cares about.
Moore's Law has been around for decades, and now more and more people are talking about the limitations of Moore's Law, suggesting that the benefits of Moore's Law will continue to diminish in the pursuit of higher chip performance.
According to the concept of Moore's Law, the density of transistors in integrated circuits can be continuously increased even without end. However, due to the physical limitations of chip size, material equipment and lithography, chip manufacturing processes are slowing down. From the original annual generation process to one generation process, the process is divided into several process versions.
In the face of these problems, is pellet technology the future? In fact, pellet technology is conceptually destined for its broad prospects.
Two chips are stacked like building blocks, combining chips of different sizes, different processes and even different functions to form a complete chip. Imagine that incorporating a new chip not only improves performance, but also perfectly runs different functional modules.
In the post-Moore era, there will be more and more explorations of new processes, new technologies, and new architectures, and the core of advanced packaging has confirmed the feasibility of industrialization.
Apple's MI Ultra chip is the epitome of chip technology, and TSMC helped Apple assemble two M1 MAX chips, greatly improving the performance and quality of transistors. If 3D advanced packaging technology is promoted in the future, the combination of upper and lower chips can save the area used for 2D packaging rollout.
Either way, it's always a good thing to explore new chip technologies. The progress of human chips is achieved in the exploration of breakthroughs, otherwise how can important chip manufacturing equipment such as lithography machines and etching machines be born, and how can chip manufacturing processes below 5nm be developed.
The domestic core technology broke through 5nm, led by Tongfu Microelectric, completed the research and development and verification of related technologies, and signed a 5-year contract with the American giant AMD. It is believed that this is only the beginning, and domestic manufacturers will make greater breakthroughs in the development of core technology.