From 5nm + process chat semiconductor process process dragon and tiger fight (on)
With the advent of September, the annual spring party of the technology session - Apple's fall conference will also arrive as scheduled. Unlike previous years, Apple's new product secrecy in the work of a tight rope, these years with Apple's global layout of the industry chain, the difficulty and operability of product secrecy is nearly imbalanced, almost every year the new Apple products will be exposed in advance, and this year is even more so.
But even so, the global technology users, still on the arrival of new Apple products, full of expectations. This year, the so-called "thirteen" iPhone series is early, from the internal structure to the appearance of the structure of the full exposure. Among them, the A15 bionic chip, which represents the core competitiveness of Apple's cell phones, is undoubtedly the focus of the industry's most attention.
According to supply chain news, iPhone13 (name to be determined) series built-in new A15 bionic chip, using TSMC's latest 5nm + process (N5P), is TSMC's current mass production of the most advanced process technology, Apple is also the first heavyweight customers under the process, and last year's iPhone12 consistent, Apple prepared in advance for this more than 100 million 5nm + process order quantity, used to The A15 bionic chip is needed for the full production of the new iPhone.
So the question arises, TSMC 5nm + process process in the end what is superior? What is the process process in chip manufacturing? Today, we'll talk about semiconductor process from the new iPhone's 5nm+ process.
What is the process?
Before understanding the process, we need to understand the working principle of the chip, that is, the use of semiconductor PN junction principle of unidirectional conductivity, and the use of multiple transistors in series "with", "or", "not " constitute a logic gate, the electrical signal into 0, 1 digital signal, to achieve signal transmission, so that the final realization of a single current switch to drive hundreds of millions of transistors to work, the output of 0 and 1 digital signal, into binary for calculation, storage.
Here, due to space constraints, we will not expand on the semiconductor PN junction, just know that it is derived from a single silicon and has a unidirectional conductive properties.
The first computer in the world, ENIAC, was composed of more than 17,468 electron tubes, 60,000 resistors, 10,000 capacitors and 6,000 switches, and was able to run only 5,000 addition operations per second.
But later, with the introduction of transistors, IBM launched the IBM 7090 all-transistor mainframe, and the computing performance got a leap, with a speed of 229,000 operations per second, which shows the importance of transistors for the modern semiconductor ecology.
After years of process evolution, the transistor has gone from a rarity requiring a glass casing for protection in the early days to the most common and indispensable key device for modern semiconductor processes.
The internal structure of a transistor is mainly composed of the source, the drain and the gate located between them, where the gate length is what we often refer to as the process size, or the process process process.
Process evolution direction
According to the working principle of the chip, we can actually know that the response speed of the chip depends on the charge flow speed between two stages inside a single transistor. Length.
These two directions constitute the main extension of the process of the major semiconductor companies in the world today, i.e., to continuously increase the number of transistors inside the chip, i.e., the density per unit space, and to continuously shorten the Gate Length per unit transistor to accelerate the charge flow.
Of course, it is worth noting that the saying "process = gate length", that process is equivalent to physical size, has actually been invalidated since 350nm. That's because, as the process advances, the physical size of Gate Length is no longer a rule, but some industry practitioners (yes, Samsung and TSMC) have named subsequent process steps in multiples of 0.7 in order to continue the previous rule of shrinking by about 0.7 every other generation.
For example, the next generation of 28nm process, i.e. 28*0.7 is equal to 22, so the next generation of process is named 22nm, and the next 14nm, even 10nm, 7nm are named according to this rule.
Transistor density is the key
So since the subsequent naming and Gate Length physical size does not have any connection, how is the industry to determine the difference between the different manufacturers of the process band?
That is the transistor density inside the chip, by comparing how much the transistor density inside the chip, you can determine whether it belongs to the advanced process or the same process.
Here, we can take the example of Intel 10nm process, according to public information, Intel 10nm process, although the name is not as loud as TSMC 7nm process, Samsung 7nm process, but in terms of transistor density, Intel 10nm far more than TSMC 7nm process (DUV) and Samsung 7nm (DUV), second only to the use of EUV lithography developed by TSMC 7nm+ process.
This can be seen, Intel 10nm and the same period TSMC, Samsung and other manufacturers process process belongs to the same level, but in the name of the non-aggressive, resulting in many users laughing about "squeeze toothpaste".
About TSMC N5+ process
Understanding the process, back to see TSMC this time for the new iPhone 5nm +, in the end what is the sacred.
In fact, compared to the hot iPhone parameters, about TSMC 5nm + process process news, revealed not much, we can try to get a glimpse of the TSMC 5nm process.
Previously, TSMC President Wei Zhejia said at a technical forum, compared to the previous generation of 7nm, 5nm process speed increased by nearly 15%, power consumption was reduced by 30%, transistor density increased by 80%, properly a new generation of process, and the process was also first used on top of the iPhone 12, according to official data, using TSMC N5 process A14 bionic chip, built-in 11.8 According to official data, the A14 bionic chip using TSMC's N5 process has 11.8 billion transistors, with nearly 3 billion more transistors, while CPU performance is increased by 40%, and GPU is increased by nearly 50%.
As for the 5nm+ process, sources speculate that it will bring an additional 5% speed boost and 10% power boost on top of the 5nm process.
As for the specific enhancements, let's wait and see.
Talking about this, the first half of the semiconductor process LW is over. In the next part, we will talk in detail about the major semiconductor makers, the current stage of the process, as well as the most promising process evolution around FinFet process in the future.