Before the machine brother in a well-known Q&A community, he would often brush up on such a question.
Presumably, what would happen if someone could copy a high-end CPU?
Such a strange problem, the heat is not low.
But the fact is that the two words of copycat have never been related to high-end chips.
After all, the technology behind the production of a chip is more complex than imaginable.
Take the most representative 7nm process now.
Etching alone requires 150 steps, and even more than 1500 steps if it is a complete process.
If you only do one step a day, it will take more than 4 years.
Don't say what a small team of cottages, even if it is a large enterprise, it is not said that it can be built.
However, some people did not believe in evil, and even made their own lithography machine in the garage and created a real chip.
Moreover, there is no "money ability" at all, the equipment is bought by the second hand to change the magic, and the process is to check the data and ponder.
Can this also make chips?
That's right, the next brother will come to tell you what it is.
First, let's introduce the protagonist, whose name is Sam Zeloof.
(Zelov Channel avatar)
He is only 22 years old and is currently a student at Carnegie Mellon University.
As early as high school, he had made a hand-built amplifier chip Z1.
The chip process reaches 175 microns.
What concept, the machine brother to help you convert it, in fact, it is 17500 nanometers.
It sounds like a bit of a crotch pull, which is equivalent to the width of two strands of hair.
Although the entire chip has only 6 transistors, the entire production step also takes 66 steps, and the whole process takes 12 hours.
According to Zelov, the yield rate of the chip is as high as 80%, but the actual situation depends on his coffee intake that day.
Of course, the chip production method is a bit wild, but it is not a good thing.
Zelov used it to make a simple LED flash circuit, and it can also be used as a distortion effect for electric guitars.
Probably because of process or design problems, the working voltage of this chip is relatively high, and it is not very good to drive.
So he had the idea of making the next chip, Z2.
The first problem facing the new chip is the venue.
In chip production, the workshop has high cleanliness requirements.
For example, the 1000-level dust-free workshop requires that the diameter of 5 microns in the space per cubic meter cannot exceed 29 particles.
Yes, it's that strict, and Zelov has only an "ordinary" garage.
However, looking back at the earliest chip production, the ancient gods also made chips on ordinary workbenches.
It's not impossible to make it work.
("Normal" garage with electron microscope)
Next is the problem of lithography machines.
If you follow the current market conditions, a lithography machine is less than 150 million US dollars, about 1 billion yuan or so.
Equally tricky is photoresist.
Even if you do have money, these two things are unlikely to be sold to individual buyers.
Zelov's solution made the machine brother sigh "can still be like this".
He made his own lithography machine, and the light source is a second-hand projector.
The magic place is coming, and the projection can be gathered into a small area through a light microscope.
Speaking of people, that is, the microscope is used in reverse, and it changes from enlargement to reduction.
And because it uses a projector, the image exposed by the lithography machine can be directly output by the computer.
Therefore, Zelov's chip design drawing was actually drawn with PS.
The photoresist used for exposure was dispensed by himself by studying "ancient" data.
About 91.475% of the people do not know much about the chip manufacturing process.
So here is a simple science popularization for everyone.
Photoresist is a material applied to the surface of a silicon wafer that changes when illuminated by light or rays.
Either it dissolves after being irradiated, or it solidifies after being irradiated, just like the yin and yang carvings in the carving.
In this way, it is easier for everyone to understand "lithography".
Based on this principle, it is possible to treat parts that are not covered by photoresist.
For example, operations such as etching and ion implantation with chemical agents are used to construct transistors.
No matter how complex the chip is, it is also built step by step with such a method.
The principle is such a principle, the actual operation is not so.
After all, even the wafers are cut by hand with a glass knife, and many steps are also made of magic equipment, and the yield rate is really difficult to guarantee.
The Z2 chip designed by Zelov contains 12 identical circuits, each containing 100 transistors.
So each chip has 1200 transistors, and the process process is about 10 microns.
The overall design is very close to Intel's first chip, the 4004, which has a transistor count of 2200.
However, judging from the growth rate of the number of transistors, Zelov even broke Moore's Law and simply beat the toothpaste factory.
This 200-fold transistor growth, Qualcomm, Intel do not feel ashamed to see it?
Next, do a simple math problem with the machine brother.
Assuming that the yield rate of 1 transistor is 99.9%, how high is the chip yield of 1200 transistors?
The answer is 30%.
So, in the end, Zelov made 15 Z2 chips, and only 1 piece could achieve 100% functionality.
The specific function, similar to a NAND flash memory chip, is of course much more rudimentary than what we use in mobile phones.
There are still 2 blocks that can only achieve 80% of the function, and the rest are basically scrapped.
After testing, some of the incomplete chips were mixed with dust.
However, for a project that costs only $20,000 (about 126,000 yuan), it is a miracle to be able to build a usable chip.
So everyone can better appreciate how difficult it is to make chips.
The machine brother looking for two data to compare, it will be more exaggerated.
The number of crystals in the Kirin 9000 is 15.3 billion, and the Apple A15 chip is also around 15 billion.
That is, 12.5 million times that of the previous Z2 chip, it is worthy of the peak of human technology.
To be honest, Zelov's wave of garage homemade chips is actually very rudimentary.
To put it mildly, it also has no practical significance.
It is nothing more than using relatively modern equipment to re-walk the old path of chip manufacturing 50 years ago.
However, the machine brother is going to say but again.
Such DIY has some popular science and educational significance.
From the perspective of the machine brother, the original chip manufacturing is a particularly advanced technology in everyone's impression.
However, after the dismantling and tossing of these folk gods, you will find that even the most cutting-edge things start from the most basic.
It is enough to arouse the interest of some people.
In fact, Zelov's garage chip was first inspired by another big guy who made his own chip.
These seemingly meaningless tosses can have some effect in some wonderful dimension.
Some things are really more important than the outcome.
But the brother feels that some things are as important as the results, such as praising the article after reading it.