Speaking out, you may not believe that the CPU wafer is the round wafer that is the core material of the CPU. Early computers were very huge. The first generation computers covered an area of ​​170 square meters and weighed 30 tons. The reduction is entirely due to the invention of semiconductors (the specific principles are encyclopediced by myself). There are quite a lot of semiconducting materials on the earth, and silicon is the semiconductor with the most reserves, the most mature refining technology, and the most stable performance, so it has naturally become the material for CPU. The refining technology of using silicon to make CPU is to provide silicon into high-concentration single crystal silicon, which is then rotated and pulled out from a high-temperature container to become a single crystal silicon ingot. Because the crystal is pulled out when it is pulled out by rotation, it is round. The wafer is the material for making the CPU. The crystal column that is pulled out is cylindrical, and then cut into very thin slices, which are the so-called silicon wafers. The size of the wafer is basically limited by its own characteristics. In the wafer production process, the farther away from the center of the wafer, the more likely it is to have dead pixels. Therefore, expanding from the center of the silicon wafer, the number of dead pixels is on the rise, so that we cannot increase the wafer size as we wish. After the size of the wafer is limited, the etching size becomes the key technology for making the CPU. The smaller the etching size, the more transistors can be engraved. Therefore, the smaller the CPU process, the better the performance of the CPU. Strong. (This is why the photolithography function becomes the key to the performance of the semiconductor chip) The crystal is etched on the wafer in advance, and then cut and processed into a CPU. Don’t underestimate a small square chip on the wafer. , There are billions of transistors on it. In the picture above, you can see that each chip is arranged in a square. In the picture below, assuming that the chips are arranged in other shapes, we can clearly see that the wafer arrangement is arranged in a square form with the highest space utilization. If arranged in a circle, the circle and the There are a lot of gaps between the circles that will be wasted. The wafer is every inch of gold. Think about the etch size and calculate the space in nanometers, so that the arrangement of circles and circles will waste space. When the wafer is cut into chips, it is cut along a straight line. The space utilization of the Tian-shaped arrangement is the highest. Of course, there are also “hexagonal” cutting methods proposed. In theory, this is feasible, but the cost of straight-line cutting Lower. Therefore, considering the comprehensive consideration of effective use of wafer space and cutting cost, the chips that are cut from the top of the wafer are founder chips. This is why the CPU or other electronic circuit board chips are square. Cut out the square Die is cut from the wafer and used to make the CPU. It is called CPU DIE, which is the CPU core. The screenshot of the CPU shown above is cut from the wafer. Is the lower DIE square? The design is more reasonable and uses space. Rate ah More CPU answers can look at these oh, what direction will CPU develop in the future?


By zhiwo

0 0 vote
Article Rating
Notify of
Most Voted
Newest Oldest
Inline Feedbacks
View all comments
6 months ago

This question is very interesting. According to the process flow I contacted; it is divided into two steps: 1: The substrate (Substrate) should be made into a round shape; the substrate and the PCB design are similar, and it should be made into a round shape, there should be no problem; (PCB has a design into Round) This is a round chip in the eyes of the user; through packaging design into various shapes, as long as the packaging factory cooperates, there should be no problem; 2: The DIE of the chip should not be made into a special shape; 2.1: Design problem: DIE should not be designed with special shapes; TSMC’s rules may not support it; for example: the world’s largest chip Cerebras wafer-level engine, a wafer is a chip, but it should still be cut into a square; if it can be made into a round shape, this side The area on the upper part can not be wasted; 2.2: cutting problem; if it is made into a special shape, the DIE cutting cannot be cut; it turned out that they have contacted the packaging factory and they cut the chip, and the cutting knife should go straight. T-shaped lines cannot be cut. Not to mention other shapes; to prevent ambiguity, the square I understand is square or rectangular;

6 months ago

Let me just think that the square mentioned by the subject includes squares and rectangles. First of all, the wafer is round, because the wafer is cut from a silicon ingot. The manufacturing method of the silicon ingot determines that it is a cylinder, so the wafer is round. So, this question It is simplified to: how to divide a circle into as many polygons of the same size as possible. This becomes a geometric problem. According to geometric theory, only triangles, quadrilaterals, and hexagons can be closely arranged into a plane, and there is no middle. The gap, the reason is not repeated here. If it is made into a hexagon, then the wafer cutting is more troublesome, because it is not a straight line to cut to the end, so the pass is left with triangles and quadrilaterals. In the floorplan stage of the die, you will find the corners of the triangles, especially The utilization rate of the acute-angle part will be very low, because it is not easy to route, it is not cost-effective. The acute-angle part of the quadrilateral is the same (in fact, the utilization rate of the right-angle part of the rectangle will be lower than the internal area). Excluding all the above options, only the rectangle is left.

6 months ago

I think there are several reasons: First, the processing substrate of the CPU, that is, the wafer, is circular. The following video is the processing process of the wafer. Simply put, it is to pull out a long cylindrical raw material through the “pull” method, and cut on this basis to form the wafer that can process the CPU. 20210410_153851CHEN’s video · 198 Play Actually, circular wafers have many advantages, such as: circular stress release is more uniform than any geometric shape: “stress treatment” during processing is one of the most decisive and critical events in the semiconductor field One (we can’t eliminate the pressure, we can only minimize it); and, in the process of manufacturing the chip, it is necessary to coat different materials (such as various polymers, photoresist, etc.) on a regular basis, and the circular wafer helps In order to coat the material uniformly, there will not be any edge competition effect due to the spin coating process. Secondly, processing square CPUs on round wafers helps to save materials. As shown in the schematic diagram below, cutting a round CPU on a round wafer will produce much more cutting waste than cutting a square CPU on a round wafer. In addition, from the perspective of the cutting process, the round Cutting is more expensive and time consuming than square cutting. Therefore, from this point of view, processing a square CPU on a round wafer is more helpful in saving materials and costs.

6 months ago

The apartment type is square, and the space utilization rate is large. What can a long corridor do? Perhaps there is a large memory that can be placed in the corridor. It is also just long, so it is just right, and the utilization rate is high. But putting such an airplane chip on the pcb hinders the utilization rate of the pcb. So from the perspective of wafer utilization and pcb utilization. Fang is the one with the best utilization rate. The round shape is not the best, because a wafer, it is round, there are many chips on it. The round shape reduces wafer utilization. Unless customized. Otherwise, a square is best.

6 months ago

Because your motherboard is square, the chassis is square, the notebook is square, and the circle is the carrier, the end product is unstable.
Therefore, corresponding to the chassis, the motherboard in the notebook, the smallest accessories, and the small packaged chips are all horizontal and vertical, and it is easier to calculate and arrange the corresponding things. A very logical answer~~~

6 months ago

Because the general wiring inside the chip is a horizontal and vertical wiring channel, each module is generally made into a square, and naturally the whole chip is also made into a square. If the chip is made round, that is to say, the sealring on the edge or other such as dicing grooves cannot be made round. Ps. Generally, each layer of the internal components of the chip, such as metal and active area, cannot be made into a curve because of the grid point problem.

6 months ago

Because the chips are of different sizes, they are arranged together, and the shape that saves the most area must be a rectangle. What I can’t understand is why it is a square. Remember that in the era of rapid development of CPU, the CPU was not soldered, but plugged in (after becoming a liberal arts student, the case was not opened in N years). In the case of “four-in-line”, the square may receive the most even force, and the “foot” is not easy to break.

6 months ago

In theory, it can be made into other shapes, such as regular polygons, but under the current process conditions, it will bring about a great reduction in efficiency and waste of economic costs. CPU is a kind of ic, ic is made on si wafer (wafer, wafer), in semiconductor process, wafer area is very important, if the pattern can be arranged regularly, it will save area; so this kind of scale The products produced need graphics that can be arranged regularly. For example, if it is set to a circular shape, there will inevitably be wasted blank area between the chip and the chip. In addition, after the semiconductor process is produced, the wafer needs to be diced. The current conventional process is to rotate a circular diamond blade to perform the dicing, and the blade travel path is a straight line. If it is not a simple and regular arrangement (such as the shape of a honeycomb or a football), the cutting requires multiple adjustments and alignments, and the time efficiency is reduced; and the interlaced graphics part cannot be processed, unless a part of other chips is wasted to cut a complete piece chip. In summary, the cpu is mostly rectangular.

6 months ago

Say a very simple truth. Suppose there is a chip design company, as the subject said, the CPU is not designed as a square but a circle. What will happen? This is equivalent to engraving many small circles in the big circle of the wafer (the chip is lithographically formed), and the effect is as shown in the figure below (to save time, I only drew a part). It can be seen that there is a considerable gap between the CPU and the CPU (that is, the small circle and the small circle), so that many silicon chips are wasted. A rough estimate is about 10%. The waste is still going on. Because the CPU is round, the electronic circuit is in a straight line. As for why it is necessary to go in a straight line, because between the two points, the straight line is the shortest, and the two components are connected by a straight line, which can not only reduce the current consumption (shorten the line and reduce the resistance), but also save space for other electronic components. The straight-line electronic circuits in the CPU are crisscrossed, and based on the principle of maximizing the use of silicon, all circuits of a CPU will form a square (as shown in the figure below). As a result, in order to design a round CPU, the silicon chips around the CPU circuit (orange square in the figure below) were wasted again, and the area of ​​this part was about 30%. A single circular CPU wastes the area of ​​the silicon wafer. When these CPUs are carved onto the entire wafer, as mentioned above, the gap between the CPU and the CPU is large, and it is wasted again (about 10%). The final waste effect is as follows Figure. The red horizontal and vertical bars in the picture are the wasted silicon area, and only the white squares are used. About 40% of the entire wafer is wasted.

6 months ago

All the chips used in daily life are made and cut on the wafer, and there is no way to install the circuit around the cutting line. Therefore, in order to reduce the wasted area, the total length of the cutting line should be controlled as short as possible when making the chip. Generally speaking, in order to make full use of the wafer, the best method is actually to cut the chip into a regular hexagon. . But because the cutting tool is an ultra-thin round blade, the thickness of this blade is much thinner than that of human hair, only reaching 1/3 of the thickness of human hair, so it is impossible to use this tool Cut the wafer perfectly into regular hexagons. So when we made the chip, we took the second place and made a rectangle. Can other shapes work? In fact, the shape of the chip is not impossible in addition to rectangles, triangles, parallelograms or even circles. However, the rectangle is the best deployment solution, and the utilization rate of the wafer can reach the highest.

Would love your thoughts, please comment.x