However, the Rectangular Core No. 1 cannot be viewed from a traditional perspective.

On the one hand, the new MIX instruction set and compiler are equipped with the new concept of MatrixOS, which does not require the CPU to run fast, but requires the CPU to have a certain fuzzy computing ability.

From this point of view, the computing power of 4 cores and 1.0GHz is sufficient.

On the other hand, matrix semiconductors are also limited by the foundry level of its partner, SMIC.

Since Loongson has cooperated with SMIC many times before, and the relationship between the two sides is good, Matrix Semiconductor is still borrowing Loongson's previous cooperation channels.

The best process that SMIC can provide at present is 40nm, which is a certain gap with the 28nm Qualcomm chip chips sold on the market, but fortunately, this has little impact on the Rectangular Core.

The most important thing is that if the 28nm process is adopted, the cost of Rectangular Core No. 1 will also increase by about 50%, which is the real reason why Qiao Zhenyu finally chose the 40nm process.

Here, we have to mention the topic of process technology.

The so-called process technology refers to the size of the transistor on the silicon wafer.

As we all know, the transistor is the smallest unit that makes up a chip, and an NAND gate requires 4 transistors, and generally there are about 500 million transistors on an ARM quad-core chip.

The world's first computer used a vacuum tube, which had the same effect as a transistor, but the vacuum tube was two thumbs in size, and the most advanced etching process in the lab today has etched a transistor that is only 14 nanometers in size.

I want to make 500 million transistors with a size of 40 nanometers on a 15mm x 15mm square silicon wafer. If this process is to be done mechanically, it is difficult to find such a sophisticated instrument in the world that can carve nanoscale transistors. Even if there were, 500 million would be carved. It is also difficult to estimate the cost and time required.

As a result, engineers can only rely on the power of light.

This is where the lithography machine comes in.

Light can etch traces on silicon wafers. The mask controls which parts of the silicon are etched. Where the mask is covered, the light does not shine and the silicon wafer will not be etched.

After the silicon wafer is etched, it is coated with an oxide layer and a metal layer, etched again, and repeated many times to make the silicon wafer.

Generally speaking, it takes more than a dozen times to make silicon wafers, and the process and mask used are different each time.

Between several etchings. The position of etching may be deviated, if the deviation is too large, the chip can not be used, and the deviation needs to be controlled within a few nanometers to ensure the yield rate, so the technology used to make silicon wafers is the most sophisticated technology invented by human beings.

Chips can be mass-produced by mask etching, but masks must be slowly processed and made with higher-precision machines, and the cost is very high, a mask costs 100,000 US dollars, and a dozen different masks are needed to manufacture a chip. Therefore, the initial investment in chip manufacturing is very large, often millions of dollars.

The chip trial production process is called tape-out, and tape-out also needs a mask. The investment is huge, and before tape-out, no one knows whether the chip design will be successful. It is possible that tape-out has been unsuccessful multiple times. Therefore, there are really few companies in China that can make high-end chips, and few companies can afford the cost of mask alone.

After mass production of chips. The cost is relatively low, and a good mask is very large. With a diameter of 30 cm, it can produce hundreds of chips at the same time. If the chip is shipped in large quantities, the profit is still very high, like Intel's chips, which sell more than 1,000 pieces, may have an average manufacturing cost of less than 100.

But if the shipment is small, then the average chip manufacturing cost is frighteningly high, and it is normal for millions of dollars to be wasted.

This is also the reason why the cost of chip research and development is so high.

Like Huawei's self-developed HiSilicon chip in the past two years, whether its price is competitive or not depends on whether Huawei's mobile phone shipments are large.

The same is true for Rectangular Core No. 1, if other manufacturers want to produce MatrixOS mobile phones in the future, then this chip will become a sharp weapon in Yang Lin's hands.

The use of 40nm process not only reduces the cost, but also increases the yield rate of Moment Core No. 1.

After the design plan was determined, Qiao Zhenyu and some members of the Moment Core No. 1 team moved to the capital and went to SMIC's production line in Yizhuang for tape-out trial production.

Half a month later, the first batch of 30 successful tape-outs were brought back to the laboratory of Matrix Semiconductor, and Yang Lin cooperated with Qiao Zhenyu to test them.

The results show that the anthropomorphism rate of MatrixOS on Rectangle Core 1 has increased from 75% to 85% compared to that on ARM chips.

Its outstanding performance has already amazed all the staff involved in R&D.

However, Yang Lin is not satisfied, and in his expectation, the anthropomorphism rate of MatrixOS on the ARM chip must reach 80%, and the anthropomorphism rate on the Moment Core 1 must reach more than 95% to be considered successful.

As a result, the more than 8 million dollars spent on the previous tape-out were completely wasted.

But this is also no way, generally speaking, scientific research is to burn money, otherwise where will Intel's nearly 10 billion dollars of scientific research funds be lost every year?

Qiao Zhenyu had to convene the team members again to further optimize the microstructure of Rectangxin No. 1.

Yang Lin also seized the time to modify the kernel of MatrixOS step by step, and integrated the applications developed by the Matrix Digital MatrixOS project team into MatrixOS.

This work is endless, for more than ten days in a row, Yang Lin basically went out early and returned late, except for dealing with necessary company business, most of the time stayed in the laboratory and discussed and studied with the team of Rectangle Core No. 1.

From the lowest hardware architecture to the kernel of the MatrixOS system, the whole team has spent countless efforts to mature the MatrixOS No. 1 chip and the MatrixOS system.

On the afternoon of August 30, after half a month of intense revision, Qiao Zhenyu went to the capital again with a new design plan and nearly 10 million US dollars in funds, ready to carry out the second tape-out.

And Yang Lin finally relaxed, rarely got off work from the company on time, and as soon as he walked into the house, he heard a familiar voice talking in the living room.

He couldn't help but speed up his pace, turned the hallway, and saw Yang Ying sitting on the sofa, talking to Lin Qingya with a smile.

Yang Lin couldn't help but say in a lost voice: "Xiaoying, why are you back?" (To be continued......)

PS: Ask for a subscription, ask for a monthly pass.

Although I had a premonition about the result of the game after seeing Perran's selection criteria, I still couldn't help but watch it.

Alas, nothing more.

Football Association stupid!