The students all dispersed, and only Li Yixuan and Yu Youcheng were left in the laboratory, and after discussing it, it was decided to let Yu Youcheng lead the development plan of the Core CPU. However, in terms of the overall architecture of the CPU, some changes need to be made.

Because the original architecture was Intel's X86 architecture, while the 6502 on the Oranges I computer was the 68X architecture, it had to be modified accordingly.

The x86 architecture does not refer to a single architecture, but to an architecture, the most well-known of which are AI-32 and AI-64, which will violate Intel's basic patents if they are not modified recently.

In addition to the X86 architecture, there was also Motorola's 6X architecture during this period, and the 6502's 68X architecture was born from Motorola's 6X architecture.

Unlike Intel, since the Z80 processor defeated Intel's 8080 processor, Intel launched the 8086 processor and 8087 digital coprocessor in 79, and in the next few years, around the 8086 and 8087 processors, has been building X86 patent technical barriers, the purpose of which is not to let its competitors rely on X86 to launch a CPU of the same type.

Motorola is different, its competitor MOS's 6502 and his 64000 are not in the same application level at the technical level, 6502 did not impact the 64000 market, so he did not build a complete set of patent barriers around 6X, which brought Li Yixuan an opportunity.

This is also the reason why he chose 6502, Intel's X86 He has no chance of getting involved.

In addition, Li Yixuan chose 6502 for another reason, that is, in this period, the most used CISC is the 6502 processor, but the X86 application is a niche in the market, and Motorola's 84000 is a niche in the niche, which is very different from a large number of X86 applications on the market in later generations.

If a 16-bit CPU compatible with 6502 can be launched, then a large number of applications built around 6502 can be fully received, just like Microsoft built a DOS system on the basis of CP/M, Li Yixuan also hopes to build his own CPU empire on the corpse of MOS, he asked the negotiation company to buy 6502 compatible instruction set license from MOS company a few years ago, which is preparing for today.

The CP/M operating system was the world's first successful microcomputer operating system. CP/M can be said to be the prototype of DOS, and later CP/M's digital research company wrote a system called 86-DOS on the x86 platform according to the operation method of CP/M.

When Microsoft and IBM began to work together, Microsoft struggled to have its own operating system for IBM's machines, so it simply bought 86-DOS completely and renamed it MS-DOS. And because MS-DOS was born out of CP/M, MS-DOS can be compatible with software applications on CP/M, resulting in data research companies losing to Microsoft's MS-DOS in the new generation of 16-bit control systems, thus completely disappearing from the market.

It can be said that Li Yixuan's method is similar to that of Bill Gates at the beginning, which is to build a new camp on the basis of 6502 by purchasing a license.

Don't blame Li Yixuan's black heart, if you want to blame MOS himself, like a digital research company, indulged in the glory of 6502, and don't know how to forge ahead. If you could have launched a 16-bit CPU that was fully compatible with 6502 sooner, would he Li Yixuan still have to do so much? Besides, MOS is even more depraved than digital research companies.

Li Yixuan spent a full $6 million on the compatible authorization of the 6502 instruction set, which is not counting the commission to the negotiating company. Although the money is spent a lot, unlike Microsoft's MS-DOS, there are a lot of hidden dangers, and he does not, although it costs a lot of money.

Li Yixuan did not buy the 6502 architecture from MOS, it is easy for him to be fully compatible with 6502, he only needs an instruction set system that can be compatible with 6502.

The architecture is not needed, and the kernel of the 6502 is not needed, so you need to use your own kernel. Directly adopt the core of the hard disk total control CPU designed by Li Yixuan as the core of the core CPU, and Li Yixuan also named this core V1.

In Yu Youcheng's view, although Li Yixuan put forward the concept of 82, provided the basic architecture system of a new generation of CPU, and even provided the basic core, but to complete the research and development of core, there are still many problems.

Especially in terms of frequency doubling technology and internal bus technology, he had to do a lot of experimentation.

Semiconductor chips are a highly capital- and talent-intensive industry, and in the entire Oranges company, only Li Yixuan and Yu Youcheng know how to design chips, except for those robot engineers.

Since Li Yixuan has his own business, it is impossible to squat in the laboratory every day, so Yu Youcheng has to do it himself, unlike in Intel and Fairchild, there are a large number of technical experts who can share his work.

And the group of students he brought with him, although they are the top elites in China, but due to their lack of development experience, Yu Youcheng still needs to take a few hours a day to give lectures to them like the ancient masters and apprentices, which is even slower.

That's not the worst thing about him, the real bad thing is that he doesn't have a good RISC model to teach his students.

Because RISC has just been born not long ago, many things belong to the stage of exploration, although the CPU designed by Li Yixuan is the first RISC in the real sense, but Li Yixuan's CPU is derived from ARM processors, which have too many CISC characteristics and are not suitable for teaching.

So he found Li Yixuan.

After listening to the other party's intentions, Li Yixuan thought about it and drew a simple RISC structure diagram on the drawing, "You can tell them about RISC with this." ”

What Li Yixuan gave him was the basic architecture diagram of the R2000 of the famous MIPS company, which is also a very popular RISC processor in the world.

It was first developed in the early '80s by a research team led by Professor Patterson at Stanford University. MIPS's R series is a microprocessor developed based on RISC's industrial products.

In later generations, the architecture diagram of MIPS's R2000 processor became the standard architecture diagram for RISC teaching in many university universities. The reason is that the structure of the R2000 is advanced and simple, and it is easy for students to understand using it as a teaching model. Eventually, MIPS became a measure of computer speed.

MIPS stands for Million Instruction Per Second – the number of millions of instructions a computer executes per second. That is, 1 MIPS is equal to the number of instructions that a computer can execute per second.

"You can use this as a brand new CPU. Yu Youcheng said.

Li Yixuan was stunned for a moment, knowing that he had inadvertently robbed someone else's achievement, because Professor Patterson, the founder of MIPS Company, was still teaching at Stanford University, and had just established the RISC research group last year, and he would not establish MIPS until 1984, and the first MIPS processor R2000 would not be released until 86. This made him the father of RISC.

Li Yixuan thought for a moment and said: "This drawing, you can publish it to the world, allowing any company to develop a CPU on this basis, but there is one condition, that is, the technical achievements must be shared with Oranges." ”

"You don't plan to make it yourself, your picture has a huge market value!"

Li Yixuan waved his hand, it is enough to have ARM and Intel's Core series, as for MIPS, he doesn't plan to do it, just leave it to others, he can't eat all the CPU architecture, leave it for others, besides, MIPS is not without flaws.

It's so pure that it takes a lot more time to write the application software than ARM and CISC.

The rise in time cost has led to the launch of application software with the same performance later than other companies, which makes the cost of software procurement, application and maintenance rise, which is why MIPS was finally defeated by Intel's Xeon CPU.

In addition, although the structure of MIPS is simple, it has high requirements for the design and writing of instruction sets, and requires very senior programmers to complete this work, which leads to the long development cycle of MIPS, which is also the reason why MIPS has been slow to update in history.

In addition to the above two points, there is a third reason, that is, MIPS has high requirements for auxiliary hardware, such as motherboard, memory, hard disk interface, total data cable, etc.

In 1992, SGI acquired MIPS Computer Company, and when it comes to SGI, we have to say that their SGI graphics workstations, before 2008, the graphics workstations on the market were the world of SGI, and in the 90s, SGI took the same approach as IBM, which was to open the workstation standard, hoping to make the SIG computer standard a public computer standard like IBM-PC, so that it could make money by selling CPUs like Intel.

Prior to 2008, SGI's program was so successful that it became the world's second-largest CPU supplier after Intel, much larger than AMD.

But in 2008, everything fell silent. The reason for this is that the cost of the SGI application is simply too high.

When was 2008, when the subprime mortgage crisis broke out in the United States, before the subprime mortgage crisis broke out, people could still accept the high cost of MIPS, but once the crisis broke out, people and companies' wallets shrank significantly, and SGI suddenly became a luxury, and people were no longer willing to pay for it, and they switched to Intel's universal graphics workstations, and SGI died.

Only the world's top special effects companies, such as ILM and Digital, are still using SGI's graphics workstations. However, even so, only the most top-notch special effects will use SGI, and generally the most demanding special effects will use Intel or AMD standard general-purpose graphics workstations in addition to cost considerations.

At the beginning of 2007, the famous Disney Company in the United States announced that its Pixar Studios would use all Intel and AMD standard general-purpose graphics workstations in the new year's computer purchase plan, and the prospects for MIPS became even more bleak.

From the lessons learned by the SGI, we have come to the conclusion that the best is not the most suitable, and that cost alone is ultimately king.

This does not only refer to the cost of chip procurement, but also the cost of design, time, use, and maintenance, any of which can determine the success or failure of a CPU type. Li Yixuan seems to understand the reason why Jerry Sanders, the founder of AMD, wanted to go all the way to the black on the road of CISC.

Cost is king.