As soon as Li Yixuan returned to Hong Kong, he received news from Yu Youcheng that DMG's first 16-bit central processing unit was successfully developed, which fully met the original design requirements.

Coming to the laboratory, Yu Youcheng pointed to the latest new processor on the table, and said excitedly: "Xuan Shao, this is the latest 16-bit central CPU, which contains 136,000 transistors and a clock frequency of 24MHz. Both the internal and external data buses are 24 bits, the address bus is 24 bits, and the addressable memory size is up to 32Mb. It adopts the latest compatibility instruction set, which is fully compatible with all the functions of the 6502 and is the upward compatible microprocessor of the 6502, so that the assembly language program of the 6502 can run perfectly and smoothly on the 80286 without any modification.

At the same time, this CPU has a real mode and a protected mode, which is a watershed moment in CPU design. ARM's reduced instruction V1 kernel is adopted, and we have also upgraded the V1 kernel to have 4 functional components inside, namely the address component AU, the instruction component IU, the execution component EU and the bus component BU. These four components are operated in a full parallel pipeline, which greatly improves the data throughput rate, makes the processor faster, and its floating-point computing power is up to 3.8 million ......"

Wow, this performance has completely exceeded the performance of the 80286 CPU released by Intel next year. First of all, the clock frequency of 80286 is 20MHz, this CPU is 24MHz, don't look at the increase of only 4MHz, you must know that the clock frequency of 6502CPU is only 1.2MHz, and the highest after improvement is only 1.8MHz. The extra 2MHz is equivalent to the processing power of 2 more 6502 pieces.

In terms of integer computing power, it exceeds 80286.

Generally speaking, the computing speed of the CPU is based on two indicators, one is the integer computing ability, and the other is the floating point computing ability.

Integer operation, also known as addition operation, as the name suggests, reflects the addition operation power of the CPU, expressed in clock frequency, in MHz. It was most commonly used in early personal computers, because most of the other aspects of daily use depended on the ability to do integer operations. This is also the reason why we go to the computer city to purchase computers, and look at the CPU mainly to see the clock frequency.

Floating-point arithmetic, also known as logical arithmetic, reflects the logical arithmetic ability of the CPU, which can be understood as the operation mode in which the decimal point can be moved, and belongs to the high-precision operation of the CPU. Floating-point mainly affects multimedia performance, what 3D graphics, video, audio, games derive real-time computing, and these aspects of performance and performance.

For early personal computers, floating-point was not valued except for game consoles. Only mainframes and supercomputers value the floating-point power of CPUs.

But with the development of personal computer technology, especially the arrival of multimedia personal computers, floating-point arithmetic has become more and more important.

In many cases, the competition of desktop processors, in addition to the traditional clock frequency, depends more on your floating-point computing ability, because the stronger the floating-point computing power, the stronger the multimedia ability of the CPU.

After entering the era of network streaming media and intelligence, the requirements for CPU floating-point arithmetic are even higher.

Float Point Unit, referred to as FPU, is a floating-point combinator of CPU, mainly engaged in floating-point arithmetic, formerly known as digital coprocessor, Intel Corporation was not included in the CPU before 486. However, Li Yixuan started with video games, and has a high demand for the computer's gaming capabilities, so he integrates the FPU directly into the CPU.

This is a lot stronger than Intel's 80286.

Because it uses a reduced instruction set kernel, and the reduced instruction set kernel itself adopts the 28 design principle, so its execution efficiency is much higher than that of the 286 processor, in order to adapt to this high computing speed, Yu Youcheng designed the internal and external data bus to 24 bits, which is much higher than the 16-bit internal and external bus of 286.

In short, this CPU surpasses the 80286 CPU that Intel will launch next year in every aspect.

"How much does this CPU cost to produce?" This is Li Yixuan's main concern, if the CPU performance is very powerful, but the production cost is high, it is not the processor he wants.

Yu Youcheng replied: "The production cost of a single chip is 80 US dollars, and if it is produced on a large scale in industry, a single processor can be reduced to less than 1 million / 60 US dollars per processor, which is completely within the range accepted by ordinary consumers." ”

Li Yixuan breathed a sigh of relief, so that it can set the final retail price of the CPU at $199 per unit, compared to Intel's CPU of more than $270, Li Yixuan's processor still has a great advantage in price.

"However, Xuan Shao, we did not use the 1.5-micron process technology for this processor, but lowered it by one level, using the 2-micron production process." Yu Youcheng reminded.

"Ah, why?" Li Yixuan was puzzled.

"If 1.5 microns are used, the clock frequency of the CPU will easily exceed 40MHz, and the floating-point arithmetic will exceed 10 million, which will cause the barrel effect caused by a single technological breakthrough." Yu Youcheng explained: "You should understand what the barrel effect is, such a powerful CPU, the hardware requirements are very high. The CPU of the Oranges I computer is 6502, and the hardware requirements for the computer are not high, even if the technology is backward, you can pay attention to it.

But this CPU is different, the Oranges II is a high-performance computer, and the performance is not weaker than the small computers on the market. With such a high performance, it is natural to use the best hardware within the scope of the force, such as memory, which has a larger capacity and faster read and write speed.

At present, the hardware for personal computers on the market generally does not meet the requirements at all, and can only use high-end parts above the minicomputer. As a result, the final price of the computer will become so impressive that the average consumer simply cannot afford it, and even the business will not buy it. Because from the perspective of cost-effectiveness of the enterprise, it is better to buy your expensive personal computer than to buy a minicomputer from DEC and HP, and your personal computer, no matter how powerful it is, is not as good as a minicomputer.

So, we finally discussed going with the 2 micron process. Although the performance of the CPU has been weakened, it can be used by the general components on the market today. After the performance of basic parts such as memory and cache improves, it is not too late for us to launch CPUs with 1.5-micron processes. ”

Considering the current computer hardware environment, Li Yixuan is not a tenacious person, and said: "Okay, anyway, you can just look at these things, I have limited energy, and I can't manage so much." However, since the process has been downgraded, the CPU should also be optimized, right? ”

"We've thought about it for a long time. Xuan Shao, you look at it, ......" Yu Youcheng picked up the chip again, and said: "After our agreement, we have set up two CPU optimization schemes, one is the one you hold, which is the CPU for general personal computers; The other one is the one I have, which is aimed at high-performance computers. ”

Is it i5 and i7? That's interesting. took a look at the CPU data in Yu Youcheng's hand...... Damn,Where is thisi7?,It's clearly the difference between Xeon and Core.。

Xeon Xeon processor is a general-purpose CPU that Intel decided to develop a new mid-level server/workstation CPU in order to distinguish the server market from the ordinary personal computer market.

Because early high-performance computers, mainframe, mid-range and small form factor CPUs, were high in performance but expensive. The starting price of any one is more than $1,500, and the high one is even tens of thousands of dollars.

So Intel came up with a cost-effective solution.

According to different enterprise markets, Xeon Xeon processors are divided into three series: E7, E5 and E3, E7 is aimed at the network server market, such as telecom and mobile basic servers, from low to high to a little more than 10,000 yuan, which is much cheaper than other companies' processors in price, but the performance is basically the same, and it will not be far behind.

E5, aimed at high-performance computer clusters and high-performance servers, the price ranges from $2,000 to $6,000, the performance is not weak, the compatibility is good, and the price is half cheaper than other CPUs.

The E3 is a bit like a minicomputer and graphics workstation CPU, and the lowest CPU price is the same as the personal computer CPU we use every day, and the highest is less than $2,000.

The CPU in Yu Youcheng's hand is completely set in accordance with the low performance standard of Xeon E3, mainly for general enterprise micro servers and general enterprise workstations.

Li Yixuan dared to pack a ticket, this was definitely the solution proposed to him by the engineer robot.

And the one in Li Yixuan's hand is like the Core i7, as for the i5 he thinks, it is not the same thing at all.