Rewriting the Technological Scenario

Meng Qian returned to Alphaham to a again talk about it, and then prepared the third day of the Global Developmenter Conference.

On the third day of the conference, everyone saw a lot of other products, and most users usually pay attention to things.

For example, ultrasonic flow meters, hard disk recorders, some car hardware, factory equipment, and some strange robots, etc.

There are many products that are exhibited, but the truth is a bit popular in this way, most products are not too popular for ordinary users because they are not too popular.

But industry insiders have seen the Today's products, but open various research and analysis.

Because today's focus is not these products itself, but the chip behind these products.

Today, all exhibited products are all adopted a special chip.

The theme of the third day of the General Assembly is a simple and rude special chip.

After a morning show, in the afternoon, Meng Qian once again appeared on the stage, you can see that Meng Qian attaches great importance to this conference this year, this is his highest frequency.

"General chip and special chip disputes are almost a topic starting in 2012. Now this dispute is not very hot, but I want to take this topic this year, because I never like to wait until things happen. Take it carefully, I see the incident habits.

In 1969, Neon's Busi selected Intel, making the world's first universal processor 4004, through rewriting software, greatly released productivity, and then thereafter, Intel's era has begun, and the era of General Computers has begun.

We can try to explore this issue from an economics, people in the field of economics know that productivity is an important privacy that measures economic growth and economic level. Since 1974, the three-third productivity has grown. Brought by the computer.

A new technology has made income through commercial success, and it will stimulate continuing to invest innovation, thereby further expanding production, users can also have a more powerful price to buy stronger strength.

Coupled with the demand growth will continue to be pulled, such a benign cycle has continuously promoted the progress of the general computing industry in the past few decades.

But we can now look at some public data. First, from 2008, the cost-effective growth rate of the chip has fallen from 48% to now 11%.

Look at this data, this is the financial statement of Intel last year, and Intel's fixed cost has exceeded variable costs.

In fact, our big wind semiconductor is about to face this issue.

That is to say, if we continue to develop by previous speed, it is very likely that the more developed, but if we don't develop, the industry's iteration speed will slow down.

But the profit is the premise of business. If you lose money year, whether Intel is still our big wind semiconductor, can you adopt more than?

How to solve this problem? "

Meng Qian spots from this very real problem from the perspective of economics, and from the development of the once, Intel's option is to slow down iteration, which is written in the financial statements of Intel. .

"I don't know what Intel is now thinking. For us, under such a big background, a problem that we have more concerned in these two years is the development of special chips.

Dedicated chips do not need to compromise in versatility and compatibility, cost-effective and energy efficiency performance is much higher than universal chips.

Moreover, the special chip can do better in a particular area, and performance can often reach the general chip to dozens of times or even hundreds of times.

However, because the iterative speed of the general chip is sufficient to adapt to the evolution of the computer, few people are placed on a special chip.

A special chip requires the development of ecology, otherwise the business path is too narrow, and it will lose business.

In addition, the dedicated chip lacks flexibility, and the new function cannot be added after the chip design is completed, and it has become a limit.

But in the past few years, the development of deep learning has made the unlimited possibilities of special chips to the market.

The future of special chips and universal chips will be, we can't get out of the economy. This problem can be developed rapidly in more than 40 years, which is due to economic benefits.

Users can use universal chips with more features under equivalents, users naturally choose universal chips, and companies will not pay this to invest in special chips.

But now, the cost of universal chips has risen, and the product is layered in the computer industry.

Smartphone is a clear layered product, and under the rapid development of smartphones, the development of AI chips quickly drives.

It turns out that as long as there is a sufficiently large market and a high-quality income, people will not hesitate to engage in dedication.

Let's take a look at the AI ​​chip data. In 2013, the global AI chip market broke through 8 billion soft sister coins. From 2014 you in the first half of the year, this year is likely to exceed 12 billion soft sister coins, and with mobile intelligence equipment The outbreak, our forecast model shows that AI chips are likely to break through the market size of 30 billion soft sister money in 2016, which is basically one-fifth of the global server CPU size.

For 2020, the AI ​​chip is expected to break through hundreds of billion. At that time, the market size of the AI ​​chip can reach half of the servers' CPU market.

However, how many years have developed, how many years have developed, which is the potential of special chips, because in the field of professional, the advantage of special chips is really too big.

Once the ecology is established, the chip is faster than the universal chip because of the dedicatedity, and the benefits can be brought about by the corresponding field.

If we use universal chips to support artificial intelligence development, artificial intelligence benefits may be more than 20 years.

Today, all products we have been exhibited at the conference, in fact, there is a problem to this market, and the chips of these products can be dedicated, then do they have enough business value? "

Meng Qian gave a warm discussion, Meng Qian made them discussed for a while, "Now no matter who to answer this question, it is a guess, but the guess is worthy of attention, at least From our awareness, we believe that the special chip era is coming.

Of course, I don't think that the general chip is not dead, just like I think the personal computer will not die, the mature market established by the general chip is not so easy to collapse, but the fact that we must accept is that the general chip will experience a period of time. The development is slow until the real technology revolution has emerged, such as the material revolution.

However, in the case of the general chip, it is difficult to continue to break through, and any law can't give us hope, but the potential of the special chip is full of possibilities. We have reason to believe that the chip and data center will gradually go to the heterogeneous era, but in this The times will have more and more dedicated chips and dedicated ecosystems.

This is the development trend of chip in our eyes. "

Meng Qian started to give the market, but this cake was still more responsible. Whether it is the various R & D results of this company, it is also the development momentum of the special chip that I have seen during the death in 2019.

The revolution of the chip material in a short period of time is indeed, so universal chips have to accept such a reality, starting from the 10nm process, and the development speed of special chips is getting faster and faster.

And all those who listen to Meng Qian's foresick as long as they are willing to take some efforts to study the knowledge in this field, they can see a thing.

Hongmeng architecture is the most suitable for development of special chips.

Compared to X86 and Power, the absolute advantage of the Hongmeng architecture will be present in the special chip field.

These words Meng Qian doesn't have to say on the spot, but if the dedicated chip can really develop, it is self-evident for the Hongmeng architecture.

Soon, there is a problem with a good news.

"Universal chip era is the X86 world, the special chip era? "

...

1603380856