Chapter 795 Don't Care About Everything

Half an hour ago, Fang Chen brought three unexpected products as the last important press conference this morning.

However, the first product brought by Fang Chen has left all the people watching the live broadcast, M9000 series, 10nm process, AI-enhanced dedicated chip.

M9000 is a brand-new series of processors of Dafeng Group. First of all, it has successfully realized the 10nm process. The market sentiment for the launch of 10nm this year is generally pessimistic, especially Intel’s delayed prediction of the launch time of 10nm general-purpose processors. Let everyone realize the difficulty of the 10nm process.

But unlike Intel, Dafeng Group has never intended to postpone the 10nm strategy, and has come up with a 10nm process chip on time. Although this is a dedicated AI chip, although Dafeng Group has not been able to produce a 10nm computer CPU, But in the 10nm competition, Gale Group has already taken the lead.

Dafeng Group will use EUV lithography machines from the beginning of 7nm, and Dafeng Group has won this step, whether it is based on facts or market expectations, it has actually won a lot.

The M9000 series will be used on tablets. In addition to the 10nm process, another highlight is AI enhancement, which further enhances the AI ​​performance. According to field data, the AI ​​performance of the previous generation of tablet chips has improved by a full 20 times.

In fact, with the development of technology, it is not so difficult to superimpose AI performance on the processor, but whether it is necessary has become a question that needs to be considered.

Fang Chen did not immediately explain why he wanted to make a dedicated chip for AI enhancement, but brought the second new product, the brand new GPU Q series.

The Q system has also received a lot of attention, because Fang Chen showed that the AI ​​performance of the Q series is 15 times higher than the company's latest GPU.

The Q series can be used in many AI scenarios. It can be said that the Q series is a GPU born for AI.

The reason why the Q series can achieve such performance is due to the newly developed sky architecture of Galaxy Technology. On the one hand, this new architecture introduces a tensor unit that can be better used for deep learning.

On the other hand, structured sparseness is used. Structured sparseness is an effective analysis method for high-dimensional data. It has the advantages of simple features, strong interpretability, and convenient calculation. The addition of this function enables GPU to process high-dimensional data more efficiently. data, which is equivalent to improving the AI ​​learning performance.

The combination of tensor units and structured sparseness will have the effect of 1+1>2.

When Fang Chen introduced the latest Q series GPU, everyone subconsciously believed that this GPU was born to better serve the AI ​​system. However, Fang Chen's next sentence surprised many people, "Our Q series series, will be installed in our ancient computers."

A M9000 series dedicated AI chip for a tablet, a Q series GPU for an ancient computer, everything points to one thing, the Gale Group is frantically improving the AI ​​performance of its products.

But what is the purpose of doing this?

Everyone looked at Fang Chen.

"There's a common question about why AI chips aren't built into personal computers.

In fact, most people know the answer, because CPU GPUs, especially GPUs, are sufficient to meet the AI ​​needs of personal computers. If you have special AI needs, you can buy an AI accelerator.

So to put it bluntly, it is a question of application.

Mobile phones have face recognition, AI shooting, and intelligent voice. Is it possible for personal computers to have breakthroughs in AI applications, even to the extent that our GPUs cannot handle them?

It can be said that now AI chip manufacturers have reached a stage of explosive growth. At present, there are more than 30 domestic AI chip companies with a market value or valuation of more than 100 million, including mining chips, mobile phone chips, box chips or pure AI chips. .

Countless AI chip players around the world have entered this market. It can be said that the hardware is now in a state of preparation, and we need applications to promote the further development of AI.

Although many companies have been working hard to come up with AI applications in recent years, it is difficult for limited AI applications to make big waves. We need an AI application with a high upper limit, which is the personal artificial intelligence that everyone heard a few hours ago.

I noticed that many friends on the Internet are asking why this technology has to wait until the end of the year to go online. Since it has to wait until the end of the year, why is it so anxious to release it in advance.

That's because we are going to release more powerful AI chips today, and even redefine the AI ​​​​computing power of PCs.

If we don't take the app out, more people will think we're doing an extremely boring thing.

Only when personal artificial intelligence is paired with our AI augmentation will everything really make sense.

As we mentioned before, as long as you have a device, such as a mobile phone, tablet or computer, you can use personal artificial intelligence technology, but how effective this technology is depends largely on the chip AI capabilities of your device. "

Fang Chen turned on the big screen and started to make a video demonstration, "We can intuitively see the gap through the comparison of experiments.

There is a huge gap in the PA computing power that can be carried by the AI ​​chips of the previous generation of tablets and the AI-enhanced chips of the new generation of tablets. This gap in computing power brings about a gap in application effects, which will directly reflect the intelligence level of our personal artificial intelligence. superior.

We often say that artificial intelligence may have no upper limit, if computing power, algorithms and data can be improved simultaneously.

This is a theoretical basis for the high limit of personal artificial intelligence, so what we bring today is not a simple application, nor a simple dedicated chip or GPU, but what we bring is how to advance the development of the entire artificial intelligence One big step approach. "

When Fang Chen said this, everyone understood that personal artificial intelligence and cloud cards are completely different things. Its computing power is placed on the local chip, and what it buys from the cloud is actually software.

And if you want to truly display the ability of personal artificial intelligence, you need to use more powerful AI chips, and these more powerful AI chips will first appear on the electronic products of Dafeng Group, the computers of Dafeng Group, and the computers of Dafeng Group. tablet, and the upcoming Hongmeng phone.

In this way, a technological innovation and technological leadership based on the internal ecological foundation appeared, and in order to further solidify this technology, Dafeng Group did one more thing.

"The further development of the chip needs to solve the problem of adaptation and standardization. Therefore, we have joined forces with Ali, Huawei, Sugon, HP, Motorola, Sony and other companies to create an enhanced artificial intelligence chip alliance, which is committed to solving the problem of adaptation and standard setting. .”

As soon as this news came out, it hit Intel's fate. This is the first time that a global chip adaptation and chip standard setting has been separated from Intel. This is the biggest insult to Intel.

It is even more insulting than a chip doing better than it, especially since Fang Chen also conducted an industry review and prediction on the development of the chip at the end of the press conference.

"From the 1960s to the 1970s, the transition from the military industry to the home appliance industry promoted the migration of the semiconductor industry from the United States to the neon country, and system manufacturers began to differentiate into IDM manufacturers.

From the 1970s to the 1980s, the era of home appliances entered the primary stage of PCs, and IDM manufacturers began to differentiate into wafer foundries, Fabless and design service companies, and the semiconductor industry moved from the neon country to Korea and Taiwan.

Now, humans have entered the mobile phone era from the PC era, the industrial division of labor has been further refined, and light design has gradually become the mainstream. With the advent of the era of cloud and intelligence, the semiconductor industry has officially tilted towards the Chinese mainland.

I hope the whole world can see this change, and next, we look forward to further cooperation with global enterprises to promote industrial development and industry development faster. The infinite possibilities of the future are there, and AI will advance again. It's just a start. "

In just one morning of the 8th Gale Global Developers Conference, the world saw the two ambitions of Gale Group.

One is full-stack AI, which is a complete smart cloud solution covering everything from chips, platforms to frameworks and scenarios.

The other is the chip structure. Dafeng Group has no taboo to announce to the world that future chip development will be led by us.

Many people watched Fang Chen's speech and couldn't help sighing:

Dafeng Group really doesn't care about anything.