Intel, NVIDIA, AMD's Grievances and Conjectures of Future AI Patterns

In the past, whenever Nvidia spoke AI stories in a victorious manner, hailed GPUs for the GPU, or occasionally used Intel's provocative attitude, Intel was more tolerant and low-key; faced with the failure of Moore's Law and slow transition in the outside world. In other areas, Intel also less publicly responded. So that at a press conference in September, Yang Xu, President of Intel China, suddenly said to the media, "The tiger does not speak, thinking it is a sick cat."

With the implementation of Intel's AI investment layout and the integration of services in place, Intel’s recent vocalization in AI has increased. From the headquarters to the top level in China, the high-level leaders frequently go to the front desk to actively “sell” Intel’s AI strategy and insight. At the same time, Intel is also actively promoting or participating in related market activities to promote the implementation of artificial intelligence strategies in all walks of life.

At the same time, Intel’s extraordinary moves in the past two months have also attracted the attention of the industry. First, in mid-October, Intel CEO Ke Zaiqi announced the launch of Nervana’s neural network processor by the end of the year. Later, Intel’s cooperation with old rival AMD was announced. , followed by Intel "dig" AMD's chief GPU architect Raja Kudry, plans to develop a separate GPU.

Wall Street also turned its attention to Intel. In mid-November, an investment company on Wall Street stated that investors should not overlook Intel’s artificial intelligence capabilities, and the company expects Intel’s potential to grow in multiple long-term growth markets and raise Intel’s target price.

If Intel's Tiger Tigers and AI's strengths break out, what will happen to the future AI chip market?

1.

If ten years ago, the acquisition of ATI was not AMD but Intel, what would happen? What is certain is that Nvidia will not have such a scene today, and Intel may not be able to give an impression of being behind on AI.

At the time, ATI and NVIDIA were both GPU vendors. At that time, ATI and NVIDIA competed on the GPU, and AMD and Intel competed on the CPU. In 2006, ATI was acquired by AMD with a huge amount of US$5.4 billion, and AMD became a chip company with both CPU and GPU R&D capabilities.

It should be said that AMD’s M&A strategy at that time was far-sighted. Unfortunately, at the time, this merger was too expensive for AMD under Intel's pressure. Prior to the acquisition of ATi, AMD had only $3 billion in cash and had to rely on debt to complete the merger. Being dragged by debt, AMD competes with different strong rivals on the CPU and GPU respectively, which brings even greater difficulties.

Nvidia’s leading position on the GPU later, it should also benefit from this acquisition – giving Nvidia a more relaxed space for development.

With the rise of artificial intelligence in recent years, NVIDIA, the leader in GPUs and graphics computing, has “stepped on the air”—although GPUs are not designed specifically for artificial intelligence, GPUs are good at graph-related calculations. Dealing with complex calculations has natural advantages, which makes Nvidia catch the AI's express train and the stock price has soared, becoming the absolute darling of Wall Street.

Intel and AMD obviously do not want Nvidia to maintain a dominant position on the GPU.

In early November, it was reported that Intel and AMD will cooperate with each other in the next century. In the future, AMD’s independent GPU will be installed in the Intel SoC. This is a reminder to Ying Wei Dao, who was facing the common enemy in the turn of the century. This proves once again that there are no eternal enemies, no enemies, only eternal interests.

However, AMD’s chief GPU architect Raja Coudry left afterwards, and Curriedly left AMD to enter Intel's door. According to reports, Lagar? Kudry will become Intel's senior vice president, in charge of the newly established "core and visual computing" group, his task is to help Intel develop GPU business.

Intel wants to develop a separate GPU! This is certainly bad news that Nvidia and AMD are not willing to hear!

In the past decade, AMD and Intel have competed in the CPU field, and AMD and NVIDIA have competed in the GPU field. Now that Intel is going to develop an independent GPU, this not only puts pressure on NVIDIA, the leader in GPUs, but also more passively on AMD, which is always in a weak position.

Imagine that if Intel joins the GPU to compete, the GPU market competitors will become Nvidia, AMD and Intel. How will the future pattern change?

2.

Previously, although Intel, Google, Apple, and other research and development of proprietary AI chips, but did not push the market, only the NVIDIA GPU is considered to achieve large-scale application and market-tested AI chip, and thus favored by Wall Street, the stock price Along the way, Intel is considered slower due to its lack of timely follow-up on the introduction of proprietary AI products in the investment and integration phase of the AI ​​chip. The stock price is in the doldrums.

On October 18th, Intel CEO Ke Qiqi announced at the WSJDLive Global Technology Conference that Intel will ship the Intel Nervana Neural Network Processor (NNP) before the end of the year. This is Intel's first industry-oriented neural network processing. chip.

Although Nervana has not yet officially launched the market, Intel has been unable to display the "killer" of this AI chip. After Ke Qiqi announced the end of the year before the launch of the Nervana neural network chip news, Intel shares rose sharply, indicating that the market expectation of this proprietary AI chip.

The Nervana neural network processor originated from the artificial intelligence startup Nervana Systems, which Intel acquired in 2016. At that time, according to Intel, this was the first company in the world to create chips specifically for AI. After the acquisition, the company’s founder, Raveen Rao, became the general manager of Intel’s AI solution division.

Intel placed high hopes on the acquisition of Nervana. Machine learning is the fastest growing AI application, and Intel is ready to attack Nervana on the AI ​​main battlefield to recover the market share lost due to GPU competition.

According to Intel, this Nervana neural network processor provides higher performance and scalability for artificial intelligence models, achieving 100 times the performance of artificial intelligence by 2020.

What is more noteworthy is that this AI chip has already been hailed - Facebook is participating in the cooperative development. But in the end, how Nervana's performance and results have yet to be tested by the market.

Intel's AI weapons, in addition to Nervana, also have FPGAs. Intel acquired FPGA manufacturer Altera for $16.7 billion at the end of May last year. This transaction is the largest acquisition in the history of Intel.

Today, Altera's FPGA has been integrated into Intel's processor products as a GPU-like acceleration technology. CPU+FPGA has become an alternative to GPUs.

For example, the U.S. Group adopted Intel’s deep learning platform including an FPGA cloud host; in August, Microsoft announced its full-line Project? Brainwave project, which uses the Intel Stratix?10 FPGA artificial intelligence platform.

Some professionals believe that although GPUs have certain advantages over CPUs, FPGAs are not inferior, and deep learning algorithms can run faster and more efficiently on FPGAs, and power consumption can be lower.

3.

At the GTC 2017 hosted by NVIDIA, Nvidia announced that the number of GTC participants has tripled in five years, and GPU developers have increased 11 times to more than 500,000...

The executives of NVIDIA will publicly publicize the growth rate of GPUs and the customers and ecological growth around the GPU in almost every public occasion.

Unlike NVIDIA's over-reliance on the GPU's AI strategy, Intel, whether it is CPU, FPGA, or Nervana, is part of the Intel AI full-stack solution. When talking about AI, Intel generally does not emphasize a single capability, but instead emphasizes a "full-stack solution."

What is Intel’s ability to do on AI? The person in charge of different levels of Intel will use the following diagram to give the answer.

Intel made a lot of investment from end to end, as well as hardware and software. In addition to the aforementioned acquisition of Nervana Systems, Altera, it also acquired artificial intelligence companies such as Movidius and Itseez. After Intel set up its artificial intelligence division in March this year, it also spent $15 billion to acquire Mobileye, and also established a $1 billion AI innovation fund. In October, it announced that it has invested in 15 AI start-up companies including China’s Horizon Group.

At the 2017 Intel Artificial Intelligence Conference held on November 15th, Fiaz Mohamed, general manager of business development at Intel's artificial intelligence business unit, said: Unlike NVIDIA's AI applications on GPU architecture, Intel's strategic breadth and depth outnumber its rivals.

Fiaz Mohamed believes that artificial intelligence is very complex, not only requires powerful computing power, different computing power, but also requires strong network transmission capabilities, advanced storage capabilities, and other aspects to provide a powerful computing platform to support artificial intelligence applications.

Intel's strategy is not only positioned on a "deep learning" link, but is instead placing the entire artificial intelligence as an ecosystem.

From the perspective of all-round investment and layout of artificial intelligence, Intel's goal is obviously not only to defeat NVIDIA, but also to target the entire AI market. This is the key to Intel's dominance over the future in the PC era and the mobile era.

Moreover, Intel's future investment and targeting are not only AI, driverless, but also the Internet of Things and 5G. After displacing the mobile chip, Intel is investing in a wide range of markets that may become high growth, and does not want to lose the next opportunity to make big profits.

4.

In the past few decades, in the CPU competition, whenever AMD challenged Intel with an outstanding technology or product, it was eventually suppressed by powerful Intel. Because, from the strength and financial resources, AMD and Intel are not in an order of magnitude, in the face of Intel's strong, each time AMD is difficult to parry, but also makes AMD degenerate into a "tragic company."

In the current artificial intelligence market, Nvidia takes advantage of the GPU. However, artificial intelligence is still in its early stages, and the market has only just started. Both Intel and Google and Apple have targeted this large market for growth.

If Intel’s full AI capability gets underway with the launch of Nervana's neural network chips, will Nvidia lead the turf a few laps ahead of artificial intelligence, and will it continue to be strong?

After all, the chip industry is both a technology-intensive industry and a capital-intensive industry. Although Nvidia holds a leading position on the GPU, Intel is much more powerful in overall technology and financial strength. After frequent investment in mergers and acquisitions, Intel still holds more than $20 billion in cash.

Analysts on Wall Street have also expressed concern: One concern is that Intel's participation in AI competition may affect Nvidia's gross margin sustainability. Nvidia is a chipless processing supplier (Fabless model), and TSMC is responsible for manufacturing the GPU. However, Intel has achieved vertical integration from design to manufacturing, and is more flexible and competitive in pricing strategy. If Intel competes for the market with price strategy, it will damage the gross profit margin of Nvidia. You know, when Intel used to deal with AMD's challenges on the CPU, it used price measures to suppress AMD.

Another concern is the threat posed by GPU alternatives to Nvidia. Currently, Nvidia's leading position in AI is mainly because there is no proprietary custom chip and GPU alternative solution to compete with it. When Intel's AI chip or Google's TPU is launched, there are competition risks to Nvidia's future development.

However, although Nvidia is inferior to Intel in terms of overall strength, it is also a veteran chip veteran. After all, it has competed with Intel in the history of development. It is believed that Nvidia is not easily defeated.

Smart Watch

Smart Watches,Smart Bracelet,Blood Pressure Smart Watch,Health Monitor Smart Watches;

C&Q Technology (Guangzhou) Co.,Ltd. , https://www.gzcqteq.com