In recent years, with the rise of mobile Internet, Intel face challenges. While NVIDIA (NVIDIA) in recent years in the field of general-purpose computing makes Intel pressure rise. NVIDIA (NVIDIA) in August announced the latest results, total revenue of $1.428 billion, up 24% from a year earlier, while profits reached $253 million, up 873% from a year earlier. This brought Intel a big threat.

Intel to launch new products, and in the test report said new Intel Xeon Phi processor capacity is higher than the market on GPU processors, directed at nVidia.

Said Intel Xeon Phi processor of 2.3 times the training speed is faster than the GPU, Xeon Phi up to one hundred and twenty-eight nodes, which is currently on the market of the GPU can't do it. At the same time, by the one hundred and twenty-eight Xeon Phi processor system composed of 50 times faster than a single XeonPhi processor.

Against Intel, NVIDIA have raised strong rebuttal, NVIDIA is pointed out that Intel's use of 18 months ago, if you use the Caffe Alex.net data update, you will find four gpus Maxwell than four Xeon Phi processor speed by 30%.

Words we are not unfamiliar, Intel and AMD, NVIDIA and ATI, then apple and samsung, domestic millet and huawei are called words. At Intel and NVIDIA saliva post-war face hidden what trends, we take a look at.

One, the new focus on artificial intelligence

NVIDIA released new products are used for deep learning operation system DGX - 1. This product design for artificial intelligence era, USES the NVIDIA Pascal core Tesla P100 accelerator.

In addition to the powerful hardware computing power of DGX - 1 also come with a variety of artificial intelligence software and development programs, including NVIDIA Deep Learning, DIGITSGPU training system, driver and SDK CUDA. The system can also support the cloud management services. According to NVIDIA, Tesla gpus 12 times faster than older GPU acceleration solutions.

And Intel also said special chip to develop artificial intelligence technology, Intel said next year to launch new models of xeon Phi processor. The new model will be introduced to accelerate the function of artificial intelligence computing tasks, will bring help to deep learning. Used in speech recognition, image recognition, and automatic driving, etc.

At present, the mainstream of the artificial intelligence systems are mostly mixed architecture of super computer, Intel CPU capability of providing cloud is together with the NVIDIA GPU.

Intel and NVIDIA spat, actual it is for the future of artificial intelligence processor war preparations of public opinion, the two are optimistic about the future of artificial intelligence market.

Second, the artificial intelligence algorithm breakthrough brings business opportunities

Artificial intelligence is not a new concept. From science fiction to the early research, artificial intelligence, at least hundreds of years of history. Why the recent artificial intelligence suddenly hot again? Thanks to the deep learning of the breakthrough.

The depth study of artificial intelligence needs a large number of data, through continuous training to enhance the level of artificial intelligence. Found a large amount of data in the past, it is difficult, and computer operation ability is not enough to cope with the requirements of deep learning.

And big data, cloud computing and cloud storage changed everything, in the era of big data and cloud storage, the data gained by the human expansion for ten million times, this gives the depth enough learning and training materials.

And the ascension of computing power, make some not too rely on chart before deep learning algorithm by spectrum.

With plenty of data, with relatively cheap, powerful computing capacity of deep learning is becoming more and more. And all kinds of algorithm is developed and applied in various fields.

In 2016, Google AlphaGO shocked the world. Go is considered to be the computer formidable originally, the traditional Mongolian topkapi luo tree search mode to chess or chess has is not a problem. But to 19 road board go is also very difficult.

And the computer, it is still difficult to judge the actor of low quality of chess form layout and the central disk, in particular, the situation of unknown phases.

While AlphaGO combine multilayer neural network with monte carlo tree search. At the same time using the depth of learning ability. AlphaGO using computer era of big data, obtained the must offer real game, learn from games to go situation of human judgment and decision making, greatly reduces the calculation of tree search.

Google also using the GPU general-purpose computing, and the use of large cloud computing resources to help computer training, through the games themselves, learning to improve myself.

Results human master lee se-dol lost nine computer, caused a sensation in the world news.

In fact, Google used multi-layer neural network, deep learning in recent years has been widely used in image recognition, speech recognition, translation, automatic driving.

Some thought the computer can't replace the human field suddenly become less absolute. Through deep study, it seems that what computers can try.

Google, baidu, Microsoft, apple and amazon... Almost all the Internet giant in research and development of artificial intelligence, are doing deep learning, and even unmanned aerial vehicle (uav) of emerging enterprises large xinjiang is trying to apply artificial intelligence to the object of unmanned aerial vehicle (uav) to circumvent the above.

This provides the opportunity.

Third, do the arms dealer of artificial intelligence

Artificial intelligence can enter almost all areas of human society, driverless cars, no one factory, unmanned aircraft, unmanned shop... , in addition to some creative jobs that almost all jobs can be replaced with artificial intelligence.

Develop useful artificial intelligence to the society, is the job of the Internet giants. Research and development itself is the vast computing power and storage capacity. In the battle for artificial intelligence. Benefit most is not necessarily a winner, and is likely to be small. Intel and nVIDIA is fighting to artificial intelligence war arms dealer.

The two companies are not satisfied in the last round of mobile Internet. Intel's admission policy mistakes late and powerless. NVIDIA started early, is dead in baseband, buy useless, finally had to drop out of the competition.

And in the next the tuyere of artificial intelligence, both are rolled up his sleeves. NVIDIA GPU general-purpose computing was very suitable for artificial intelligence, is vested interests, natural regress, DGX - 1 ambition have integrated control industry.

And Intel also is not a vegetarian, all the nuclear XeonPhi architecture is inherently is very similar to the GPU, and Intel X86 instruction set, development and transplantation is relatively easy, plus Skylake core bottom, Intel is also preparing to make a difference in the field of artificial intelligence.

At present, the NVIDIA has certain first-mover advantage, Intel will next year will launch for high performance computing products of artificial intelligence.