Can Intel beat NVIDIA GPUs after winning Nervana?

Artificial intelligence is getting hotter and hotter in the world. The artificial strategies of Google, IBM and Microsoft have been released one after another. The industry has long been waiting for Intel's artificial intelligence strategy map. From receiving Saffron, Movidius, nervana, to layout drones, autonomous driving, precision medicine, what kind of chess does Intel play in the AI ​​market? Not long ago, Intel announced its artificial intelligence strategy in the United States. On November 30, Yang Xu, Intel's vice president and president of China, said in an interview: "Learning internal skills, doing ecology, focusing on key AI applications are the three key factors of Intel's AI strategy."

Deep learning chip is 100 times faster than GPU after three years

On November 30th, the Intel Artificial Intelligence Forum was held in Beijing. Intel’s internal artificial intelligence related fields were all attended, including Naveen G.Rao, general manager of Intel Artificial Intelligence Solutions, and Jason, General Manager of Intel Data Center Solutions. Wasilman and Wiliam Savage, General Manager of Software and Services.

From the perspective of the AI ​​map released by Intel, Intel's dimension of AI's “integration” includes the “function and experience”. At present, everyone's attention to Intel AI is more focused on the "functional" dimension. After accepting Saffron, Movidius, and Nervana, Intel has been enhanced in AI "functional" capabilities such as machine learning/deep learning, inference systems, and machine vision.

90 days ago, Intel completed the acquisition of Nervana Systems. Nervana, a supplier of artificial intelligence ASIC chips, acquired Nervana and is seen as a response to Intel's “one-sided” GPU adoption against the AI ​​market. At present, most of the deep learning and machine learning in the market are handled by the GPU. It is for this reason that the stock price of NVIDIA with GPU has risen and rises. Can you win the Nervana and let Intel beat the NVIDIA GPU?

Nervana co-founder Naveen G.Rao is now the general manager of Intel's Artificial Intelligence Solutions division. He believes that GPU is not born for AI, so it is impossible to be better than Nervana. "NVIDIA's products are mainly for image rendering, which is similar to deep learning, but it is not exactly the same. From the current choice, GPU is the most suitable product to accelerate deep learning. But in fact, people can also build More optimized and more suitable for deep learning products. From the architectural point of view, Intel's deep learning chip is ten times faster than GPU accelerated deep learning."

Naveen G.Rao said. Currently, Intel has announced more details on how to integrate Nervana's technology into existing product roadmaps. Intel will test the first chip (codenamed Lake Crest) in the first half of 2017 and will be available to major customers in the second half of the year. In addition, Intel has added a new product (codenamed Knights Crest) to the artificial intelligence roadmap that tightly integrates Nervana technology with Intel Xeon processors. From the information revealed, Intel hopes that three years later, Nervana can shorten the training time for deep learning models to 1/100 of the GPU solution.

Not long ago, Intel announced its AI product portfolio in San Francisco. In addition to Nervana, Intel also includes Xeon processors, Xeon Phi processors and FPGAs. Intel has given such a product portfolio. Because of the data, in the data center, the current AI-related calculations account for about 7%, which is still a relatively small part. Different chips are still needed for different workloads.

Intel’s executive vice president and general manager of the data center business unit, Bo Anna, revealed that the next-generation Intel Xeon Phi processor (codenamed Knights Mill) can improve its deep learning performance by four times compared to the previous generation of processors. Listed in the year. In addition, Intel has provided an initial version of the next-generation Xeon processor (codenamed Skylake) to specific cloud service provider partners, which uses Intel's advanced vector instruction set AVX-512 integrated acceleration technology to greatly enhance the machine. Learn the reasoning performance of workloads.

Despite the roadmap, Nervana and Intel are technically integrated and the production of chips still takes time. Future routes may still be fine-tuned with the market.

Promote the key application of AI ecological layout

Naveen G.Rao worked at Qualcomm and later founded Nervana. Why choose to "marry" Intel instead of Qualcomm or another company? He expressed his appreciation for Intel's capabilities in data centers and ecology. To get Nervana to release more energy, it is better to join an enterprise like Intel that has ecological capabilities and strong data center and chip manufacturing capabilities.

Another important dimension in Intel's manual strategy map is the AI ​​ecosystem. Intel hopes to create an open artificial intelligence ecosystem through a series of development tools that combine ease of use and cross-platform compatibility through alliance building and education.

In the United States, Intel and Google have teamed up to collaborate on technology integrations such as Kubernetes (containers), machine learning, security, and the Internet of Things. At the artificial intelligence forum in Beijing, Jingdong appeared on the platform. Jingdong revealed that it can use the Intel Xeon processor and Caffe version solution to respond more flexibly to the image data processing tasks of the billion-level. For example, the online performance of illegal image recognition has increased by more than 4 times. At the same time, it is revealed that the two sides will continue to cooperate in the field of artificial intelligence.

Not long ago. Intel also introduced the Intel Nervana Graphics Compiler to accelerate the development of deep learning frameworks in Intel chips. At present, various IT giants are strengthening the development of a deep learning framework. Whether this platform can gather a larger group is still unclear.

Jason Waxman, vice president of Intel Corporation and general manager of the data center solutions division of the data center business unit, believes that autonomous driving is a typical artificial intelligence application and will be a trillion dollar market. Intel has set up a new dedicated department, the Automated Driving Group (ADG), to drive the development of driverless solutions. Intel previously announced that it will invest $250 million in venture capital to invest in autonomous driving technology. On Tuesday, Intel announced a partnership with Mobileye and Delphi.

In fact, Intel is not only investing in these applications that are “strongly associated” with AI, Intel is also active in sports, entertainment and other fields.

For Intel's extensive penetration and involvement, Yang Xu said that Intel is a company focused on data processing. The changes in data characteristics must be understood by Intel. This is why Intel should pay attention to and deepen its involvement in all aspects including sports. The reason for each industry is to provide better data processing capabilities and data experience, and to release data value.

Magnet Pc Speakers

Magnet Pc Speakers,Manget 2.0 Speaker With Bluetooth,Pc Speaker With Bluetooth Function,Magnet Thin Small Speaker

Comcn Electronics Limited , https://www.comencnspeaker.com