ARM CEO: The car will become a supercomputer, pushing AI into smart hardware

650) this.width=650, "src=" http://s2.51cto.com/wyfs02/M00/7E/9B/wKiom1cFFtiRtgd3AAGzACbb-dI936.jpg "title=" ARM CEO Simon segars.jpg "alt=" Wkiom1cfftirtgd3aagzacbb-di936.jpg "/>

(For ARM company CEO Simon Segars)

Since the first generation of GSM cellular devices in 1990, ARM has become the preferred CPU architecture for cellular modems. To date, more than 20 billion cellular devices worldwide have opted for arm-architecture CPUs. The ARM architecture has been applied to 85% of the world's smart mobile devices, with more than 95% of smartphones based on arm design.

Now, ARM is becoming the standard for smart hardware and IoT devices.

February 2016 arm released a new processor architecture design, primarily for 5G modems and high-capacity storage soc embedded devices, which will become the basis for future AI (AI) popularization, which can significantly reduce power consumption and cost. In March, ARM CEO Simon Segars told reporters in the new Chongqing Arm Industrial Park that only when the cost and price of AI can be borne by ordinary people will it really achieve the desired goal.

Improve the computing power of automobiles

According to Arm's earnings, in the fourth quarter of 2015, ARM authorized a total of 51 chip licensing, the main applications are: Mobile computing, Smart cars, security systems and the Internet of things. Among them, it is worth mentioning that the direction of the smart car, including NVIDIA, Qualcomm and Japan Renesas Electronics (Rensas) are based on arm design and development of the driver-oriented auxiliary system supercomputer.

Segars believes that the intelligent system on the car is like a supercomputer on a wheel, because it is not possible to get more computing power from the back-end cloud in the car, so it needs to retain the appropriate computing power on the car. How to improve the computing power of automobiles in a compact scale is a great opportunity in the future.

In March 2016, Nvidia announced the first embedded system for deep-learning neural network design nvidia Jetson TX1 landed in China, focusing on applications such as drones and advanced Driver Assistance Systems (ADAS). The core of the Jetson TX1 is the NVIDIA Tegra X1 processor, with 1t-flops architecture for Maxwell Operations, 256-core GPU and 64-bit four-core arm A57 CPUs, 4K Video codec support and up to 1.4 billion megapixel/second camera capture, In some ways it's comparable to supercomputers 15 years ago.

Renesas Electronics in Japan last September announced the production of the R-car T2 SOC Embedded system in 2016, based on the Ethernet AVB video/Audio communication protocol, can achieve 360-degree real-time video surveillance of the car periphery, while significantly reducing power consumption and latency, while providing 1,280x 960 HD Video Real-time transmission.

Popularize AI in Smart hardware

Compared with the research of artificial intelligence in algorithm field, ARM has taken a different direction to artificial intelligence. Segars says Ai has a high demand for computing power, and the industry is still using traditional mainframes for computing power, and arm focuses on how to implement AI decision making and even embedded intelligence in a small portable device.

Chongqing Mayor Huang Qifan at the signing ceremony, said: "All the source of artificial intelligence, concentrated on the chip." ”

In the fourth quarter of last year and throughout 2015, Cortex-m was the most licensed chip license by ARM, with a cumulative license of over 360. The main applications of the CORTEX-M series are intelligent interconnected devices, including microcontrollers, industrial controllers, medical devices, smart sensors, and low-power wireless communication chips. The University of Michigan even developed a CORTEX-M0-based world's smallest computer, with only 0.4mmx0.8mm size.

The ARM Mali image processing series, which is related to the AI display, was also harvested last year. In 2015, the arm Mali-based GPU chip shipped up to 750 million tablets, an increase of 36% compared to 2014. The new generation of Mali GPU chips is performing well in mobile virtual reality (VR), bringing the visual experience of smartphones into smaller smart devices. In March 2016, ARM also launched a development kit for the VR Android Application SDK for Mali.

In the field of intelligent hardware ecology, the arm eco-Industrial park, which cooperates with Chongqing, is an intelligent hardware and software industry cluster which includes chip and sensor design, machine development and manufacturing, Internet of things and big Data application, and both sides jointly carry out cooperation in integrated circuits and intelligent hardware and software. The first companies to be settled include: Chuang Space Accelerator, middle-move Internet of things, Chongqing Horizon AI Robotics Institute, etc.

Among them, the space accelerator is to enrich the arm ecosystem resources as the goal of intelligent hardware and Internet of things acceleration and incubation platform, focusing on AR/VR, robotics and artificial intelligence, smart cars, intelligent medical, smart home and intelligent city, such as six areas.

Designing next-generation data centers

As early as 2011, experts from Nvidia thought that the arm architecture was emerging in the embedded (SoC) domain, and the ability to implement the chip under low power consumption forced the arm's chip designer to go the other way. And that's what the future supercomputers need: Low power, high performance, and optimized cost.

Also in March 2016, arm and TSMC announced the signing of a long-term strategic cooperation agreement for 7 NM FINFET process technology, including a future low-power, high performance computing Soc, which further extends both process technologies from mobile phones to next-generation networks and data centers. Both sides have successfully cooperated in the 16 nm and 10 nano-chip technology, the industry is currently still in the 14 nano-chip technology.

Although the 7-nanometer chip may not be mass-produced as early as the 2018, this is significant for the design of next-generation servers and data centers. The 7 nm process can greatly improve the efficiency of energy utilization and improve the speed of calculation and processing. Segars believes that the current demand for server chips, mainly focused on the emphasis on efficiency and energy consumption, arm and TSMC's cooperation from the final production and manufacturing process to achieve optimal performance and cost.

For the next generation of data centers, there is also a need to address computing requirements for specific applications, resulting in a large demand for highly integrated silicon chips. ARM is also increasing its software investments, such as a custom chip with PayPal, which combines arm processors and digital signal processors (DSPs) to rewrite parts of the code, resulting in a 85% reduction in data center power consumption.

ARM is also actively involved in the Facebook organization's Open source Hardware Alliance OCP (Open Computing Alliance) to redesign next-generation servers and data centers from a new perspective. ARM is currently working with OCP Alliance members to accelerate related software and hardware development. Segars said ARM expects to be able to occupy 25% of the world's data center market share in 2020.

Chongqing Mayor Huang Qifan predicted that in the future in the chip arm will be the mainstream, so working with Arm is the future of the world's giants to cooperate. Considering that about 11% of the country's computers and mobile phones are made in Chongqing, Chongqing is also actively layout smart hardware, Huang Qifan mayor of this prediction can be said to be quite a weight. (Text/Ningchuang, this article first Titanium media, Itvalue)

"More exciting content in the era of cloud technology No.: Cloudtechtime "

This article is from the "Cloud Technology Age" blog, please be sure to keep this source http://cloudtechtime.blog.51cto.com/10784015/1761079

ARM CEO: The car will become a supercomputer, pushing AI into smart hardware