As the demand for smartphones is hitting the roof and the trend of smart homes is gaining momentum, the need for artificial intelligence (AI) chips that can carry out operations efficiently is high. The tech giants and startups have taken various steps to launch new chips that can bring innovation across the globe and facilitate the functioning of smart systems. Intel has collaborated with Facebook to develop an inference chip.
The research and development activities are ongoing. The Taiwanese research institute and a tech firm collaborated to develop the smallest AI chip. The world’s leading tech giant Apple has announced its newest A12 Bionic AI chip with an ability to perform nearly 5 trillion operations per second. These activities have been focused on the development of new chips.
Some of the activities show that the implementation of the chips to improve the performance. Moreover, the Chinese tech giant Huawei has begun utilizing AI chip at its data center to improve the network performance. The market for AI chip is hitting the roof. According to the research firm Allied Market Research, the global artificial intelligence chip market is expected to garner $91.18 billion by 2025. Following are some of the activities taking place in the industry:
Intel & Facebook collaboration for new AI chip
Intel Corp. has collaborated with Facebook Inc. to develop an AI chip by the second half of this year. This will be Intel’s big step in the fastest growing AI computing segment. This chip would assist in what scientists termed as inference, in which AI algorithm performs certain functions such as tagging friends in the photos. The company is expected to face the competition from Nvidia Corp and Amazon.com Inc.
In September 2018, Nvidia offered its inference chip. In November 2018, Amazon also announced that it has developed its own inference chip. However, the chip from Amazon will not present direct competition to Intel and Nvidia as it will not sell its chips. The e-commerce giant will offer services to its cloud client base that would run chips from the next year. However, if Amazon offers its own chips, there would be a major loss of customers for Nvidia and Intel.
Intel outlined that Dell Technologies will equip its XPS line laptops with Intel’s next generation of processors. Its 10-nanometer chips have not been included yet due to delays. Navin Shenoy, the data center chief of Intel, stated that new chips would be available for laptops by the holiday shopping season of 2019 and in data centers by early 2020.
Researchers claim to design the smallest AI chip
The Taiwanese research team and a tech firm claimed that it has designed the smallest AI chip of the world, which has measurements of 7mm by 7mm. The scientists at the National Taiwan Normal University’s (NTNU) Department of Electrical Engineering collaborated with the Taiwan branch of T1 Technologies Inc. for development of the “edge AI chip.” The team also launched a system development kit (SDK), which presents a collection of software that can be utilized for the development of applications for various devices or operating systems. The chip developed by the team consists of a deep-learning convolutional network that can be utilized for various applications.
Professor Wang Wei-yen, one of the team members, outlined that the edge chip can be utilized into mainboards and employed in various applications. This chip has the capability to transform the devices into a smart digital assistant through an AI identification and transmit the results to the master chip. The SDK can be used for subsequent verification and processing. The researchers believe that this innovative chip can shorten the product development cycles, save money, time, and manpower, and transform many regular audio & video devices into AI products.
Apple’s innovative and faster AI chip
The leading tech giant Apple has announced its new AI chip A12 Bionic. A 7nm smartphone chip has nearly 6.9 billion transistors and a neural engine along with a graphics processing unit (GPU), which is 50 percent faster than its A11 chip. It has the capability to perform nearly 5 trillion operations per second. Its system on a chip (SoC) functionality will enable the tech giant to become one of the leading companies in AI chip market across the globe. Its previous versions A11 Bionic and A10 Fusion boasted GPU and neural engine to empower AI accelerators. As the market for smartphones has been evolving rapidly, manufacturers have been developing chips that can offer optimum capabilities.
AI chip to improve the network performance
The data center of Huawei has begun an AI chip to improve the network performance. The Chinese firm has announced its product at its spring 2019 event. This product comes into implementation as the firm faces political pressure and bans in the U.S. and European regions. Kevin Hu, President of Huawei Network Product Line, outlined that its CloudEngine 16800 contains three features crucial to the “data center switches in the AI era.” Along with the embedded chip, there is 48-port 400GE line card per slot. Moreover, it has the ability to evolve with an intent-based network.
The traditional 100GE data center networks are not able to handle the data flood. In addition, the traditional techniques cannot meet the requirements with the increase in a number of data centers and the storage network, data network, and computing network converge. The new switch overcomes these problems with the embedded AI chip by utilizing iLossless algorithm for implementation of auto-sensing and auto-optimization. This offers the high throughput and low latency on the basis of zero packet loss.
Moreover, the computing power limitations can be tackled, the AI computing power can be enhanced from 50 percent to 100 percent, and the data storage Input/Output Operations Per Second (IOPS) can be improved by 30 percent. The AI chip also improves the level of intelligence of devices that are deployed at the network edge along with offering the switch an ability to carry out local inference and real-time decision making. Moreover, it offers analyzer to determine faults in few seconds and locates the faults in few minutes.