In March this year, the newDynamIQ TechnologyAt Computex 2017, ARM unveiled the higher-performance Cortex-A75 processor architecture design and the Cortex-A55 processor architecture design, which is targeted at a wide range of applications. It also unveiled the Mali-G72 GPU, which is targeted at virtual reality, machine learning, and higher-performance display applications. ARM also provided developers with a technology licensing package that included development tools such as the ARM Compute Library.
According to ARM, by redefining multi-core architecture design and integrating DynamIQ technology for AI applications, it is expected that ARM architecture processors will help drive the growth of AI technology with performance improvements of more than 3 times over the next 5-50 years. In addition to specifically improving the computing performance of mobile devices, it will also specifically expand the growth of technologies related to virtual reality, deep learning, autonomous driving, and richer multimedia experiences. ARM TrustZone technology will also ensure the security of devices that currently cover 70% of global usage.
The newly launched Cortex-A75 processor architecture design not only introduces DynamIQ technology to increase overall computing performance by more than 50%, but also can increase the computing performance of mobile devices by more than 73% compared to the previously launched Cortex-A20 architecture, while maintaining the same basic power consumption performance as the Cortex-A73. It even improves basic computing performance by more than 72% compared to the Cortex-A40, and can be used in laptops, network communication equipment or server devices.
The Cortex-A55, designed for a wider range of device applications, will be positioned as the successor architecture to the Cortex-A53. In addition to incorporating DynamIQ technology, it also integrates instruction sets related to artificial intelligence applications, resulting in approximately 53 times higher computing efficiency than the Cortex-A2.5. It can also form a big.LITTLE core combination with the Cortex-A75, thereby achieving higher execution performance and more battery-efficient operation. DynamIQ technology also enables more flexible computing core configurations and core architecture design flexibility, such as achieving core combinations such as 4 big and 4 small, 1 big and 7 small, or 4 small.
The Mali-G40 GPU, designed with the Bifrost architecture and offering over 72% performance boost, is designed for high-end gaming and deep learning applications. It can also handle more stable mobile VR and AR applications, as well as a wider range of computer vision calculations. According to ARM, the Mali-G72 GPU improves battery life by 25%, increases machine learning efficiency by 17%, and boosts computing efficiency by 20% compared to previous products. It also reduces data bandwidth usage by 68% in gaming applications.
ARM stated that the future development trend of mobile devices will gradually move towards high-performance computing and low power consumption. At the same time, they will be able to complete device-side deep learning, always-on connectivity, and meet the needs of HDR and high-resolution image output. They will also further promote the integration and development of technologies such as virtual reality and augmented reality.
To enable the growth of ARM architecture-based design applications, ARM has announced the release of a new suite of development tools and technology licenses, including the ARM Compute Library. This is expected to enable ARM architecture-based hardware to drive 10-15 times the growth of artificial intelligence and machine learning.
