British tech firm, ARM, unveiled new and enhanced chip offerings designed for 5G-capable mobile devices, which offer massive improvements in machine learning (ML) performance.
These chip offerings include the ARM Cortex A77 CPU, the Mali-G77 GPU, and an improved machine learning processor. Aside from enhanced ML processing capabilities, ARM claims that its latest products can satisfy the performance requirements of handsets for increasingly demanding tasks like augmented reality, virtual reality, and gaming.
The Cortex A77 CPU is the latest high-performance chip design from ARM and compared to the Cortex A76 CPU, the newer processor offers up to 20-percent improvement in performance at the same clock speed and process node technology.
Moreover, the new processor continues ARM’s push of improving the machine learning performance of its chip offerings, with the two latest high-performance CPUs offering 35 times improved ML processing performance compared to previous ARM processors.
Aside from boosting the performance of its CPU, ARM managed to continue improving the power efficiency of its chipsets through improvements in microarchitecture. For example, to reduce the power consumed by its CPU, ARM enhanced its chips’ data prefetching capability, which refers to the processors' ability to obtain data before the device actually needs it. Together, improvements on chipset performance allow smartphones to offer features like more secure face recognition, artificial intelligence-powered cameras, and visual scene detection.
Another product that ARM unveiled recently is the Mali-G77 GPU, the first graphics processor from ARM built using the new Valhall architecture. Compared to the previous generation Mali-G76 GPU, the newer graphics chipset offers 40-percent improvement in performance. Furthermore, at the same conditions, the Mali-G77 offers 30-percent better power efficiency, 30-percent improved performance density, and a 60-percent increase in machine learning capabilities in comparison with its predecessor.
Improving the power efficiency of the GPU is an important goal for ARM, with the company aiming to boost the battery life of the devices equipped with its latest GPU even if the handsets process and display complex content like games and augmented reality.
Last but not the least, ARM also announced substantial enhancements to its existing ML processor, with the newer implementation of the chip offering twice the energy efficiency, improved memory compression, and better scaling to multi-core CPU designs.
ARM continues the development of its product offerings with the aim of improving the machine learning capabilities of handheld devices. These improvements allow developers and handset manufacturers to process neural network and artificial intelligence algorithms directly on the device, rather than on the cloud. By processing data on the handset, users will experience lower latency when dealing with machine learning-powered tasks, and it also improves the security of the user, since the information no longer has to be transported and stored in a remote data center.
While the performance and power efficiency improvements claimed by ARM are based on the 7nm process node technology, the enhancements made by foundries to their respective fabrication technologies should further improve the capabilities of the new chip designs from ARM.
Moving forward, consumers should expect that the new chip designs from ARM will appear on smartphones starting next year.