ARM releases new chip designs aimed at AI, machine learning

Softbank-owned silicon designer ARM has released two new mobile processor variants targeting artificial intelligence and machine learning as well as augmented and virtual reality.

Source: ARM

Based on its ARMv8.2-A architecture, the Cortex-A75 and Cortex-A55 designs use the company's DynamIQ technology, allowing them to be clustered with as many as eight processors in one system.

The Cortex-A75 design promises 20 percent better performance than the earlier Cortex-A72 and Cortex-A73 processors, along with faster single instruction multiple data (SIMD) and floating point units, making it suitable for machine learning applications, the company said.

Fitted with a new high-performance memory subsystem, the Cortex-A75 could be used in laptops running Windows as well as massively scalable servers, encroaching on fields traditionally dominated by rival chip vendor Intel.

ARM's Cortex-A55 design promises 15 percent better power efficiency compared to the existing Cortex-A53 processor, yet provides up to 18 percent better performance and improved scalability for mobile devices.

The company also released a second-generation graphics processing unit design, the Mali-G72, based on its Bifrost architecture.

The Mali-G72 is similarly faster than earlier Bifrost-based GPUs, with up to 40 percent better performance while slashing energy use by up to a quarter.

ARM said the 850 MHz Mali-G72 - with up to 32 shader cores - enables console-quality gaming on mobile devices, supporting up to 16 times multi-sampling anti-aliasing for graphics.

The GPU also being designed for augmented and virtual reality applications, and ARM said the chip design offers 17 percent improved efficiency for machine learning.

Mobile chip makers and device vendors such as Qualcomm, Samsung and Apple utilise ARM processor designs which power billions of smartphones and tablets.

The British company was acquired by Japanese tech giant Softbank last year in a blockbuster A$40 billion deal.