ARM, the UK-based semiconductor design firm, introduced a new chip targeted at markets ranging from self-driving cars to artificial intelligence.
The new design, DynamIQ, is an update to ARM’s existing Cortex-A offering. The latter made up almost 20 percent of all ARM-based chips in the first three quarters of 2016, according to the company’s last available regulatory filing.
The new design may help ARM, bought for $32 billion last year by SoftBank Group, compete with chips engineered for neural networks, a promising type of artificial intelligence software. Rivals like Intel and International Business Machines (IBM) have recently unveiled chips designed for these applications.
It could also give ARM a better chance of making inroads against Intel in the lucrative server and data center market.
SoftBank Chief Executive Officer Masayoshi Son snapped up ARM to bet on the Internet of Things — the idea that everything from refrigerators to industrial robots will be connected to the Internet in the future. Son hopes ARM will become the leading designer of chips in these devices.
The new design by ARM is aimed at higher-end IoT applications — such as autonomous cars and robots — that must handle complex computing tasks, such as running AI software, locally while communicating with other devices and remote servers.
Energy efficient
ARM is well known for designing energy-efficient chips for power-constrained devices, such as mobile phones. More than 85 per cent of the world’s most sophisticated smartphones include ARM-based chips.
The new design can handle up to eight processor cores of varying size on a single chip in almost any configuration. That will give customers more flexibility than ARM’s existing designs, Nandan Nayampally, general manager of the company’s Compute Products Group, said.
This is especially true in cases where a device has to switch rapidly between different tasks, for instance, using neural networks for facial recognition one moment and then handling a voice call.
He said DynamIQ would be more efficient than existing architectures because the processors can share memory and switch rapidly between tasks with different power requirements. The technology will also work well in devices like industrial robots and self-driving cars that require high levels of safety and redundancy and have to process most computing tasks locally, Nayampally said.
Data centre
The new system could also be configured to handle difficult computing tasks in data centres, a market long-dominated by Intel and where ARM-based chips have only slowly started to catch on.
When coupled with special ARM software, the new design will perform as much as 50 times better on AI tasks than the existing chips within three to five years, he said.
Late last year, Intel unveiled a chip aimed specifically at AI applications. Google has also developed its own AI chip design, and several startups have announced rival offerings.
ARM believes most systems will need a combination of general processing alongside specialized AI capabilities, Nayampally said.