February 14, 2018 // 11:14 a.m.
Cambridge-based low-power chip specialist Arm has announced its formal entry into the machine learning (ML) acceleration industry, announcing the Trillium platform for energy-efficient acceleration on edge devices - beginning with future smartphones.
Artificial intelligence and machine learning are hot topics, and companies working on increasing their performance while reducing their power draw are doing well for themselves in this latest boom. Arm, now a subsidiary of Japanese technology giant SoftBank, is fully aware of this and eager for a slice of the pie with its latest platform, Trillium.
'The rapid acceleration of artificial intelligence into edge devices is placing increased requirements for innovation to address compute while maintaining a power efficient footprint. To meet this demand, Arm is announcing its new ML platform, Project Trillium,' explains Rene Haas, president of Arm's intellectual property (IP) products group. 'New devices will require the high-performance ML and AI capabilities these new processors deliver. Combined with the high degree of flexibility and scalability that our platform provides, our partners can push the boundaries of what will be possible across a broad range of devices.'
Project Trillium begins with a pair of products, which in traditional Arm fashion are being offered as soft IP which its customers can have fabricated with or without modification once licensed from Arm. The Arm ML processor concentrates, as its name implies, on machine learning acceleration, offering a claimed 4.6 trillion floating point operations per second (teraflops, t/FLOPS) with a further performance boost between two and four times in real-world scenarios through an intelligent data management system and a power efficiency of three t/FLOPS per watt. The Arm OD, meanwhile, is designed for computer vision tasks including facial and object recognition with a claimed 80x performance boost over traditional digital signal processors (DSPs) and the ability to process Full HD video at 60 frames per second.
The Trillium name, Arm advises, is temporary, with the commercial launch due in mid-2018 under new branding. The first preview IP, meanwhile, will be available to Arm's customers in early April, along with specialised neural network software designed for use with Arm ML, Arm OD, and the company's traditional Cortex CPU and Mali GPU products.
More information is available on the Arm blog.