Even as Apple shifts away entirely from Intel chips and components on its PCs, its newest custom chip, M2 Ultra, seems to follow a similar philosophy to chips that Intel, AMD and other chipmakers are taking. Intel’s Meteor Lake chip generation, showcased as prototype at the recently concluded personal computing show Computex 2023 in Taiwan on 30 May, featured a dedicated ‘Versatile Processing Unit’ (VPU) — a proprietary, dedicated co-processor within Intel’s Meteor Lake chips, designed to enable local processing of artificial intelligence (AI) and machine learning (ML) workloads.
At Worldwide Developer Conference (WWDC) 2023, Apple’s annual developer event hosted on Monday, the company also seems to be pursuing AI processing. In a press statement and keynote, Apple said that its M2 Ultra custom chip doubles the total amount of memory on a PC that its processors could support, which “can train massive machine learning workloads in a single system, than the most powerful discrete GPUs.”
Both Apple and Intel, therefore, are pitching local, on-device AI processing — which typically takes massive amounts of processing power and memory to work. Apple’s M2 Ultra chip, the company said, uses a proprietary technology called ‘UltraFusion’, which combines the chip die of two M2 Max chips to increase performance. According to a press statement, M2 Ultra can locally process 2.5 terabytes per second of data in a 24-core CPU layout, and support 192GB of memory — which the company said in its keynote as the key factor to enable on-device AI and ML processing.
Apple also added that its M2 Ultra chip features a 76-core GPU that is 30% faster than its predecessor, and a 32-core dedicated Neural Engine that is 40% faster than its predecessor.
Apple and Intel’s announcements come amid a growing interest from companies around the world in AI, specifically generative AI, which has seen a worldwide boom following the public launch of Microsoft-backed OpenAI’s chatbot, ChatGPT, in November last year. Since its unveiling, experts have highlighted a major roadblock in the form of massive amounts of computing power, which is required in order to train AI and ML models required for such AI tools to work.
To be sure, Mint reported on 16 April that while work on generative AI has been undergoing at top Indian engineering colleges for at least four years now, the lack of availability of adequate computing power is a key hindrance in academia pursuing projects in this sector.
While training of these models still remains a task for massive cloud computing infrastructure, Apple and Intel’s announcements in the past week both highlighted increasing the ability of consumer-end devices to locally process AI workloads — with Apple’s press statement underlining the use of the M2 Ultra chip in training ML workloads, locally.
Apple’s latest generation M2 Ultra chip will feature in new generation Mac Studio and Mac Pro desktop PCs, which were also announced at WWDC 2023 on Monday.
Download The Mint News App to get Daily Market Updates & Live Business News.
More
Less
Updated: 06 Jun 2023, 12:12 AM IST
For all the latest Technology News Click Here
For the latest news and updates, follow us on Google News.
