close
close

How Apple’s M-series chips are pioneering AI and energy efficiency

How Apple’s M-series chips are pioneering AI and energy efficiency

Apple is outpacing overall industry progress with its M-series chips. Image credit: apple


How Apple’s M-series chips are pioneering AI and energy efficiency

Apple executives believe that by designing their own devices Apple silicon chips and AI, the company now has a significant advantage over traditional chipmakers that need to serve a wide range of markets and customers.

Apple’s Vice President of Mac Product Marketing Tom Boger and Vice President of Platform Architecture Tim Millet spoke about the new M4 line of chips used in recent Apple product updates in a interview of The Indian Express. The company believes it is its goal own chip design gives it “a huge strategic advantage,” Millet said.

“We are not a silicon trader,” he added, summarizing Apple’s advantage. “We don’t build chips and sell them to other (companies).”

By creating chips that are tailor-made for the devices they are used in, the company avoids compromises in overall performance. Boger added that “no other platform can match our power performance per watt. That is the tangible benefit for users.”

Leading chip innovation

Boger noted the dramatic year-over-year performance improvements as successive generations of Apple Silicon come to market, outpacing the progress of the rest of the industry. The new M4, Apple says, brings customers “the world’s fastest CPU core, delivering the industry’s best single-threaded performance.”

The two executives say there is more to Apple Silicon’s success than just delivering speed with minimal energy consumption. “We benefit from the three key components: the architecture, the design and the process technology,” Miller said.

“Our fourth tool, really our secret weapon, I think, is our ability to design these amazing chips together with the systems teams and the product designers as they imagine possibilities.” Miller pointed to the new M4 Mac mini as an example of this.

“The opportunity was for us and for the design team to come together and build this incredible new platform,” he told the newspaper. “There’s no way that machine would have come to life without that collaboration. And that’s really what Apple is all about.”

Miller noted that competing chipmakers “can’t just move to the latest cutting-edge technology like second-generation, three-nanometer, but we (Apple) benefit from it in a way that we think is worthwhile. It delivers for us and our products and our customers, we try to leave nothing on the table.”

Boger added that it was rare to see the “pace of innovation year after year after year,” noting that the first Apple Silicon chip only debuted four years ago. “That’s the promise. That’s a promise we make to our teams to deliver innovations as they become available to us,” he said.

The rise of the neural motor

Commenting on the rise of artificial intelligence in PCs and Apple’s response to it Apple IntelligenceBoger noted that “intelligent” features have been in Macs for years. He noted that Apple first incorporated a Neural Engine into its program iPhone chip designs in 2017.

Two scientists in lab coats work at a table with a laptop and microscopes, surrounded by laboratory equipment and glassware.

The Neural Engine and M-series chips attract customers with more intensive workloads. Image credit: apple

Millet added that “this was inspired by our recognition of the importance of computational photography. We saw the amazing research that people at the University of Toronto showed (that) these new neural networks were able to do image recognition beyond people’s capacity, or at least fit together, and they were on a trajectory that was clear.”

“And so we took the opportunity to build that built-in capability into our camera processors for the phone,” Millet said. Boger added that the Neural Engine was a core part of the first M1 chip.

“We have a great architecture for AI, and we also have developers taking advantage of Apple Silicon to bring intelligent features to our customers,” he said. “So the M-series chips have always been built for AI.”

Boger said his team saw “an interesting article” in 2017 discussing transformer networks, now the engine behind Large Language Models (LLMs) used in AI. Boger’s team saw that the technology could have a major impact on the Neural Engine and introduced it in the first M-series chips.

“It shows you the dedication with which we spend all our time figuring out where the ball is moving,” Miller said. “We’re trying to make sure we get there before it gets there.”

Innovation driven by users

Boger said in the interview that Apple Silicon continues to push boundaries in performance and energy efficiency “because that’s what our customers do.” He used the M4 MacBook Pro line as an example.

“For example, you run the most demanding workload with it plugged in, and then you unplug it (and) you get the exact same performance.”

Noting the addition of the M4 Pro and M4 Max chips, Millet said that memory bandwidth is a key differentiator over the regular M4. “M4 Max effectively has about twice the memory bandwidth of M4 Pro, which will help someone who really pushed the limits for a very, very large model.”

Millet said Apple is working closely with software partners “to look for the best opportunities to accelerate not only generic benchmarks, which we are often judged on, but, more importantly, based on the workloads we actually deliver to our customers.”

“We know what the hardware system and thermal design will look like, and we understand what the process technology nodes are, and we are aggressively pursuing our best silicon options,” Millet said. “I’ve been doing this for over 30 years (and) it’s the best situation to be in.”