A few days ago, ARM Holdings (NASDAQ:ARMH) announced its next-generation low power processor IP core known as the Cortex A72. This is the successor to the Cortex A57 from 2012, and ARM says the A72 brings solid performance-per-watt improvements over its predecessor.
According to a post on the Real World Technologies forum by Peter Greenhalgh, ARM's director of technology, the Cortex A72 increases performance per clock over the A57 by anywhere from 10%-50% while also reducing power consumption. Greenhalgh also noted the Cortex A72's power improvements are "achieved on the same process with the same library as Cortex-A57" and that the efficiency improvements come from design enhancements.
Longtime followers of Apple's (NASDAQ:AAPL) chip technology likely know that while the company licenses the ARM instruction set -- which allows it to design chips that run the same code as ARM's own Cortex A-series designs -- its core designs are custom. I don't think Apple is going to drop its own core in favor of a Cortex A72. Given Apple's focus on performance, this is actually a good thing.
You don't build custom cores for "fun"
Designing and validating increasingly complex high-performance, low-power CPU cores is difficult and expensive. ARM designs among the best mobile processor cores available today, so a company needs a darn good reason to embark on a custom design.
Now, a company such as ARM -- which licenses cores to many customers aiming for many different uses -- must build "one size fits all" cores. These chips need to offer good performance within a reasonable power budget, and to fit in a certain silicon area on a given manufacturing technology.
If a company feels developing a custom core better fits its performance, power, and area objectives, and has the money to throw at the project, then it will try to do so. That's what Apple does with its A-chips.
Apple likely feels it can do "better" than the A72
If Apple is sticking with a custom CPU core for the A9 (and I would bet heavily that it is doing so), then the Apple team must feel it can build something better for its needs than the Cortex A72. Since the A72 and whatever Apple names its next core principally target performance-oriented smartphones, my guess is the CPU core inside the A9 will offer improvements over the current "Enhanced Cyclone" inside the A8 similar to the improvements the A72 brings over the A57.
In fact, since ARM continues to design its "big" CPU cores for use in quad-core big.LITTLE configurations, while Apple seems to like using just two cores in its iPhone chips, I wouldn't be the least bit surprised if the next Apple core is much more powerful than the A72 on a per-core basis. The trade-off here could be that Apple's core would use more silicon area than does the A72 (but Apple's margins on the iPhone are so high that this is likely lost in the noise).
At any rate, given how impressive the performance of the A72 seems to be, based on ARM's comments, I can't wait to see what Apple will do with its own next-generation custom CPU core slated to show up in the 2015 iPhone.