Apple makes a fantastic product, including some of the best laptops in the world, but sometimes it can feel like Apple’s biggest innovation year after year is hype.
And so now that Apple has announced its new Apple M5 chip that will power the 14-inch MacBook Pro, 11-inch and 13-inch iPad Pro, and Vision Pro headset, a simple look at the press release is enough to see the Apple hype train clearing the station.
Will the Apple M5 be a good chip? Almost certainly, considering that its predecessor, the Apple M4, was a good chip, too, and the Apple M5 is more akin to what’s already on the market than “the next big leap” Apple is always promising.
What’s new about the Apple M5
That’s not to say that there’s nothing new in the Apple M5, because there is. First, Apple is introducing “neural accelerators” into its GPU cores, something that Intel, AMD, and Nvidia have been doing for a couple of generations now. This should help improve graphics performance in games and applications that choose to leverage those AI components.
For the GPU compute units themselves, Apple says the “next-generation” GPU delivers up to 30% faster performance compared to the M4 and up to 2.5x faster performance than the M1.
For its CPU component, the M5 will get up to 15% better multicore performance than the M4 with similar core configurations, though no word on the single-core performance.
The M5 will also feature an expanded bandwidth for its unified memory of up to 153GB/s, which is a roughly 30% increase over the previous M4’s, and frankly says a lot more about memory-intensive graphics performance than anything about its new architecture.
Why none of this is likely to matter
To understand why I am very underwhelmed by this chip, even without seeing a single independent benchmark, is the second sentence of Apple’s press release announcing the new chip, which begins: “Built using third-generation 3-nanometer technology…”. That’s certainly one way of saying that you’ve been using the same TSMC N3 process node for the last three generations of your processors, severely limiting the kind of gains you can get from performance.
Second, I’ve covered a lot of processor announcements and launches, and every company claims that their new chips get “up to” some impressively high-percentage better performance in this category or that. AMD, Intel, and Nvidia, however, actually show you what those benchmarks are, which lets us, as journalists, hold them accountable. What’s more, I’ve sat through any number of Intel and AMD briefings and announcements that show games and applications that don’t show any performance improvement at all.
Apple is nothing if not completely opaque when it comes to its performance claims. Saying the Apple M5 can “deliver up to 15 percent faster multithreaded performance over M4,” is functionally meaningless if you don’t know what application or test got that score. Likewise, its performance claims regarding its GPU. If anything, this chip should be called the Apple M4 Plus, or even the Apple M4 AI. It would be more honest.
The only indication we get from Apple’s announcement about its performance claims is a footnote that reads, in part: “Performance measured using select industry‑standard benchmarks. Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Pro.”
Apple has done this kind of thing for years, but in the case of the M5, it feels worse since I get the sense that Apple is trying to hide the ball here from consumers and journalists in order to hype up a chip that isn’t likely to see the kinds of overall performance gains that Apple wants to tout and that customers expect.
That’s because the only real add here is neural processing capacity that, quite frankly, won’t make all that much of a difference in the small smattering of apps that use Apple’s neural engine or new GPU neural accelerators. Apple isn’t an AI company the way Nvidia is. Apple makes very fine computers and designs very good processors, but few will turn to a MacBook to run actual on-device AI workloads that matter.
The neural accelerator-enhanced GPU cores and neural engine on the M5 likely don’t have the juice to compete with even a two-generation-old Nvidia RTX 3060 laptop GPU.
Meanwhile, since this is also about the MacBook Pro and iPad Pro, actual “Pro”-level AI work is done on Nvidia GPUs, end of story. CUDA is the language of today’s AI, and the entire AI ecosystem is built around it. CUDA is also a proprietary software framework that is exclusive to Nvidia GPUs, and so, try as they might, AMD, Apple, Intel, and Qualcomm simply cannot break that industry lock-in with Nvidia.
Apple has its own AI framework that it is encouraging its developers to use, but this runs the risk of those developers locking themselves into Apple’s framework, which is different and incompatible with what everyone else in the field is doing, which makes no sense for an AI professional.
So ultimately, the biggest sell for this chip and the products it will power is negligible in most cases, while the actual day-to-day functionality won’t be noticeably different than an Apple M4 chip you might already have in your laptop, or which will now be cheaper if you are looking to upgrade this year, so will be a much better value than anything with an M5 in it.
Should Apple have just waited a year or so to launch the M5?
Yes.
Apple has a good working relationship with TSMC and Samsung, and TSMC and Samsung’s respective 2nm process nodes will be ramping up production in 2026. It could have fabbed the Apple M5 on a more advanced node next year and seen real, across-the-board performance gains.
It chose not to because it’s under pressure right now to “catch up” on AI and is attempting to do so when just about everyone—including OpenAI’s Sam Altman—is acknowledging that we are probably in the midst of an AI bubble. How that will work out for Apple and everyone else remains to be seen, but what it does show is that Apple is struggling right now to find the kind of direction that it had when it made the decision to launch its own in-house silicon, and is now chasing after ideas instead of innovating.
Which makes sense. Apple’s been coasting on the move to its M-series for half a decade now, but the car is starting to sputter a bit as TSMC hasn’t been able to get its 2nm node into production in time to keep pace with Apple M-series releases, leading to three generations of Apple M-series chips using the same 3nm process technology.
Apple needs something more than just hype right now, and I’m not sure what it is, but it’s not going to be the new Apple M5.
Read more about the latest Apple news…
https://cdn.mos.cms.futurecdn.net/SDh7V2HvhLF4PjefPVEDyR-2000-80.jpg
Source link
John.Loeffler@futurenet.com (John Loeffler)
