Apple M4
Apple (Hacker News, Slashdot):
Built using second-generation 3-nanometer technology, M4 is a system on a chip (SoC) that advances the industry-leading power efficiency of Apple silicon and enables the incredibly thin design of iPad Pro. It also features an entirely new display engine to drive the stunning precision, color, and brightness of the breakthrough Ultra Retina XDR display on iPad Pro. A new CPU has up to 10 cores, while the new 10-core GPU builds on the next-generation GPU architecture introduced in M3, and brings Dynamic Caching, hardware-accelerated ray tracing, and hardware-accelerated mesh shading to iPad for the first time. M4 has Apple’s fastest Neural Engine ever, capable of up to 38 trillion operations per second, which is faster than the neural processing unit of any AI PC today.
[…]
M4 has a new up-to-10-core CPU consisting of up to four performance cores and now six efficiency cores. The next-generation cores feature improved branch prediction, with wider decode and execution engines for the performance cores, and a deeper execution engine for the efficiency cores. And both types of cores also feature enhanced, next-generation ML accelerators.
M4 delivers up to 1.5x faster CPU performance over the powerful M2 in the previous iPad Pro.
[…]
M4 can deliver the same performance as M2 using just half the power. And compared with the latest PC chip in a thin and light laptop, M4 can deliver the same performance using just a fourth of the power.
Previously:
Update (2024-05-08): Jason Snell:
Why the M4 now? It mostly has to do with Apple shifting chip production at TSMC (the company that fabs Apple’s chips) from the first-generation 3nm process to a new, more efficient second-generation 3nm process. There’s a whole backstory about TSMC’s change in 3nm processes that’s not worth getting into here, but suffice it to say that the first-generation process is largely a dead end, and the company is moving to a new set of 3nm processes.
As expected, the performance “gains” of the new M4 chip Apple is using in the new iPad Pros are mostly due to the N3e process. Apple advertises a “1.5x” speed gain: but they slyly compare the prior 8-core Pro M2 to the new 10-core Pro M4 (25% more
performancecores, right off).
Update (2024-05-10): Omar Sohail (via Hacker News):
An early look at the M4’s performance did not deliver the best positive first impression because we believed that Apple lowered the clock speeds to achieve better efficiency. However, we are pleasantly surprised by the latest results, as Apple’s new SoC powering the 11-inch and 13-inch iPad Pro models runs circles around the M2, handily beats M3, and zips past the M3 Pro and Qualcomm’s Snapdragon X Elite, which are two chipsets occupying a higher performance bracket.
Apple's M4 processors have become convincing leaders in the Geekbench single-core leaderboard. Several scores of roughly 3,800 points have appeared in the Geekbench online database over recent hours. This is significant as single-core benchmark scores of this magnitude put clear blue water between the M4 and Intel’s flagship Core i9-14900KS. A little Geekbench database checking shows that, in single-threaded tests, Apple's M4 outpaces Intel's power-hungry desktop champ by about 16%.
Apple said that the M4 delivers up to 1.5x faster CPU performance than the M2 in the prior-generation iPad Pro, which is accurate based on the benchmarks we’ve seen so far.
Update (2024-05-16): See also: Hacker News and MacRumors.
4 Comments RSS · Twitter · Mastodon
But why is this a thing? Nick Heer quoted someone saying it was because TSCM (or whatever the ored of the letters are) had to retool their factory and Apple had no other choice than to upgrade their babysitters with killer chips.
People speculated that TSMC's processes were the cause, and Apple's marketing claim of "second-generation 3nm process" seems to confirm that. Which does not alone explain why they brand it as an M4. They have in the past changed processes while leaving processor generations.
I think the other piece in the puzzle is really just a mix of marketing and AI hype. They wanted an SoC they could tout as AI-focused, so that's gonna be the M4.
How does the M4 actually differ from the M3? Very unclear to me. We do know that memory bandwidth is up from 100 GiB/s to 120, and that cores are up from 4+4 to 4+6. I have a sense the process change means we can't directly infer that this leads to (slightly) faster performance, though; given that the new process is allegedly "worse", and that the M3's clock was already quite high, I wonder if the M4's clock is slightly lower, which those spec changes then offset, in order to lead to roughly the same performance.
There's no real indication that the CPU/GPU cores have changed at all.
The only exception is the Neural Engine, where they seem to _suggest_ they've made improvements, but even there, it's unclear if that's in comparison to the M2, or the M3. (Which, fair enough. Their marketing here was firmly towards iPad Pro purchasers, and an M3 iPad Pro never existed.)
Well, I got deets in that post wrong! The 2 additional cores aren’t Performance cores, they’re Efficiency cores. But I suppose the surprising gains reflect the admirable performance capabilities of those Efficiency cores in Apple silicon, as shown by Max Tech several times. It’s going to be interesting to see what numbers we get when M4 jumps to the Mac, specifically the frequencies and thermals; and I hope it happens sooner than later. It was said during the event hat M4 was giving same performance at half the power, which indicates quite a bit of runway to boost performance at same power, and we know how much more power efficient N3e is over N3. Which may really highlight how disappointing M3 was and why Apple dragged their feet so long before finally adopting it (and likely only doing so begrudgingly and only thanks to really, really good pricing from TSMC).
"[…] This doesn’t sound to me like a man about to announce iPad Pros with M4 chips. […]" - John G., Apple Pundit.
https://daringfireball.net/linked/2024/05/02/cook-comments-analyst-call
Oops…