M7
It's a shame they seem to have dropped the ball on other types of cores.
M7 could do with an update.
Arm has confirmed it is working on a CPU core expected to deliver a jump in performance, thus taking aim at the closing the gap between its own chips and those produced by Apple. The Brit chip design house is developing the next iteration of its Cortex-X line of CPU cores to be the most powerful available at launch, which …
If ARM is preparing an offering to compete with the Apple M-series of processors I'd say: "Don't bother."
Apple is able to do this because they control the entire hardware and software stack of their Macs, whereas Microsoft only makes the software bits. Also, Microsoft has a huge legacy of software it can run on its x86 based operating system and isn't likely to switch to another processor platform that could eliminate its backwards compatibility advantage.
Also, it seems that the migration to a non-Intel platform has adversely affected sales of Mac computers. I'm unsure if Apple will be willing to backtrack on its move to its own processors, but if they don't it may become a "bet the company" decision.
Mac sales have increased compared to when they used Intel CPUs, but they are down from the initial peak when they introduced the Apple Silicon models. When you have a lot of people upgrading at once, that obviously depresses sales subsequently as people aren't replacing PCs every year or two.
I think most people make the "Mac vs PC" decision first, then choose from available options. Few people were switching from PC to Mac because of Apple Silicon, even if it performed better than any Windows laptops in power/performance. Likewise few will leave Mac if the equation changes and you can get better options from Windows laptops, regardless of whether they are x86 or ARM.
The performance increase is due to Apple tightly integrating DRAM memory with the processor, physically bonding the two together.
Not so. Using LPDDR actually REDUCES performance compared to DIMMs for anything sensitive to memory latency which typical tasks are to some degree at least. The LPDDR standard is designed for low power and sacrifices memory latency to achieve it. Tasks sensitive to memory bandwidth (i.e. GPU) benefit from LPDDR.
I think Apple would be able to switch to that new LPCAMM2 standard and get essentially the same performance. The only catch is that since they increase the memory channels as they step up you'd have to fill 4 LPCAMM2 slots for a "Max". Not sure if they have that amount of board space for a Macbook Pro with the Apple Silicon Max, so while it would be possible for them it may not be practical.
The memory is not physically bonded. It's just much closer to the CPU and it's technically possible to upgrade memory by replacing memory chips. It's actually relatively simple, but you need a few tools.
Problem is that Apple unlikely is going to make those chips available so that people can do upgrades.
Yes you can do this at home. I've done it many times just not with Apple.
If you get pre-balled new chips, it's really few minutes to do replacement (I am not counting time to disassemble and reassemble the laptop).
If you are replacing reclaimed chips, then you need to reball them which is a bit of faff, but doable as well.
You need a microscope, hot plate, hot air station, good tweezers, flux, wick, solder paste and steady hands. You can get professional quality tools for under a grand.
No, Apple's work is totally independent of ARM. ARM provides the ISA specs (instructions and what they do) but implementations are the responsibility of architectural licensees like Apple. ARM has no ability to use Apple's technology in their designs. And vice versa, since Apple is not licensing ARM's core designs.
> Don’t bother.
Well, it’s just about the phones. But why not make an Mx competitor? Everyone talks about vertical integration but actually Asahi Linux goes like the clappers on Apple Mx, even ARM Windows in a VM is nippy proving that Apple Mx chips are great general purpose chips and someone else could do the same. In fact Qualcomm/Nuvia are trying to do just that, though it’s obvious that the Qualcomm Snapdragon X Elite is originally a server chip hastily re-moulded for PCs but it’s a start. ARM do have cores trying to compete closer to this space but they’re not quite “there”.
Well, actually I’m dreaming about RISC-V on the desktop but if it happens it’ll take a bit longer. Companies like Qualcomm and AMD (yes, they’re looking at it again) are working on Aarch64 at this time..
Oh, don't get me wrong. I am firmly in the ARM camp as the only bulwark against the stagnating and increasingly creaky world of x-86/64. Yes...yes...I know....AMD EPYC/Zen.....Intel iSomethingMeaningless/Xeon_whatever.
I mean....that world is crowing about how from 2024 and onward AMD and Intel chips will come with their own NPUs....(yawn).
But let's get really real. No matter how close ARM comes to developing an Apple Silicon killer, computing, regardless of the platform, from Supercomputer to wearable is NOT just about the hardware. It's about the firmware that undergirds said hardware. It's about the kernel of the OS that runs on said hardware. It's about the software that runs on said hardware.
And it REALLY depends on the support of not only the hardware designers but the OS designers. And you will never.....never....get the theoretical performance of said hardware such as this new ARM X SoC when the hardware designer and the OS designer are not on the same team much less the same company.
And you will never get the theoretical performance maximum out of this new ARM X using Windows or Android. That's doubly true when the support lifetime of any given Windows computer and particularly Android phone or tablet is relatively short as compared to an Apple product.
Used to see this on the old gaming consoles. Early games were great to play, but when the model was going to be replaced with the 2/x/one/3, etc, the quality of the games being developed on the older console were so far superiors as the developers had learnt the secrets and tweaks. Took them ages, but they get ther. Apple has that straight away.
As for Windoze. I guess the devs write on the biggest/fastest they can get and then TEST (if they do any) on the same kit.
Years ago at one company we developed on the fastest kit we had, but our test rigs were the slowest out of date piles of crap we had. We got the functionality going and then set about optimising, rewriting - I don't see that happening these days