Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Found M4 and M5 Mac identifiers
- Thread starter Jamie I
- Start date
- Sort by reaction score
Ever since the M3 Pro, I have the fealing that Apple has building blocks to put together a chip. CPU, NPU, GPU, I/O, etc.
So for their server chip, they put together exactly what they need, what is best.
For the Ultra, they put two chips together to look like one to applications. These are applications that are proposed on the server anyway. Do they need the chip we want in our consumer machine?
Last edited:
M4 is out for a while and according OP there are already M5 identifier. Would expect more rumors about M5.
I am curious to see how Gurman times it, my guess is he will stay silent about it until early 2025, when he will use it ahead of whatever event or press release they have scheduled.
The problem right now is no one knows, really, if it’s the Air or the iMac, or something else. We’re just guessing (edit: my current guess is in the comment directly below). Agree with the OP that it’s likely the Air, but Gurman won’t say anything until he can plausibly claim he got it from a source. That’s what’s most important to him (or rather, to his employer, Bloomberg News), maintaining the illusion that he has sources, when in reality it’s mostly just gossip.
Last edited:
I've posted this elsewhere with the identifiers listed in order (which appears to correspond with Apple's internal product-development roadmap), but this version is arranged by model, as you can see.
I believe it is correct re: the M3-generation MacBook Pro identifiers, with GB/s memory bandwidth as the distinguishing feature. If there are really only twelve M4-generation Mac identifiers, then they are probably going back to having just one memory bandwidth for the Max:
M2
Mac14,2 :: M2 100 (8/8, 8/10) MacBook Air 13"
Mac14,15 :: M2 100 (8/10) MacBook Air 15"
Mac14,3 :: M2 100 (8/10) Mac mini
Mac14,12 :: M2 Pro 200 (10/16) Mac mini
Mac14,5 :: M2 Pro 200 (10/16, 12/19) MacBook Pro 14"
Mac14,9 :: M2 Max 400 (12/30, 12/38) MacBook Pro 14"
Mac14,6 :: M2 Pro 200 (12/19) MacBook Pro 16"
Mac14,10 :: M2 Max 400 (12/30, 12/38) MacBook Pro 16"
Mac14,7 :: M2 100 (8/10) MacBook Pro 13"
Mac14,8 :: M2 Ultra 800 (24/60, 24/76) Mac Pro
Mac14,13 :: M2 Max 400 (12/30, 12/38) Mac Studio
Mac14,14 :: M2 Ultra 800 (24/60, 24/76) Mac Studio
M3
Mac15,3 :: M3 100 (8/10) MacBook Pro 14"
Mac15,4 :: M3 100 (8/8) iMac (Two ports)
Mac15,5 :: M3 100 (8/10) iMac (Four ports)
Mac15,6 :: M3 Pro 150 (11/14, 12/18) MacBook Pro 14"
Mac15,8 :: M3 Max 300 (14/30) MacBook Pro 14"
Mac15,10 :: M3 Max 400 (16/40) MacBook Pro 14"
Mac15,7 :: M3 Pro 150 (12/18) MacBook Pro 16"
Mac15,9 :: M3 Max 300 (14/30) MacBook Pro 16"
Mac15,11 :: M3 Max 400 (16/40) MacBook Pro 16"
Mac15,12 :: M3 100 (8/8, 8/10) MacBook Air 13"
Mac15,13 :: M3 100 (8/10) MacBook Air 15"
M4 (projected)
Mac16,1 :: M4 120 (8/8) iMac (Two ports)
Mac16,2 :: M4 120 (10/10) iMac (Four ports)
Mac16,3 :: M4 120 (10/10) MacBook Pro 14"
Mac16,5 :: M4 Pro MacBook Pro 14"
Mac16,7 :: M4 Max MacBook Pro 14"
Mac16,6 :: M4 Pro MacBook Pro 16"
Mac16,8 :: M4 Max MacBook Pro 16"
Mac16,9 :: M4 Ultra Mac Pro
Mac16,10 :: M4 120 (10/10) Mac mini
Mac16,11 :: M4 Pro Mac mini
Mac16,12 :: M4 Max Mac Studio
Mac16,13 :: M4 Ultra Mac Studio
M5 (projected)
Mac17,1 :: M5 MacBook Air 13"
Mac17,2 :: M5 MacBook Air 15"
I believe it is correct re: the M3-generation MacBook Pro identifiers, with GB/s memory bandwidth as the distinguishing feature. If there are really only twelve M4-generation Mac identifiers, then they are probably going back to having just one memory bandwidth for the Max:
M2
Mac14,2 :: M2 100 (8/8, 8/10) MacBook Air 13"
Mac14,15 :: M2 100 (8/10) MacBook Air 15"
Mac14,3 :: M2 100 (8/10) Mac mini
Mac14,12 :: M2 Pro 200 (10/16) Mac mini
Mac14,5 :: M2 Pro 200 (10/16, 12/19) MacBook Pro 14"
Mac14,9 :: M2 Max 400 (12/30, 12/38) MacBook Pro 14"
Mac14,6 :: M2 Pro 200 (12/19) MacBook Pro 16"
Mac14,10 :: M2 Max 400 (12/30, 12/38) MacBook Pro 16"
Mac14,7 :: M2 100 (8/10) MacBook Pro 13"
Mac14,8 :: M2 Ultra 800 (24/60, 24/76) Mac Pro
Mac14,13 :: M2 Max 400 (12/30, 12/38) Mac Studio
Mac14,14 :: M2 Ultra 800 (24/60, 24/76) Mac Studio
M3
Mac15,3 :: M3 100 (8/10) MacBook Pro 14"
Mac15,4 :: M3 100 (8/8) iMac (Two ports)
Mac15,5 :: M3 100 (8/10) iMac (Four ports)
Mac15,6 :: M3 Pro 150 (11/14, 12/18) MacBook Pro 14"
Mac15,8 :: M3 Max 300 (14/30) MacBook Pro 14"
Mac15,10 :: M3 Max 400 (16/40) MacBook Pro 14"
Mac15,7 :: M3 Pro 150 (12/18) MacBook Pro 16"
Mac15,9 :: M3 Max 300 (14/30) MacBook Pro 16"
Mac15,11 :: M3 Max 400 (16/40) MacBook Pro 16"
Mac15,12 :: M3 100 (8/8, 8/10) MacBook Air 13"
Mac15,13 :: M3 100 (8/10) MacBook Air 15"
M4 (projected)
Mac16,1 :: M4 120 (8/8) iMac (Two ports)
Mac16,2 :: M4 120 (10/10) iMac (Four ports)
Mac16,3 :: M4 120 (10/10) MacBook Pro 14"
Mac16,5 :: M4 Pro MacBook Pro 14"
Mac16,7 :: M4 Max MacBook Pro 14"
Mac16,6 :: M4 Pro MacBook Pro 16"
Mac16,8 :: M4 Max MacBook Pro 16"
Mac16,9 :: M4 Ultra Mac Pro
Mac16,10 :: M4 120 (10/10) Mac mini
Mac16,11 :: M4 Pro Mac mini
Mac16,12 :: M4 Max Mac Studio
Mac16,13 :: M4 Ultra Mac Studio
M5 (projected)
Mac17,1 :: M5 MacBook Air 13"
Mac17,2 :: M5 MacBook Air 15"
Last edited:
I have a hunch that Apple can now introduce a yearly schedule for its Mx devices. Considering the iPad Pro’s schedule of one and a half years, I anticipate seeing numerous M5 devices in October 2025.
Gurman said no Mac Studio until mid-2025. I would be very surprised if Mac mini and Mac Studio is refreshed ahead of MacBook Pro...
Gurman: No New Mac Studio and Mac Pro Until Mid-2025
Apple will not refresh the Mac Studio and Mac Pro with next-generation high-end chips until the middle of 2025, according to Bloomberg's Mark...www.macrumors.com
My timeline prediction:
October 2024 event
- M4/M4 Pro Mac mini
- 24" M4 iMac
- 14"/16" M4/M4 Pro/M4 Max MacBook Pro
March 2025 event
- 13" and 15" M4 MacBook Air
WWDC25
- M4 Max, M4 Ultra Mac Studio
- M4 Ultra Mac Pro
One of the points for ditching Intel was to be able bring new processor updates quicker to market.
The GPU of M4 introduced hardware-accelerated ray tracing.
Studio is targeted at high powered video editing on the desktop.
So why leave it out from getting that hardware-accelerated ray tracing for more than a year?
By WWDC2025 we'll probably be at M5 and 3 generations on from where Studio is now.
I suspect if it is delayed that long it'll be as a sacrifice to the cash cows of MBP, Mini, iMac, Air etc.
TSMC will no doubt be churning out base M4, Pro & Max.
Could they ditch Ultra to update Studio if there's no capacity for that highly specialised chip?
I suspect it's the Ultra chip fab that holds back Studio if anything is.
How much of the Mac line up refresh is now dictated by TSMC's capacity, tooling and costs?
It is possible, even likely, that the lack of an M3 Ultra is a result of a decision by TSMC to not support InFO (Integrated Fan-Out) advanced packaging (used for UltraFusion) for the first-generation 3nm process.
If so, it explains a lot. If Ultra was never in the cards for M3, then Apple was free to experiment with other things, like two different memory bandwidths for M3 Max, as well as to focus on GPU features in M3, as you mentioned (note your “M4” typo).
But that would, then, just be a one-off. N3E and N3P support the full range of TSMC’s advanced-packaging technologies, so UltraFusion for M4 Max and M5 Max is probably in the cards. Plus, there’s no reason to think N2 and beyond won’t also support it.
Apple may be TSMC’s only customer that is using the InFO chip-first approach to advanced packaging, but InFO is also being incorporated into chip-last designs used by Nvidia, AMD, and so on. So there are plenty of incentives to invest in that approach, even if it is (apparently) unique to Apple in terms of actual, shipping products.
EDIT to add a comment on your last question, a very good question. I don’t think we know, but obviously it’s dependent on it, if not truly “dictated” by it. TSMC’s partnership with Apple is longstanding, disruptive, and highly valuable, and we can probably be certain, for example, that Apple would be involved on the highest levels in a strategic decision affecting them, like the one mentioned above.
Last edited:
According to this article back in May N3E was in volume production that coincided with M4 for iPad Pro launch with N3P scheduled for 2nd half of 2024.
Which might explain why no Mac updates until October due to them waiting for N3P, which if N3P supports InFO, would include the Ultra variation.
I think the OP's Mac17,1 & Mac17,2 are most likely to be the two Studios and unlikely to be launched in October with any variant of M4. Why else give them a major version number bump?
Although I could be wrong and they are larger iMacs to cover >= 27".
Apple hasn't got a great track record with keeping it's flagship Mac updated.
Which is nuts given the HW ray tracing in M3+ as well as how much AI is used in video editing these days.
Seems to me that Apple considers the MacBook Pro as its flagship Mac product. For Apple, desktop are not even close to being a primary concern.
It almost is, with comparable benchmark numbers, why would you buy a Studio when you can get a keyboard, screen and trackpad in addition to the machine that can almost keep up for the same money?
EDIT : See this post I wrote last week.
I've had my eye on the refurbished Studio M2's for a while. They aren't shifting.
The M3 MBP's on the other hand sell quickly.
It's the flagship in as far as revenue is concerned.
Last edited:
N3E volume production didn’t “coincide” with the launch of M4 in May 2024. N3E volume production began in the 2nd half of 2023 (4th quarter, but we only learned that detail later on). So M4 launched at least six months after the start of N3E volume production.
This timeline is also what’s expected for N3P. Volume production will begin in the 2nd half of 2024, and M5 will probably launch in the 1st half of 2025.
Keep in mind the fact that Apple is on the “bleeding edge” (quoting the Anandtech article you cited) of TSMC’s nodes. Tape outs for M4, M4 Pro, and M4 Max would have been completed in 2022, more than a year before N3E went into volume production.
It’s not clear from your comment that you understand the basic premise of the OP’s prediction (and my comment #32 above, which provides details that support the premise), which is that all M5-generation Macs will have “Mac17,x” identifiers, and all M4-generation Macs will have “Mac16,x” identifiers. So the “major version number bump” you refer to is simply the next generation of Apple silicon.
Could it be the Studio? It’s not inconceivable, but it would be unusual and it would probably mean it wouldn’t launch until the 2nd half of 2025, even if announced earlier.
Their record with regard to the Ultra Macs is good, if we assume (as I do) that M3 Ultra doesn’t exist due to factors not in Apple’s direct control. The Ultra is always going to lag behind the Max on the roadmap, that’s the reality of Apple’s chip-first approach to advanced packaging. The Max comes first.
? I'm a little confused by this statement. As far as I can tell, indeed the full M2 Ultra doubles the full M2 Max: 2x(8/4+38) (P/E CPU+GPU) -> 16/8+76? Unless you mean something else?
This thread isn't really giving us any new information. Every Mac will likely see an M4 update, it's not too surprising and at this point I'm happy if Apple gives us the M4 updates already and doesn't try to gap the update cycle with giving current M2 models the M3 based SoCs that are outdated already.
And whether there even will be a M3 Ultra, or a M4 Ultra for that matter, is a question relevant to very few users and not necessarily because nobody wants the kind of performance the Ultra provides, but because this design has inherent scaling problems.
At least on the CPU side for most apps there is little to no benefit going beyond the 16 cores of a M3 Max. Apps that can even parallelize to the extent of using 12 performance cores let alone all 16 threads are few and far between. Even though there is a lot of software that can max that out with loads of computations running independently of each other, such as media conversion/transcode, very rarely will you have such a workflow that relies on maxing out all those resources for hours at a time. (I do and I run these on a dedicated machine, sometimes over night, as it can take days to finish, I don't run that on my main machine.)
What most workflows benefit greatly by is single core performance and that's another weakness of that ARM based architecture. It's definitely the best chips to have especially in mobile devices like the Macbooks, but amongst desktop computers in terms of raw performance these SoCs are just average and the Ultra especially so when taking hardware cost into account. For most users they certainly aren't worth paying such a premium for.
The GPU scales almost linearly but even then you'll find overall performance to be average.
And whether there even will be a M3 Ultra, or a M4 Ultra for that matter, is a question relevant to very few users and not necessarily because nobody wants the kind of performance the Ultra provides, but because this design has inherent scaling problems.
At least on the CPU side for most apps there is little to no benefit going beyond the 16 cores of a M3 Max. Apps that can even parallelize to the extent of using 12 performance cores let alone all 16 threads are few and far between. Even though there is a lot of software that can max that out with loads of computations running independently of each other, such as media conversion/transcode, very rarely will you have such a workflow that relies on maxing out all those resources for hours at a time. (I do and I run these on a dedicated machine, sometimes over night, as it can take days to finish, I don't run that on my main machine.)
What most workflows benefit greatly by is single core performance and that's another weakness of that ARM based architecture. It's definitely the best chips to have especially in mobile devices like the Macbooks, but amongst desktop computers in terms of raw performance these SoCs are just average and the Ultra especially so when taking hardware cost into account. For most users they certainly aren't worth paying such a premium for.
The GPU scales almost linearly but even then you'll find overall performance to be average.
no? not in the CPU department which is what you seem to be claiming ...
It most definitely is not ... the little M4 has the highest ST performance of any P-core, desktop included, released so far.
In what workloads?
The GPU is actually the part that doesn't scale as well to desktops, not the CPU.
That ARM based way of doing things is to parallelize as much as possible. With most software there quickly comes a point where additional cores can only help so much and what is actually needed is single core performance. Most users would never see any benefits going from 12 cores to 24. Either apps just aren't optimized for that many cores or computations rely on previous computations meaning that the next computation cannot start until the currently ongoing computation finishes.
There is no telling when M4 Macs are going to hit the shelves. Apple might start out with a base M4 SoC Mini and refresh the passively cooled Airs as well. People to whom the extra performance matters for actual work will buy a Studio or MBP. And those might not refresh until 2025. The M4 might be super fast but what are you gonna do with that on an iPad? You are comparing unreleased next-gen computer hardware to the current-gen Intels/AMDs.
If you buy a 100W+ desktop computer chip in 2025 you'll have plenty of M4 alternatives. Just look at the M2 Ultra and where that slots in compared to the Intels and AMDs of the time. It was the most expensive way to go but not that fastest, neither single threaded nor multi and certainly didn't have the fastest graphics compared to Intel/AMD computers with dedicated graphics.
In...any workload? Compare a high-end 13th gen Intel with a RTX 4080/90 or something to the M2 Ultra and you'll find the Ultra is the worse yet more expensive option. I mean it's no surprise when these computers require 1000W power supplies and more - when you are looking for the "ultra" performance the M2 Ultra is unlikely to be what you'll be looking for.
Huh? The amount of GPU cores you can choose when buying M SoC based Macs scales pretty much linearly with GPU performance.
Ugh, you are right. Not sure how I missed that. I've edited my comment #32 to include that...
The whole point of the original thread is to point out that there is evidence that not “every Mac is likely to see an M4 update,” that two Mac identifiers will go straight to M5 and skip M4. That is new information.
It’s just a list of M4 and M5 identifiers found in the Sequoia builds, but it’s something. It’s possible the M4 list is incomplete, but if it is complete then it is two short, which just happens to be the number of M5 identifiers on the list. Maybe coincidental, but certainly thread-worthy.
No ... according to whom? If anything, the "ARM based way" is to have fast, efficient ST performance cores with no SMT. If anything, historically x86 has shined in parallel workloads buoyed by SMT, which Intel is now dropping because according to them it no longer makes sense even for x86. So you kinda have your thesis backwards here ...
EDIT: wait are you confusing instruction level parallelism within a thread with multithreaded parallelism? Because, yes the former is definitely a hallmark of how Apple managed to produce a very fast ST core that was also efficient. The superiority of ARM in doing ILP but with lower clocks is "the ARM based way" for even ARM and Qualcomm and Apple. I'm genuinely trying to figure out what you're talking about.
The Ultra class chip is literally designed for those people who need exactly that. And you better not tell that to AMD or Intel each of whose top-of-the-line "9-series" chips come with 32 threads ...
No telling? You do know Apple has pretty standard events every year around September for iPhones and October/November for Macs. We may not know which exact models will launch but we have pretty good idea about when M4 Macs are coming ...
But why you just said that only single core performance matters? So since for the Mac those all basically have the same ST performance (okay the top line M2 Max/Ultra got a tiny bump) why not just get the mini?
The M4 exists it is not unreleased hardware ... we know what the performance characteristics are and AMD just released its newest desktop lineup and we know the M4 even in the iPad is faster. AMD aren't going to be releasing a new set of desktop processors in the next couple of months. Intel's Arrow Lake is unreleased but if it's faster, it won't be by much, and judging what we've seen from Lunar Lake so far and given Intel leaks and promises it might not be faster at all, again depending on workload.
But why would you do that? 100W+ desktop computers come with all sorts of parallel CPU cores which you just said nobody needs ...
Hence why I said Apple's GPU was underwhelming for a Desktop ... and you disagree with below for some reason
Yes ... and that's why it is underwhelming for the desktop ... which you agreed with above saying: "certainly didn't have the fastest graphics compared to Intel/AMD computers with dedicated graphics." Desktop GPUs on PC don't scale linearly with laptop GPUs because Nvidia and AMD also crank up the clock speeds which yes wastes a lot of power that can't be wasted in the mobile design, but results in greater desktop performance. Thus, Nvidia and AMD scale desktop chips in both cores AND clocks. Meanwhile, Apple's advantages in TBDR GPUs isn't as pronounced as its CPU microarchitecture advantages so it can't quite translate the same "desktop-level performance, mobile-level power" for the GPU that it can for the CPU and it is limited by reticle area and die costs in scaling the GPU by core count alone. So, for desktops, Apple GPUs fall behind in relative performance to AMD/Nvidia for many, but not all workloads.
Interesting ...
1000W power supplies aren't to support 35W ST performance. They're there to support hundreds upon hundreds of watts of unrestrained MT and GPU performance. So what's actually important to you? Anyway here is GB below, I could do the same with CB R24, but I'm tired and not going to because you said embarrassingly parallel CPU workloads were unimportant even for someone buying an Ultra even though those are actually the workloads the Intel and AMD processors excel at:
ASUS System Product Name - Geekbench
Benchmark results for an ASUS System Product Name with an Intel Core i9-13900KS processor.
browser.geekbench.com
ASUS System Product Name - Geekbench
Benchmark results for an ASUS System Product Name with an Intel Core i9-14900K processor.
browser.geekbench.com
Mac Studio (2023) - Geekbench
Benchmark results for a Mac Studio (2023) with an Apple M2 Ultra processor.
browser.geekbench.com
Gigabyte Technology Co., Ltd. X670E AORUS MASTER - Geekbench
Benchmark results for a Gigabyte Technology Co., Ltd. X670E AORUS MASTER with an AMD Ryzen 9 7950X processor.
browser.geekbench.com
MacBook Pro (14-inch, Nov 2023) - Geekbench
Benchmark results for a MacBook Pro (14-inch, Nov 2023) with an Apple M3 Pro processor.
browser.geekbench.com
Yeah the overall ST score is low in comparison - but again depends on workload, remember GB is a collection of individual subtests and in fact the M2 Ultra actually wins more than a few, especially those that are not AVX(-512) heavy (which matters less now that benchmarks can make use of Apple's matrix coprocessor). And the Ultra's MT scores is roughly equivalent here to the Intel and AMD processors ... and if you complain that GB 6 MT tests aren't embarrassingly parallel in order to reflect real world MT use cases, remember your baseline is that massively parallel CPU tasks are outmoded ... and if it's ST that's all important, why wouldn't I just buy an M3 mini with nearly as good ST performance as the best Intel and AMD models? The M3 above is the Max variant but the base M3 has the same ST performance and the Max is useful to demonstrate something about MT performance of the M3 family.
Oh also about any workload and price ... just out of curiosity how much does an Nvidia GPU with more than 24GB of VRAM cost? You might find that edifying ...
Look the reality is Apple never released an M3 Ultra, there are some hypothesis as to why, a couple in this thread even, but they didn't. However, to test your theory about ARM, we can still extrapolate with napkin math for what it would've look like at 24 P cores and 8 E cores with 80 GPU cores. The ST performance would been within a few percent of the best AMD/Intel chips and the MT performance would've about the bottom rank of modern thread rippers. Sure the GPU would've done say Blender ray tracing rendering benchmarks at about the level of a 4080 (less actually) which is indeed underwhelming for the price (at least for small scenes, again massive VRAM pool not typically reflected in benchmarks and those are incredibly pricy from Nvidia), but otherwise the CPU would've been pretty dam competitive in just raw performance even before we get to the ancillary benefits of perf/W, heat, noise, and long term processor stability.
The M2 Ultra was Apple's 2nd homegrown desktop processor of the new Apple Silicon era. According to Gurman, and I take this with a mountain of salt, Apple will be doing a desktop-specific die, dubbed "Hidra", in the near future. I'm not going to hold my breath, but even without that we can see the development trajectory of where they're taking their processors based on the M3 and M4. Apple competes extremely well in the CPU space in performance in both desktop and laptops in both ST and MT. The development of these chips simply doesn't line up with your theories at all. In fact it's hard to tell what your theories are because your posts are also filled with self contradictions about the importance of ST, MT, and GPU performance that's it's hard to gauge what it is you actually want to see or think doesn't scale about Apple's cores? This conversation really reminds me of the early days of the Apple Silicon, especially post A7, and people coming up with all sorts of excuses for why AS couldn't scale and would never compete with PC chips. Hell same thing post M1 even prior to the release of the M1 Pro/Max which took a long time come out. And yet here we are.
Last edited:
I wonder if he's convinced himself that this is the "ARM way" because there are Arm server chips where they pack up to 192 Neoverse cores into one chip? Low performance per core, lots of cores is a thing that's popular in datacenters.
If so, @okkibs , please understand that implementation choices are not always hints about the fundamental nature of an ISA. They are often just engineers optimizing a product for the market it's supposed to sell into.
Also, if that was your idea, if you'd been paying attention you'd still know it was an incoherent bad idea because Intel and AMD both build server chips with exactly that design philosophy - low performance per thread, lots of threads. For example, Intel's recently launched Xeon 6 family has 144-core models which not only peak at 3.0 GHz, the cores are what they call an "E core" when used in desktop SKUs.
The single thread performance on these is not great relative to desktop x86, but you get a lot of threads. There is actually a market for that - lots of internet servers scale best with thread count, since there are so many clients hammering on them at once, and each client needs its own thread.
You are contradicting yourself here. Boosting the number of physical cores/threads to improve performance is the common strategy of the x86 makers. We see Intel making CPUs with 24 cores and 32 threads to claim that performance crown. Similarly, AMD uses 16 cores and 32 threads. In fact, it is all over the industry, since scaling single-core performance is hard. Another fact: Apple uses less CPU cores in their configs compared to similar designs because Apple is a single core performance leader. The M3 Max has only 12 performance cores and 4 efficiency cores (so an equivalent of 13 P-cores aggregated), yet it handily outperforms 16+ core mobile workstation solutions both in single and multi-core.
Besides, there is no such thing as “ARM way”. Different designers use different strategies. Apple in particular uses excellent single-threaded performance without product segmentation. The M3 core has the same performance across all the products. This is very different from the x86 product strategy where you have a range of models with different configuration. And it certainly costs Apple some performance in the enthusiast class segment as they don’t sell factory overclocked CPUs unlike other companies.
Yes, Apple currently fails to convince in the high-performance desktop segment. There are multiple reasons for that. None of them have to do with ARM architecture however. They are consequences of Apple business decisions. Will this change in the future? We will have to wait and see. M4 is the first chip where Apple comfortably outperforms even enthusiast-level desktop CPU cores, and it doesn’t seem like others will be able to catch up in the next year. So an M4 Ultra will be in a much better position today than M1 Max was on release, for example. And it is still possible, although unlikely, that Apple will develop a desktop-focused product. With their single-core performance leadership they certainly have a theoretical advantage in this area. Of course, having a theoretical advantage is not the same as having a business need.
Still, everything points to M4 Max laptop outperforming Zen5 9950X, which would be quite sad for the x86 desktop state of the art. Zen5 retains advantage in some vector processing workloads, unless of course your code can take advantage of SME.