Interesting. I’m also thinking now that Apple’s tweaking to determine what frequency provides which yields and then that’s what they produce. It could be more of a balancing act where the power used and the heat produced are a factor of the performance and reliability curves they’re looking to hit. Like you say, it could be that some solution in the future might require more cycles, but, with the current way they’re fabricating chips, we can expect that the highest single threaded performance will be always be reasonably close to the lowest single threaded performance.
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.
M2 is really just m1.1 (Update, I was wrong).
- Thread starter macguy360
- Start date
- Sort by reaction score
If Apple has the ability, I don't see any reason for them not to. Technology will continue to push forward, and if Apple doesn't keep pushing the envelope, their competitors will.
So in essence you largely agree that the existing M1 SoC is already complicated enough, so it is probably to Apple's advantage to not make it more complicated, from the point of view of the SoC architecture and macOS development?
I don't think anyone is saying Apple is purposely limiting it's system performance, probably other than yourself?
My view is that Apple is designing for a target performance level and would like all their Macs to have that as a baseline and scale from there, i.e. the end user feel of how responsive their system is will be identical from the MacBook Air all the way to the Mac Studio, and to do it such that it doesn't introduce too much complexity into the macOS code base, which is already complicated enough.
Apple could have boosted the M1 Ultra single core clock to a ridiculously high rate, but doesn't mean that they should. Engineering is always about trade off. Maybe Apple thought that the benefits of such an action does not justify the effort and sales?
Oh ... anyone can write software with the proper tool and base knowledge. There are school curriculums for kids that does programming now compared to say 20 years ago. Doesn't mean it's easy, but anyone can do it with the right mindset.
Coding for an OS on the other hand, is really hard to get right.
I'm personally glad they didn't. On the Mac Studio (assuming they even could raise clock speeds that high), the power consumption would be a non-issue and they could draw as much as they needed. But raising the bar that high for single threaded performance would pressure them to do the same on their notebooks, and the same kind of power consumption on MacBooks would negate some of the biggest advantages of Apple Silicon (and throw battery life out the window).
I want Apple to push the envelope forward and to continue to raise single threaded performance, but I like the strategy they've used so far to focus on IPC rather than just raising clocks. I hope they don't reverse course on that.
I’d imagine Apple’s OS folks are glad they don’t have to consider something as poorly performing as an i3 for any future features
Wait, why? And do you think hiding their performance potential from customers so they don’t want it, is the right competitive move?
Is there something inherently bad about high frequency clocks? What is the frequency at which a clock becomes “ridiculous”? I would have thought performance was performance— there’s not good performance and bad performance. As I keep repeating though, I’m not sure why everyone keeps assuming higher performance means “over clocking”. What if you could increase the IPC at the cost of power?
“What customers really love about the all new Mac Pro is that it provides them the same great performance they’ve come to expect from the iPad Air.”
Yeah, especially considering the iPad Air also ranks in the top 25
The issue people are pointing out is that power consumption doesn't scale linearly with clock speed. It scales exponentially once you reach higher clock speeds, and it's the result of having to raise the voltage sharply in order to reach these kinds of clocks. It's a lot cheaper to raise the IPC by improving the chip's design than it is to rely on ratcheting clock speeds through the roof.
Apple likely couldn't even raise the clocks this high at all unless they wanted to redesign their chips (and the chip's low-level design is just as important as the fab for this kind of thing). But even if Apple did, I'm not sure it would be a good thing. The pentium 4 relied on raising clocks through the roof as well, but we know the colossal failure that strategy was. Intel themselves recognized it when the Core 2 Duo slaughtered it with significantly lower clock speeds a few years later.
Let’s pretend I understand the physics, the logic design challenges, and that there’s multiple ways to improve performance.
Why aren’t people upset that Apple is clocking their design a 3.2GHz instead of 2.8GHz? Because everyone is assuming that’s some kind of “natural” frequency and having a kneejerk reaction that saying “the high end desktops can support higher power processors” means “I saw on YouTube you can overclock it like crazy if you have a dewar of LN2”.
If you have more thermal headroom, you have more performance headroom. Just because Intel got backed into a corner and kept ramping up MHz to compete with AMD in the day, doesn’t mean every effort to get the most from a machine is somehow dirty.
Anyway, this is going in circles.
Makes sense. On the desktop, it's a perfectly valid point and makes sense for a lot of users. Why waste potential by clocking a processor lower than its thermal limit when you have plenty of cooling headroom in the desktop to handle it?
I think that raising clocks is eventually inevitable, but I am personally glad they've not done this yet. Let's suppose they release a Mac Studio with an M3+ Ultra in a few years, and it's clocked at 4GHZ. Now that's the top tier performing machine, and because there is a market for those who want the very best that exists, there will be inevitable market pressure for Apple to do the exact same thing in laptops (partly how we got an i9 MacBook Pro to begin with). If Apple were to embrace that strategy (and they've done this before), they would create a situation where getting the very best single threaded performance required chasing diminishing returns at a huge power expense (we don't want a repeat of the i9 MPB where even the non-turbo base clocks couldn't be sustained because of thermal overload).
For most users, that's completely impractical. And most users would be recommended to stay away from these kinds of notebooks in favor of the lower tier ones. But if Apple were to try to push the power consumption to its absolute thermal max in its notebooks again, it wouldn't be a terribly good look for a company that doesn't exactly have the best reputation for its cooling systems. That was one of the best things about Apple silicon. It was able to totally slaughter its competition with barely a blip on the radar in terms of power consumption, and you always got the very best that Apple Silicon had to offer (whether you were plugged in or on the go). x86 can't claim that. X86 can't even get close.
It's inevitable that eventually the desktop will be clocked higher. I think it's the smart move in the long run, but I do hope that Apple doesn't reverse course on their strategy. They showed us that we didn't have to choose between great battery life and great performance. They gave us both, and that's part of what made Apple Silicon so popular to begin with.
Last edited:
At the moment, there’s not much pressure to draw more power in laptops— for one thing, they’re already about as good as the competing desktops, and for another Apple has gotten enormously good press on their battery life.
I don’t see hiding the fact that they can make a faster desktop as the solution to that problem though. Frankly, if people want a luggable studio, I’m not so paternalistic as to deny them the ability to buy one. I’ll be disappointed if Apple stops making the style of laptop we have today but I don’t think they will though.
The product line today is a little odd— there is little if any reason to buy a 13” MBP. It’s essentially the same machine as the Air but you can shave a few seconds off a media encode which seems like small potatoes. Now they’re talking about a 15” Air. The Studio machines are clearly designed to dissipate more heat than they actually do today. It looks to me like we’ll see more differentiation in performance in the line in the future— top performance in the Studio, the Air will be the type of machine we love today, the MBP’s with the Max chips might start to carry more grunt.
And I think the Pro is going to be yet a different beast.
If they keep the same chip lineup they have today for future generations, then it would make sense to make the Max slightly different in core architecture from the Pro and pick a different point on the power/performance curve. The base Mx chip and the Pro would be optimized for battery life and the Max (and Ultra being 2x the Max) would be optimized for higher performance. The real MBPs currently has a choice between Pro and Max, so you could choose whether you want your MBP optimized for battery or processing. The desktops would be Max or Ultra where power doesn’t matter.
It also kinda depends on whether Apple is targeting single threaded or multi threaded performance on their future Mac Pro. They can honestly probably target both and do it reasonably well. Apple's efficiency cores are about 1/3 as fast as their performance cores (at least on the M1), but they consume about 1/8 of the power. All they need to do is pack "only" 16 performance cores (at their full potential at 3.5ghz+ or so, use the power where it's needed), and then pack a huge number of efficiency cores (32, 48, or even 64 or 96 of them) for multithreaded performance gains.
As ludicrous as that would be (and as terrifying as the transistor counts would be), they could absolutely do it in their power budget. Intel would have no prayer of ever catching up for many years.
Last edited:
I’m actually curious… Apple’s focus on mobile processors is because that’s where they make their money (with the Mac, iOS and iPad). They take what they create there (which performs admirably) add a few extra cores to give it a multithreaded performance edge over the mobile processors and there’s the desktop. Given that they have a pretty good idea how many units of a completely separate chip line (not iterated from the Mx) would sell (maybe in the tens of thousands a year) AND how much developing and maintaining a separate line would cost, it’s possible that they’ve already ruled it out. We may not see much higher clocks because they’re continuing to iterate on the IPC and just “coring” that up for the lesser selling desktops.
I'm glad that I simply bought an M1 and then an M1 Max machine near the beginning of their life-cycle (after a brief period to assess the real-world reviews) and don't have to fret about the improvements, or lack thereof, in the next generation.
I think nearly everyone would agree that there was a significant improvement between the last Intel versions and the first Apple Silicon version of the respective models, and this seems to be the sweet spot in terms of getting access to these improvements before getting into a never-ending loop of "version N+1 will be so much better" procrastination. Getting on the train at the first station seems to be a more satisfactory choice than trying to second-guess whether the next "station" will be worth waiting for.
I'm enjoying my M1 machines for work and play (and have been for some time), and only have a mild academic interest in whatever M2 brings. It will no doubt be slightly better than M1, but not enough to tempt me to upgrade. I expect my next upgrade will be M3 or M4 Pro/Max or whatever they call it.
Those who fret about their expectations of technology are just missing out on actually using it to do stuff.
I think nearly everyone would agree that there was a significant improvement between the last Intel versions and the first Apple Silicon version of the respective models, and this seems to be the sweet spot in terms of getting access to these improvements before getting into a never-ending loop of "version N+1 will be so much better" procrastination. Getting on the train at the first station seems to be a more satisfactory choice than trying to second-guess whether the next "station" will be worth waiting for.
I'm enjoying my M1 machines for work and play (and have been for some time), and only have a mild academic interest in whatever M2 brings. It will no doubt be slightly better than M1, but not enough to tempt me to upgrade. I expect my next upgrade will be M3 or M4 Pro/Max or whatever they call it.
Those who fret about their expectations of technology are just missing out on actually using it to do stuff.
Same here, love my M1 / M1 Pro. My M1 Pro is my work horse and works pretty well, I'm just a tad disappointed I went with 16GB instead of 32GB, this way I can't play much with VMs.
But other than that I been loving the performance and battery out of it! I am hoping we can reach Air levels of battery on our M1 Pros by the time the M3 Pro lands.
Yeah, it’s possible they’ve written off the desktop or they try to somehow make a desktop out of mobile chips, but I hope not. Anchoring the top end on the desktop is important to the overall product line from both a customer and marketing perspective.
There are a number of professionals that use Apple mobile hardware because they use a Mac for heavy compute on the desktop. If they are forced to switch to another OS because it runs their workloads faster, they’ll likely abandon the mobile stuff too. I think this is why Apple keeps the truly Pro hardware and software alive. For those users, they’re willing to pay a premium for the performance, so there may be fewer units sold, but they can sell them at higher prices keeping the product line relevant in their portfolio.
There’s also a marketing importance to having a flagship device at the high end— as long as they don’t their competitors can continue to claim they have the fastest chips, and Apple is forced to say “our are better once you read the list of caveats”. It’s a clearer message to be able to say they’re trouncing the industry in all form factors.
Multithread performance can only get you so far, and not all workloads are amenable to running on infinite cores.
And I’m pretty sure the desktop Macs sell more than 10’s of thousands. They sell 25 million Macs a year. I suspect more that 0.1% of them are desktops.
My theory on all is - assuming recent rumours are correct - is that Apple was happy to design M2 in its current manner because that was the most cost effective way of introducing this chip for its intended hardware.
A few interesting things to note about M2; it has more cores (in addition to the media engines), said cores are clocked higher, and it's a physically larger die. All this added up shouldn't make sense when the process node continues to be 5nm, which means that the chip will draw a little more power and thus generate more heat. Reviews have backed this up.
However, there is one caveat with all this - the reviews thus far have only been for the 13" MacBook Pro. The M1 chip was used in a wide range of products, but those Macs are either going to be, or have been redesigned.
This is important to note on the recent topic of performance, because it's arguably not the best subject for analysis. Yes, there's clearly something going wrong with the fan curve and when it kicks in, but M2 and the 13" MacBook Pro were also clearly not designed for each other. The fact that Apple continues to sell it as their second best selling notebook is telling that they can get away with this.
The same can't be said for the other Macs that will potentially use M2 vanilla, which include a Mac mini (which has more than adequate thermal capacity) and the new MacBook Air (we will wait and see, but I'm sure Apple has figured this out for an entirely new design...)
And the 24" iMac? Rumours have suggested that it will skip M2 and go to M3. But if this doesn't turn out to be true, then the machine has two fans on the mid and higher tier models that should be adequate.
Finally, what about the M2 Pro/Max variants? Again, rumour suggests that these will be produced on 3nm processes, which will offset the increase in performance and make for a series of chips that will be just as efficient as the M1 series.
Just my two cents for what it's worth.
A few interesting things to note about M2; it has more cores (in addition to the media engines), said cores are clocked higher, and it's a physically larger die. All this added up shouldn't make sense when the process node continues to be 5nm, which means that the chip will draw a little more power and thus generate more heat. Reviews have backed this up.
However, there is one caveat with all this - the reviews thus far have only been for the 13" MacBook Pro. The M1 chip was used in a wide range of products, but those Macs are either going to be, or have been redesigned.
This is important to note on the recent topic of performance, because it's arguably not the best subject for analysis. Yes, there's clearly something going wrong with the fan curve and when it kicks in, but M2 and the 13" MacBook Pro were also clearly not designed for each other. The fact that Apple continues to sell it as their second best selling notebook is telling that they can get away with this.
The same can't be said for the other Macs that will potentially use M2 vanilla, which include a Mac mini (which has more than adequate thermal capacity) and the new MacBook Air (we will wait and see, but I'm sure Apple has figured this out for an entirely new design...)
And the 24" iMac? Rumours have suggested that it will skip M2 and go to M3. But if this doesn't turn out to be true, then the machine has two fans on the mid and higher tier models that should be adequate.
Finally, what about the M2 Pro/Max variants? Again, rumour suggests that these will be produced on 3nm processes, which will offset the increase in performance and make for a series of chips that will be just as efficient as the M1 series.
Just my two cents for what it's worth.
Is it important to the overall product line, though? The majority of folks buying computers want some form of mobile system. There’s nothing about a stationary computer that excites the majority of the computer buying public, regardless of how fast it completes some arcane benchmark that means nothing to them.
Yes, there ARE a number. And that number that haven’t ALREADY moved on, is small indeed. Apple’s been marketing heavily to the growing mobile computing market and that means less real concern about the extreme high end. The bottom could drop out of the high end desktop market and it wouldn’t move the Mac revenue needle by a noticeable amount.
Apple doesn’t have to trounce the industry in all form factors, though.
They currently (and, due to AMD’s and Intel’s business practices, will continue to) trounce the industry in the most important growing form factor, mobile. They’ll make a desktop processor, sure, because there are enough folks that will still buy it, but it’ll always be based on an extension of their mobile hardware.From information previously provided by Apple, assuming 25 million, 80% of what they sell is mobile, MBA, MBP, etc. That’s 20 million. Of the 5 million left, 80% of those are iMac/iMac Pro, (which they no longer make, so plug in a few low spec Mac Studio’s there). The remaining desktops that are NOT iMacs would be that last million.
Anything Mac Pro level has already been confirmed by Apple to sell in the single digit percentage a year, and that fits right… and likely in the low single digit percentage (since 5% is 1.25 million and there’s not that many left in the last group), so somewhere between 1-3%. And, one can assume that the most expensive highest spec’d machine would be just a sliver of that.
As the numbers of potential customers for ever higher performing systems get lower and lower, it makes sense why Apple’s just capitalizing on the amazing performance of their bread-and-butter mobile systems and not making any effort for a custom solution that might perform a few percent better, but would sell well under 1,000,000 units in a year.
I'm not sure why people get so angsty about "mobile". Yes, M series chips use mostly the same building blocks (CPU cores, etc) as Apple's mobile chips. This is perfectly fine, you can make a great workstation chip this way:
In the 1990s, Intel conquered the workstation and server markets with x86, even though x86 was already correctly regarded as an archaic and horrible ISA. They started the decade as a joke relative to the incumbents, and ended it as the dominant force. They did it without designing special CPU cores exclusively for servers or workstations. They simply used the enormous cash flow generated by the Wintel duopoly to keep making their PC CPUs better and more capable than they strictly needed to be for PCs, and used that to attack upwards.
Apple's been doing much the same with Apple Silicon. The iOS devices Apple Silicon was originally created for gave Apple an enormous R&D budget. They've been using it to improve their mobile CPU and GPU core designs far beyond what's necessary for mobile. The M1 and M2 families are the fruits of that long term project.
Yes, thanks to different design priorities, Apple isn't offering a huge spread in single thread performance between the lowest power phone CPUs and highest power workstation CPUs. Who cares? If the ST performance they offer is good enough, it doesn't matter. And so far, it has been.
Because, the things AMD and Intel have been putting out as “mobile” chips have performed poorly compared to their desktop variants, and this is their intent.
They need to make their customers believe that putting a “mobile” chip in a “desktop” is a “bad idea”, when Apple has clearly shown that, if it’s a performant “mobile” chip, why wouldn’t you?In order for Intel to do something similar, they’d have to kill everything but, say, the i9’s. From the cheapest to the most expensive would just be i9’s of varying core counts. But, of course, they can’t do that.
Even their mobile i9’s are SO much different from their desktop i9’s that, if you’re in the market for a Windows laptop and want the BEST performance, you’ll have to buy one with a desktop i9 in a mobile form factor. This is the kinda thing that folks have been taught is “the way things should be”. It will take several years of Apple continuing to iterate on their designs for people to understand it fully.