I disagree. The one-year is arbitrary. It might have more to do with people and seasonal cycles than annual cycles. You could argue why not 11 or 13 months too. More time would be nice for additional QA.
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.
The yearly M-series release cadence
- Thread starter Bodhitree
- Start date
- Sort by reaction score
I disagree. The one-year is arbitrary. It might have more to do with people and seasonal cycles than annual cycles. You could argue why not 11 or 13 months too. More time would be nice for additional QA.
Nvidia doesn't roll out new GPUs annually but around every 2 to 3 years. Innovation cannot be scheduled in a MS project plan
A great example of this is Apple's own AI goals, promises, where they announced a number of AI initiatives in the 2024 WWDC where they failed, so much so the iPhone 16 was marketed with a number of AI services that never materialized and they're facing a class action suit.
Don't get me wrong, I think Apple has done a fantastic job with apple silicon, but like the iPhone, big updates/improvements slows down as the technology matures. There will come a point where fab improvements, die shrinkage and other improvements with the current technology slows down.
Moore's law stopped being a thing in the 2010s due heat and power limits, foundry node limitatins, and problems (particularly with intel trying to move off of 10nm and in 2019 Jensen Huang explicitly stated that “Moore’s Law is dead” in terms of cost efficiency
Apple has a significant advantage over Intel in this regard because all of the process/node development and retooling is handled by TSMC rather than by anyone in Cupertino. Look at how the M-series has moved from 5nm to the 3rd iteration of TSMCs 3nm process in five years while Intel is stuck in a rut from a node perspective. Intel has so many issues with their own fabs and process nodes that they're actively looking to use TSMC and possibly even Samsung to build CPUs.
I have a Lunar Lake laptop (Yoga 2-in-1 14) and it's the most impressive Intel laptop I've seen. Good single core performance (better than M1), good multicore performance (a bit less than my M1 Pro), very good battery life. Minimum idle CPU package draw that I observed was 0.3 watts. This is not a gaming laptop, not a creatives laptop - just a nice laptop that's thin and light for students, office workers, and some other categories that require x86 Windows without super-strong multicore. If they can do that with TSM 3nm, what could they do with TSM 2 nm?
We may never find out as Intel supposedly was only going to do this with Lunar Lake. I've heard that they are using TSM for Nova Lake but I don't think that's an ULV project for them.
AMD has a nice Niche with Strix Point. Moderately better multicore than Lunar Lake, can be tied with discrete GPUs and it is put in thin and light laptops but I don't think that's a good idea. It's better in thicker laptops with good cooling. Good for gaming, creative work.
Neither of these compete against M4 in CPU performance, single or multicore but there are a lot of people out there that are fine with less performance than M4. And that's why Apple needs to compete in areas besides the CPU. And rumors are that they are doing exactly that. I don't know that I really care about whether I get M4, M5 or M6 from a CPU perspective. There are other things that I'd like in a laptop.
So what you're saying is Apple have an advantage because intel management are incompetent at both CPU design and fab execution right now?
Fab is one of intel's core business units and used to be a competitive advantage for decades that they had over AMD and everyone else. Since ~2012 they've mismanaged it into irrelevance.
Define "decent amount". 10%? 15%? 50%?
You're paying the engineers whether you're releasing chips or not.
My purchase history may be illustrative.
I bought a base M1 Max Studio in 2022. The performance of the base M1 chips would have been fine but I needed 32 GB of RAM and support for more than one external monitor. In late 2023, I found a base 2017 iMac Pro for sale for $800 and made an immediate offer when I saw it. I was lucky as he said that a lot of other people were interested in it.
Geekbench 5 on the Studio are 1,757 / 12,420 while the iMac Pro scores 1,073 / 8,057. The iMac Pro is my daily driver because of the display, speakers, microphones, 1 TB SSD and 32 GB of RAM standard. The loss of CPU performance is negligible for stuff I do on the desktop.
I bought an M1 Pro MacBook Pro in 2021. Geekbench 6 scores are 2,371 / 12,228. I bought a Lenovo Yoga 2-in-1 14 this past July. Geekbench 6 scores are 2,759 / 11,253 in performance mode but I mainly run it in efficiency mode where it scores 1,233 / 3,755. If I need more performance, I can just hit a button that cycles through three performance modes. Battery life in efficiency mode is about 14 hours for casual stuff. The battery performance of the Yoga is comparable or maybe a bit better than the M1 Pro in Low Power mode. But I have one program that cuts battery life in half on Apple Silicon but doesn't on Windows.
I'm on a trip with the Yoga and I love it as it is so much smaller and lighter. It has a 14 inch OLED 4k display, and I really appreciate the USB-A port for my Logitech dongle mouse. It runs as cool and quiet as my MacBook Pro. It can also be used in tent mode or as a tablet and has touch and pen support. There are rumors that Apple may put in a touchscreen for M6 MacBooks which could explain the larger UI elements in Tahoe. But the ability to take handwritten notes, draw and use touch is functionality that may appeal to a lot of people like creatives or students.
As you can see for me personally, the percentage doesn't matter because I'm happy with performance of modern chips or even a lot of older ones. The performance of this Lenovo Yoga is comparable to a 2015 MacBook Pro 16 without the heat and fan noise and with much better battery life. The vast majority of computer users do fairly light tasks which is why you still run into a lot of people using older equipment and getting annoyed that they have to spend money for something newer.
The question is does a 40% improvement in CPU performance matter to you getting more work done? Does 18 hours of battery life over 14 hours of battery life get you to upgrade? Or do factors like the screen type, speakers, upgradeability, price, flexibility and utility matter more? The upcoming MacBook Budget may wind up telling us that less performance is even more attractive than more performance when it comes with a lower price.
Intel and AMD make mobile chips with far more performance than Lunar Lake but the idea to make a chip that doesn't have the strongest CPU scores while providing x86 compatibility and long battery life is quite the surprise. What's more surprising is that they haven't really done a lot of marketing on this line of chips, likely because it's a one-off with TSMC.
Last edited:
10% of 100M. is 10M. 10% of 1B is 100M. The pure percentage isn't going to get to the real scope of the potential savings. Saving 1% on $1B is real money. ( 10M would cover a staff of engineers that have a real cost of $250,000 per person for a whole year. 80 engineer's for 6 more months. )
From initial prep to fully finished chip (of M-series class complexity) takes from 3-4 years to do. To get one being completed every 12 months takes pipeline development process. [ There was lots of dubious commentary when Apple bought Intel/Infineon modems and speculation that a 'restart from scratch' modem would show up in 1-2 years. ]
If three years. [ somewhat an over simplification for illustration: D -- base design and new approach research B -- build concrete components and 'soft' debug (simulate, fix cycle ) V-- verification real hardware really works. ]
Code:
design n D B V
design n+1 D B V
design n+2 D B V.four years than then yet another "team" operating concurrently. Can try to do position the stages as 100% siloed groups of engineer. [ so once at year 3 always have a vertical slice that covers all the stages so all 'busy'. ]
The disconnect there is that each stage is 100% independent of the others. If 'build" is broke in some way then may have to fix that ( with more build/simulate/test before can complete. ] Similar design disconnects from the latter two stages need to bubble back to that previous stage.
Furthermore, the scope ( depth and breath ) of the workload similations are very likely no human resource constrained. What software is optimized for is limited to that can get the hardware verification on time to hit whatever deadline there is. The design team could be throw at broad pathfinding development tasks. The verification could be thrown more security/robustness tasks to do. Every single stage there trims off possible optimizations to hit the deadline. (nevermind skill development could send engineer staff through if not always in crunch/panic ship mode . )
The other issue is that Apple's scope of dies needed to be created is getting larger. As Apple shifted into M-series the Watch updates got slower. In part, because the watch cores were 'fast enough'. Also likely in part because the human design/build resources are limited and the watch fell into 'copious spare time' zone. [ i.e., the watch update will fall into a A/M E core cycle which where E cores were not a focus and can timeslice in the watch cores. ] ( Apple collects several generation of E core improvements before doing anything with a new chip. )
The 'idle' time costs of the engineers is not likely to be the largest cost/overhead choke point.
Last edited:
Why not improve chip performance AND screen, speakers, etc?
Believe it or not, Apple is competing against many players including Intel, Qualcomm, Arm, Mediatek, AMD, Nvidia, Broadcom. Market forces tells them they should make improvements to chips every year.
I think it's better to think in terms of market forces instead of your personal experience.
I want an iPhone Mini Pro. Apple is not making it. Probably never making a small phone ever again. Market forces have spoken. My own personal experience is not what the market dictates.
Yes, let's look at market forces. Right now Apple has to discount Macs in third-party channels to get them to sell. They have the best CPU, so, if that were the most important market metric, they should have much higher marketshare.
Why haven't they been doing the other stuff? Maybe it's expense, logistics, supply chain. But they're rumored to be doing a lot of that for M6. I will say it again in that M4, M5, M6 does not matter for the vast majority of users.
I think that’s unfair to the people who work brutally hard to deliver the products we love at the blistering pace they do.
There are just so many moving pieces in development, manufacturing and supply chain that even Apple is unable to update all computers to new chips all at once.
The MacPro didn't even make the enumerated list above. every 4-6 years is a blistering pace? Not!
The pacing here is in part set on cost recovery of the M-series chips being utilized. To 12 month 'churn' the MBP 14/16" models on Max's means the Studio with Max's goes slower because there is no "hand me down" product for the SoC package ( which pragmatically needs to be sold for more than just 12 months. Intel , AMD , Nvidia ... general chip makers do NOT make chips for only 12 months. The hype train may more onto the next cycle in a year, but the previous processor products are still made and sold.).
The iMac is getting similar 'treatment' the Mini got in last 8-9 years of Intel era.
The Mac Pro not being upgraded to a M3 Ultra is blocked with what high hurdle from the list above. The SoC itself is in the Studio; so it exists. Development. The ports on the MacPro are different how on a Studio ( other than more of them)? Huge development task? Nope. Supply Chain... massive numbers of Mac Pro's being sold now (or with a M3 Ultra )? Nope. Same case, Same power supply, same fans, same 8-pin aux power connector, etc. etc. Huge logistical hurdles? Nope.
The Mac Pro is a 'copious spare time product'. Was back latter half of Intel era and still there now.
The churn on the lower part of the Mac product line up is supported by they have other places to stuff plain M-series to. iPad Pro and then "hand them down" to iPad Air as 'new' . Throwing a chip the size and development complexity of the Max into the trash can every 12 months extremely likely doesn't make any economic sense at all. So it doesn't happen.
[ cost recovery for development is not 'development' itself. Making profits is a factor here at least as high as those other 3 mentioned. ]
In PC space they are getting a bit constipated with multiple generations with not enough places to go.
"... Chen said that multiple generations of x86 chips from both Intel and AMD are already proving to be complicated— the entry of Nvidia as a third potential supplier will likely make procurement and inventory management much more challenging. ...
Nvidia’s entry into the CPU-making business through Intel will definitely give tech enthusiasts more options — but it will also make it harder for manufacturers to find the right pairing of CPU, GPU, RAM, PSU, etc., that buyers will appreciate. And for the common person, this could just lead to decision fatigue, where they will just settle for the brand they’re familiar with. ..."
Acer chief says Nvidia’s Intel investment will complicate PC supply chains — says Lutnick's 50/50 proposal could take 50 years to realize
Acer leadership has some thoughts on future chip supply chain developments.
There is a substantial among of dogma that got built up around the general PC industry when it was in serious double digit , year-over-year growth phase. Part of the that is this 12 month cycle thing.
Last edited:
"... Chen said that multiple generations of x86 chips from both Intel and AMD are already proving to be complicated— the entry of Nvidia as a third potential supplier will likely make procurement and inventory management much more challenging. ...
Nvidia’s entry into the CPU-making business through Intel will definitely give tech enthusiasts more options — but it will also make it harder for manufacturers to find the right pairing of CPU, GPU, RAM, PSU, etc., that buyers will appreciate. And for the common person, this could just lead to decision fatigue, where they will just settle for the brand they’re familiar with. ..."
Acer chief says Nvidia’s Intel investment will complicate PC supply chains — says Lutnick's 50/50 proposal could take 50 years to realize
Acer leadership has some thoughts on future chip supply chain developments.www.tomshardware.com
On the Intel side, you have worries that 14th and later generations might have instability problems. So you make the 12th gen available for those that are paranoid about it. Prices aren't even all that attractive on 10th, 11th, 12th generations - don't really know why.
Last I checked 40xx nVidia GPUs cost more than their 50xx equivalents - again, no idea why.
AMD is still selling CPUs for old sockets so that those that want a cheap upgrade can get one. I kind of wish Intel would do that but it must make an absolute mess of SKUs.
Yes, consistency is certainly an issue.
I also feel a sort of laziness that creeps in with yearly updates. If you look at the MacBook Air M2, M3 and M4 … they’re basically the same machines. They feel like “lazy” updates to me, plug a new processor in and be done with it.
Next year’s M5 Air will probably be the same again.
Apple has turned into a cash machine and somewhere along the way some of its magic has gone missing.
I see we’re at the ‘spec updates never happened with old Apple’ part of the thread.
Lol indeed. We aren’t gonna get entirely new designs every generation and tbh why would we want that? What matters is how good the machine is, not how new it is.
There used to be a time when an Apple keynote was actually exciting because we didn’t know what to expect. Today’s keynotes are a lot more predictable.
Not a complaint, just an observation.
Makes me wonder if we need a keynote when the only change is a processor upgrade.
Yeah we also have an Apple rumour industry that spoils everything if you haven’t noticed. Pretty difficult to be
excited given that. Who said there’s a keynote? We didn’t have one last year for M4 ¯\_(ツ)_/¯
Rumours that are often so right that you wonder if they’re just carefully controlled leaks to keep people talking.Yeah we also have an Apple rumour industry that spoils everything if you haven’t noticed. Pretty difficult to beexcited
Who said there’s a keynote? We didn’t have one last year for M4 ¯\_(ツ)_/¯
The closer leaks occur to release it is more likely a problem of "too many" people have to know. If 25 iPhone case makers need to have precision dimensions so they can make cases ,then it is far more likely to leak from 25 (or more) vendors than it would leak from 1-2 vendors ). More folks involved the more likely info will pop out.
Inside of Apple also. Once specs for device has to be translated into 40 languages , then more likely that more folks are 'read into' the information. Same issue when 'reading in' major retailers of the products. Even bigger problem when Apple pushes manufacturing into "as cheap labor as possible" contexts where a USA $25-50 bribe is 'megabucks' and folks will talk.
On very few occasions , the outside talk spins so vastly off base that Apple will leak something to pop that 'bubble' before it gets worryingly large. ( 'Decent chance we are not going to do a M4 Ultra ; not doing Ultras every iteration. ' = Sometimes that is done more indirectly than that one was. )
If you think Apple keynotes product launches are boring. Google Pixel stuff leaks like a huge, pourous sieve. There is almost nothing left by the time they do the talk.
Apple has so many successful devices that they have to keep going and iterate on that the magic has kind of gone missing. For me, the Vision Pro was a misstep that would never have happened while Jobs was around. It’s heavy and has a ton of compromises, it’s just not a mass consumer product.
I think a core concept people are missing here is that CPU/GPU designs take years.
You can't just decide "oh this is fine guys, take some time off" or "work on something more impressive than 12 months" because these decisions are made years in advance based on projections.
If you find that a competitor releases something that blows your year N device out of the water, you can't "react" because you'll be 2-3 years down the track before you see results. And there's a lot more risk riding on each individual design.
No. You need to put the best you can out as soon as you can.
Or die.
You can't just decide "oh this is fine guys, take some time off" or "work on something more impressive than 12 months" because these decisions are made years in advance based on projections.
If you find that a competitor releases something that blows your year N device out of the water, you can't "react" because you'll be 2-3 years down the track before you see results. And there's a lot more risk riding on each individual design.
No. You need to put the best you can out as soon as you can.
Or die.
Let’s be honest here, the only thing that is going to “die” is a soulless, profit-obsessed corporate entity. The people will just walk into another job.
Nvidia stopped production of the 40xx series in advance of the 50-series announcement. Many people have speculated that Nvidia did that for two reasons:
a) avoid having to discount 40xx GPUs like they had to do with 20xx and 30xx when the next generation was released
b) artificially increasing demand of both 40xx and 50xx GPUs to drive prices upwards.