The machine you need now is always more valuable then the machine you want later.
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.
Looks like we may see an M3 pro/max powered MBP this year.
- Thread starter A1423
- Start date
- Sort by reaction score
I don't think we'll see M3 until next year, probably some time after they release the Vision Pro.
Last edited:
If Apple is releasing new models as soon as it is able, then Apple is doing its job properly.
Effectively you're saying that, even if Apple were able to release an M3 MBP 7-10 mos after the M2 MBP, it should instead delay that release just so you don't feel emotional regret for your M2 purchase.
That seems like a selfish position to take, because while you aren't hurt by Apple releasing a new version after 7-10 months, customers waiting for the M3 would be hurt by subjecting them to an unnecessary delay.
Furthermore, 7-10 months is a reasonably long time in the processor update cycle. There were a few years that Apple's update cycle for the MBP was only 6 months. That provided a great benefit to customers since, regardless of when you needed to purchase, you were getting a very recent model (never more than 6 months old).
Last edited:
Since when does Apple release things as soon as they are able? They plan on releases around a certain time that they more often than not abide by.
What I am saying is, if they are shifting the schedule to earlier than what it is now it’s going to piss off/annoy people who rely on the pattern already practiced and planned their purchase accordingly. If they pick 12 months like the iPhone than they pick 12 months. If it’s 18 months than 18 months. If it’s 6 months it’s 6 months.
That's exactly what they've done with Apple Silicon Macs (releasing products as they become available), as evidenced by the fact that their release dates have been all over the map; it's not followed much of a pattern. You can't annoy people by deviating from a regular schedule, since there has been no regular schedule. It most decidely has not been like the regular release schedule of the iPhone.
Moreover, Apple Silicon has had only two generations. Even if what they've done thus far did have the appearance of consistency, that's too short to establish a pattern. Again, it's not like the iPhone. For each processor, we only have two data points! How do you get a pattern out of that?
If anything, customer expectation should be that what Apple has done thus far is not predictive of future release cycles, both because these are just the first two generations, and because we had COVID interfering. I recall disussion that Apple hoped to get onto a yearly update cycle with the Mac, to synchronize with the iPhone's (something they've clearly not done yet).
Last edited:
I think most people are overestimating the performance jump from M2 to M3 for the reason not to make people with M2 angry. I think we see am smaller SoC chip area, some efficiencies and a "modest" 10-20% increase in performance. Maybe more increases on the GPU side especially on the Max chip. The only rumour is the implementation of hardware RT which will be a real generational shift in capabilities. If you don't game or do 3D rendering, HW RT will have minute effect on the productivity.
I like symmetry so I would rather see a yearly "Mac" event where all Mac changes the SoC generation at the same time. iPhones are so wonderfully predicable.
I like symmetry so I would rather see a yearly "Mac" event where all Mac changes the SoC generation at the same time. iPhones are so wonderfully predicable.
The main reason I expect a large performance uptick is simply because M2 is still essentially A14 tech, with fairly minor evolutionary tweaks. At some point we will see a new u-arch, which will set the platform for the next two to four years. Usually Apple introduces a new u-arch every two years, and there is good reason to believe it was planned for 3nm (but got delayed because of production issues). If their next-gen CPU microarchitecture does not come with a substantial performance improvements, Apple might be in trouble. They have a chance now to take the initiative in high-performance personal computing and beat Intel/AMD/Nvidia at their own game.
Amen. Thanks for bringing some logical analysis into the usual madness
Is 20% much? I think so. The CPU and more to the point the GPU microarchitecture might be much better but who says Apple will use its whole potential the first year? I think they spread the potential out over the entire architecture cycle. A steady increase between M generation would be preferable.The main reason I expect a large performance uptick is simply because M2 is still essentially A14 tech, with fairly minor evolutionary tweaks. At some point we will see a new u-arch, which will set the platform for the next two to four years. Usually Apple introduces a new u-arch every two years, and there is good reason to believe it was planned for 3nm (but got delayed because of production issues). If their next-gen CPU microarchitecture does not come with a substantial performance improvements, Apple might be in trouble. They have a chance now to take the initiative in high-performance personal computing and beat Intel/AMD/Nvidia at their own game.
Amen. Thanks for bringing some logical analysis into the usual madness
Yeah, extrapolating from two data points (M1 and M2) is meaningless. I guess it has to do with chip industry node progress as well.
I'm sure there are internal discussions at Apple about spreading performance increases over a few years. It's natural human instinct to "relax" when you're on top.
However, they just had huge declines in Mac sales. One way to bring Mac sales back to growth is to blow it out with an M3 SoC that makes M1 owners want to upgrade and have the undisputed best performance (not performance/watt) for any laptop.
So you'd prefer it if tech companies delayed their latest products for the sole reason that it annoys you? It doesn't change anything about your already-purchased device and the faster the tech advances in the meantime, the more modern tech you'll get for the least amount of money the next time you upgrade. Would it be less inconvenient for you if they held back M3 until you're ready to replace your Mac a couple years down the road? If I can't have the latest tech, nobody else can!
And where will the transistors come from required for such large performance uptick?The main reason I expect a large performance uptick is simply because M2 is still essentially A14 tech, with fairly minor evolutionary tweaks. At some point we will see a new u-arch, which will set the platform for the next two to four years. Usually Apple introduces a new u-arch every two years, and there is good reason to believe it was planned for 3nm (but got delayed because of production issues). If their next-gen CPU microarchitecture does not come with a substantial performance improvements, Apple might be in trouble. They have a chance now to take the initiative in high-performance personal computing and beat Intel/AMD/Nvidia at their own game.
Amen. Thanks for bringing some logical analysis into the usual madness
N3 process brings 40% logic density increase, and SRAM, and alongside it: memory controllers density increases by 5%.
M2 on N5 node has 20 bln transistors with 150 mm2 die size.
M3 in 150 mm2 die size will get, at best 25-26 bln transistors, because of the characteristics of the process.
Thats 25% more transistors. If Apple wants to increase die size - then go ahead, they can, but it will be ridiculously expensive for consumers, since each 3 nm wafer costs around 20000 USD with a yield that is around 60% and increases by 5% each quarter.
Because of the yield Apple HAD to release M2 Max and Ultra Mac Studio and Mac Pro, because we should not expect next gen M3 Max and Ultra for at least 18 upcoming months.
Its beneficial for Apple, and for consumers if we will see M3 series next year, because they will have stacked enough chips to fulfill demand, and the yield is reasonable for mass production, with low costs.
And secondly: don't expect meanigful changes to the architectures performance.
I would not expect higher core clocks because it will cost transistors, Apple may want to push for higher bandwidth for those cores, with larger and faster caches, while also cutting down the memory controllers because they are the largest consumers of precious die space on smaller, denser nodes, which is why I expect that M3 Pro will come with 192 bit bus, and not with 384, like Previously I thought.
IMO, M3 will have 128 bit bus, 12 GPU cores, 4P/4E CPU cores, both will have new architectures, that will deliver 25% performance increase over M2 series, and thats it.
I changed my mind, about M3 series AFTER I learned about how woeful N3 process is. Its by far the smallest density and performance uplift with node shrink in history.
Remember that just makes the SoC cost $64 instead of $40 (price example). Not really that big of a deal when the best selling models start at $1099.
New 3D packaging with a 3nm logic die and a 5nm SRAM/memory controller die.
Or you might be right and see a monolithic die with only modest improvements. Who knows.
A few things:
- Let's assume M3 is N3E.
- There's no SRAM improvement.
- However, power is a whopping -34% compared to N5 or speed increases by +18%.
- A16 is 11% faster in CPU over A15 (M2). If Apple got lazy and used A16 ported to N3 for M3, then M3 would already be 1.3x (1.1 * 1.18) faster than M2 by default.
- Let's assume that Apple uses A17 instead. Let's assume that A17 is 11% faster than A16 on the same node. Then M3 would be 1.42x (1.1*1.1*1.18) faster than M2.
- 1.42x improvement in one generation is not unheard of for Apple - especially if M3 is 2 generations of A series SoC improvements and 1 major node ahead of M2.
- Let's ignore the logic and math and be very conservative and say M3 is 20% faster than M2. That's still an incredible leap for 1-1.5 year.
Last edited:
That patent mentions copper pads. Does that mean copper to copper bonding? I think that's part of TSMCs 3D strategy. Are Apple's low power designs ideal candidates?
Mark Gurman mentioned awhile back that Apple’s internal goal is to get the Mac SoCs on an annual update cycle like the iPhone. When Apple released the M2 Pro/Max MacBook Pro in January, instead of last fall as was originally planned, that told me Apple’s perhaps setting that cadence for the future.
January 2024: M3 MacBooks (M3/M3 Pro/M3 Max)
March 2024: M3 iMac, M3/M3 Pro Mac mini/iPad Pro M3
May 2024: M3 Max/Ultra Mac Studio/Mac Pro
June 2024: WWDC
September: iPhone/Apple Watch/low-end iPad update
January 2024: M3 MacBooks (M3/M3 Pro/M3 Max)
March 2024: M3 iMac, M3/M3 Pro Mac mini/iPad Pro M3
May 2024: M3 Max/Ultra Mac Studio/Mac Pro
June 2024: WWDC
September: iPhone/Apple Watch/low-end iPad update
You're right.
I had forgotten about this. https://forums.macrumors.com/threads/apple-silicon-timeline-gurman.2330395/
It doesn't stop a lot of people who insist that Apple will not have an annual release cadence despite an Apple exec saying it, Marc Gurman reporting it, and logical speculation suggesting that it's Apple's true intention.
Assuming they don't have problems with manufacturing process from TSMC.
Last edited:
I suppose Apple wants an annual cadence but they will delay/fastrack certain generations based on TSMC's nodes.
It seems like TSMC is settling into a 3-year new node cadence.
No use getting angry. This has been the same situation since computers were invented. There’s always a new model around the corner.
There was a video floating around of Steve Jobs at a whiteboard when he was at NeXT talking about his goals, one of the goals being a 9 month upgrade cycle. When he returned to Apple, models were regularly updated every 9 or 10 months, sometimes sooner. And to think this was before the hyper connected Internet we have now!
When Intel started missing its targets, or the targets didn't meet Apple's internal plans, Apple either skipped the update or waited until Intel's next at bat. This is when Macs started to whither on the vine and Apple starting making plans to move off of Intel. So here we are.
Apple could be planning 18-month product cycles like it's done with the iPad Pros. Who knows?! The bottom line for me is that Apple seems like it's lining things up for an annual Mac upgrade cadence, starting in mid-January, conveniently after the holiday return period is over. 😆
When Intel started missing its targets, or the targets didn't meet Apple's internal plans, Apple either skipped the update or waited until Intel's next at bat. This is when Macs started to whither on the vine and Apple starting making plans to move off of Intel. So here we are.
Apple could be planning 18-month product cycles like it's done with the iPad Pros. Who knows?! The bottom line for me is that Apple seems like it's lining things up for an annual Mac upgrade cadence, starting in mid-January, conveniently after the holiday return period is over. 😆
CPU/GPU/SoC designs are done many generations into the future, it is mostly manufacturing that is going to be the hold up. I would assume Apple is targeting ~20% improvement per generation on the CPU side and maybe more aggressively 30% increase on the GPU side. What this means in practical terms is to expect CPU performance to double every 4 generations and GPU performance to double every 3 generations. Some of this is built with the assumption of process improvements on the process node.
Apple can tweak this by give us more cores at the cost of more die area but I am sure they are always trying to balance that with production cost and battery life.
And Apple can for sure pull this off given how many time they have done this with the iPhones which has higher manufacturing complexity and volume that your typical Mac.
Apple can tweak this by give us more cores at the cost of more die area but I am sure they are always trying to balance that with production cost and battery life.
And Apple can for sure pull this off given how many time they have done this with the iPhones which has higher manufacturing complexity and volume that your typical Mac.
Higher volume than the Mac, yes. However, M series SoCs are more complex than their A-series counterparts used in the iPhone. Here's a good comparison of the M1 and A14 SoCs...
Define “more complex”.
Why would you link known clickbait website? CPU monkey steals benchmark results and speculations from other sources and publishes them unvalidated. It’s the worse source you can take.