Ok, if that gets sooo much attention why not start already with it?
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.
M4+ Chip Generation - Speculation Megathread [MERGED]
- Thread starter GoetzPhil
- Start date
- Sort by reaction score
Base stats:
- 40 high-performance cores
- 20 high-efficiency cores
- 100-core GPU
- 80-core Neural Engine
- 8GB Unified memory
Assuming a 19.5 month refresh cycle
- M1: Q4 2020 5nm
- M2: Q3 2022 5nm
- M3: Q1 2024 3nm (N3)
- M4: Q4 2025 2nm (N2)
- M5: Q2 2027 1.4nm (A14)
- M6: Q4 2028 1.4nm (A14)
- M7: Q3 2030 1nm (A10)
- M8: Q2 2032 0.7nm (A7)
- M9: Q4 2033 0.5nm (A5)
- M10: Q3 2035 0.3nm (A3)
There was rumors of Apple wanting to align Mac chip's 19.5 month refresh to iPhone chip's 12 month refresh cycle.
This would help simplify and align chip generations.
Last edited:
The M4 will make all your wildest dreams come true!
The M4 will add a bunch of Vision Pro R1 tech so you can control your Mac cursor with your eyes and hand gestures.
Finally you'll be able to have zoom meetings on your Mac with a virtual 3D avatar, so instead of just not needing pants, now you can be completely naked while presenting the financial numbers for Q4.
Finally you'll be able to have zoom meetings on your Mac with a virtual 3D avatar, so instead of just not needing pants, now you can be completely naked while presenting the financial numbers for Q4.
I cannot imagine the R1 cores being put into the M4 unless other Macs will receive Vision Pro parts. In the same way iPhones aren't using M1 chip as a lot of its microarchitecture isn't useful to it.
M6 will have wireless brain-chip interface, eliminating keyboard and mice once and for all. You will be one with your computer. You will batch outsource your thoughts for Mac Mini to process for you, at the rate of 20 thoughts per minute (new benchmarks will need to be created)
M10 will eliminate monitor and project signal straight to your retina via upgraded brain chip with the speed of 120 human "looks" per second in full 8k 360 view.
M12 will add all other sensory connections: smell, touch, taste, vestibular and proprioception.
Watching pHub will be next level experience, not to mention playing fortnite might have real life consequences in "hardcore" mode, having one life only one will have to be carefull.
M10 will eliminate monitor and project signal straight to your retina via upgraded brain chip with the speed of 120 human "looks" per second in full 8k 360 view.
M12 will add all other sensory connections: smell, touch, taste, vestibular and proprioception.
Watching pHub will be next level experience, not to mention playing fortnite might have real life consequences in "hardcore" mode, having one life only one will have to be carefull.
Reduction in feature size will not happen. Already "3 nm" is a process name and not a description of the size of the transistors.
All technology has a curve that is non-linear. Change is fast when a new technology is introduced and then over time the rate of change slows. We see this in airplanes, cars, electric toasters, and bicycle tires. Eventually, the technology becomes more stable with only slow incremental changes.
What I hope will change is the kind of software we want to run on our Macs and then Apple will evolve the design of the Mac to better run this software. There was a software revolution 30 or so years ago when mouse-driven point-and-click user interfaces became popular and then small computers no longer spent the majority of their "compute-power" actually computing, most of the power was used to run the graphical interface and computation because a secondary function. This made computer hardware to become VERY different and the GPU was invented. Today most of the raw computational power is in the GPU.
What if the user interface were driven by a large AI model? What if cameras looked at your face and hands and listened to your voice? If this kind of software is what runs on computers in 10 years then Apple will need to build systems with thousands of neural network cores. These cores are simpler than GPU cores and there really could be thousands of them in a Mac.
Simultaneously disappointing everyone on Macrumors, with endless discussion threads complaining that it didn’t cure the users gonnorhea, or get the cute girl in the office to notice them.
There is no 2nm fab ready for mass production in 2025, it will be 2026 at the earliest if there is no unexpected delay (and I think there will be, because of China's ban on gallium and germanium).
M3 Late 2023/Early 2024 on N3 (3nm)
M4 Mid/Late 2025 on N3P (3nm+)
M5 Late 2026/Early 2027 on N2 (2nm) or N3X (3nm++) if there is production delay
Attachments
Last edited:
Bi-directional? My Mac will slow down because I need a shower? Or I put my foot through the glass top of my coffee table because the Mac told me it was above the ottoman?
And M20 will out our asses as we will become chip connected zombies already 🤣
Interface or otherwise, AI is only going to become more important in the future. So that’s probably the next frontier in processor development. More powerful and numerous neural cores will probably be the next measurement of computing power.
The M2 Ultra is honestly embarrassing for a $4K desktop computer. You can get vastly more performance in a PC for that price. Apple Silicon was made for laptops and phones. There’d be no shame for Apple if they threw in the white flag and admitted that on desktop, where performance per watt is irrelevant, Apple silicon can‘t compete.
Except that performance per watt is not irrelevant, you might have something there.
Apple isn't likely going to touch N3X with a 50ft pole.
"... TSMC claims that N3X will support voltages of (at least) 1.2v, which is a fairly extreme voltage for a 3nm-class fabrication process. The leakage cost, in turn, is significant, with TSMC projecting a whopping 250% increase in power leakage over the more balanced N3P node. ..."
For a company that preaches regular sermons on Perf/Watt as being their primary metric... N3X is the exact opposite of that. Honestly, I don't see any computationally focused cores using that ( not just Apple , but Nvidia, AMD , Intel , etc.). There are some 'priciness and the pea' latency problems in networks and a few other corner cases that might want to use that , but general computation ... nah.
There is a small group of folks eager for Apple to take the 'top fuel dragster' single threaded drag racing crown away from Intel at any costs. However, highly doubtful any of them are calling the shots in the M-series design meetings. 'N3X' is basically useless for Apple. They aren't in the ST drag racing business.
Far, far , far better thing to do would be to spend more money updating architecture on N3P and using more die area space and just waiting for the kinks to get worked out of N2 a year or so later.
TSMC is starting to talk about a N3S (which solves some density limitations). Whether that is a follow on to N3P ( as denser N3P ... getting back to what N3B hoped to be) or trades density ( more so a smaller N3E. ) is a bit unclear . But something other than 'X' is far more likely candidate if N2 proves problematical. [ Slimmer chance a N3S is a better N3B with better fab equipment to cut down the multipatterning (and/or backside power ). ]
An efficient SoC provides what some people who use desktops value: a near-silent computer.
You forgot SSD's that are slowed down to 750 MB/s because SSD speeds don't matter.
M4 might be on 2nm, but here's some 3nm ASi possibilities...?
- A17 - N3B
- M3 - N3E
- M3 Pro/Max/Ultra - N3P/S
By M4, we might see high-density STTRAM modules that would obviate the SSD disadvantage entirely and change "off" into the same thing as "sleep". STTRAM is currently at around DDR3, but improved density could maybe also improve speed.