Time to issue a stop sale order on the M1, M2, and M3 to avoid unsuspecting customers getting severaly outdated hardware.
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.
Apple Reportedly 'Just Started Formal Development' of M4 MacBook Pro
- Thread starter MacRumors
- Start date
- Sort by reaction score
I don't think there's much to develop, we know the design will not change until at least the M5 MBP. It's just an evolution as they have been so far.
This makes sense, the M4 will finally up the ante with substantial neural engine improvements, and I suspect may be the first chip that has variations there between the base / pro models which are desperately needed. It's stupid that I can buy an M3 MBP for $5,000 that has the same amount of Neural Engine cores the Macbook Air has, they need to segment the lineup better.
I also expect M4 will share some of the new hardware from the A18 which is the reason for the "just started development" part of Gurman's report, this time we will finally have parity between Phone and Mac CPUs as far as Neural engine enhancements – not core counts obviously but the underlying tech.
It's going to be a tough decision to get an M3 Ultra Studio or M4 Macbook Pro... I'd probably lean toward the laptop given the local AI stuff Apple has cooking but I don't want to wait another year+ either and need a new DAW. Might go for a mid range M3 Studio to hold me over and then sell it to upgrade but I hate switching main computers and setting up development environments etc. all over again.
I expect for the Pro Laptops the M4 or M5 generation will either introduce faceID or shrink the notch. Probably the M5 generation, if the M4 is so Neural Engine focused that will be enough to sell once MacOS gets some features that really leverage it which are definitely coming in the next 18 months.
I also expect M4 will share some of the new hardware from the A18 which is the reason for the "just started development" part of Gurman's report, this time we will finally have parity between Phone and Mac CPUs as far as Neural engine enhancements – not core counts obviously but the underlying tech.
It's going to be a tough decision to get an M3 Ultra Studio or M4 Macbook Pro... I'd probably lean toward the laptop given the local AI stuff Apple has cooking but I don't want to wait another year+ either and need a new DAW. Might go for a mid range M3 Studio to hold me over and then sell it to upgrade but I hate switching main computers and setting up development environments etc. all over again.
I expect for the Pro Laptops the M4 or M5 generation will either introduce faceID or shrink the notch. Probably the M5 generation, if the M4 is so Neural Engine focused that will be enough to sell once MacOS gets some features that really leverage it which are definitely coming in the next 18 months.
Last edited:
Sounds like trouble. I still remember Get MX Out of Europe protests from the 1980’s
Last edited:
These MBPs are a pretty new design so I would not expect any visible design changes for another few years. There is no pressing need for it. The previous design debuted in 2012 and wasn’t replaced until 2021 with the only visual change being the addition of the touchbar.
Last edited:
Apple: since it is too high-end, we finally decided to name it "Mac Ultra" 🤣What if Mac Pro IS the car? Just add those Apple wheels, carefully balance (Cirque Du Apple) and drive... well hang on and hope Siri can get you somewhere near where you want to go.
Somehow I did not get Aoplecare on my m1 🍸🙀,and they were reluctant to add me after a couple years🍸😿, so, easy solution. Buy a new computer. M4 in 2024? I’m in!🍸🐈
It looks like you're the smartest here. A wise old man once said (my father, mainframe computer engineer): You can only tell a difference if the new computer is 2x as fast. At 4x you can upgrade. So, let's keep looking at the benchmarks. I, still happy with a 2019 I9 MBP 16.
I think they’re going to stick with the paradigm they’ve created in that the lineups are segmented by the ports/storage configurations/RAM. Mainly because by keeping everything pretty much the same, they can use and reuse the same chip architecture across the entire like, from their high volume Airs, to their extremely low volume Studios and Pros (which they don’t sell enough of to make it worth a unique effort).
I just upgraded from that same i9 to an M3 Max (I like to buy a top-end machine and keep it a while), and I'll be darned if it isn't 4x as fast. I watched a photo export that had been just over 2 minutes per image go down to 30 seconds. I can manipulate a 1.5 GB panorama in real time - which is a stunning amount of processing power.
The other thing that stands out about the M3 Max is that it is absurdly efficient. Not only does it have the CPU power of a biggish desktop workstation and the GPU power of a mobile workstation (although not quite the GPU power of the biggest mobile gaming rigs), it has the power consumption of an iPad when I'm just writing on it. As I'm writing this post, I'm watching iStat's system power consumption meter. While I'm actively typing, it is in the mid 4-watt range. When I take a break for a few seconds, it drops in to the low threes. That's insane even for an Ultrabook, let alone a top-end mobile workstation. Most PC-based mobile workstations can't idle below 10 watts, and 15 is common (that's why they get 6 hour battery lives in absolute beet cases ).
I also haven't managed to get it over 50 watts for any sustained period, other than with Cinebench. DxO Photo Lab's famously piggish DeepPrime XD noise reduction/debayering? 30 watts or so, with a 2-3 second boost to 80 watts at the end of the 30 second procedure. Any laptop PC workstation would take longer, and it would peg both CPU and GPU to the wall at well over 100 watts, possibly 200. The draw would be so high that it would refuse to run on battery. It would run at half speed, and it would have an hour or so of battery life. This thing will give me close to all-day battery life while EDITING 100 MP PHOTOS. That is, of course, assuming that I'm actually editing (it would probably only run about 3 hours in a single batch export on battery).
The M4 Max is almost certainly going to be a mild upgrade to this (like M2 Max over M1 Max), not what this is over M1/M2 Max, nor what M1 Max was over Intel. They're not going to manage to squeeze in 50% extra P-cores, PLUS get a 15% improvement in per-core performance, PLUS add efficiency, all in the same generation again for a while. If we're lucky, M4 will be a particularly good upgrade for the base and Pro chips - the spectacular one in the M3 range is the Max (and presumably the Ultra we've yet to see). Maybe it'll just be 10-15% across the board like M2.
At least for photographic use, it's beginning to beg the question of what more we need. I'm manipulating large numbers of 100 MP medium format image files, and it's well past the line of fast enough. Every operation that should be real-time is, even on a stitched panorama. Export times are fast enough not to worry about. 64 GB of RAM sees no pressure, even on that big pano stitch (I was watching that especially closely, because I didn't get 128 GB, I was still inside the return window when I stitched that pano, and I wanted to make sure I wasn't wrong). Sensors aren't increasing in resolution especially fast, and I'm extremely confident this thing will handle the next generation (~160 MP), and probably the one after that (~220 MP), even going by high-res medium format sensors (yes, I know about Phase One's 150 MP chip that uses a larger version of medium format, but that's around $50,000, so it's not relevant to most of us).
The other thing that stands out about the M3 Max is that it is absurdly efficient. Not only does it have the CPU power of a biggish desktop workstation and the GPU power of a mobile workstation (although not quite the GPU power of the biggest mobile gaming rigs), it has the power consumption of an iPad when I'm just writing on it. As I'm writing this post, I'm watching iStat's system power consumption meter. While I'm actively typing, it is in the mid 4-watt range. When I take a break for a few seconds, it drops in to the low threes. That's insane even for an Ultrabook, let alone a top-end mobile workstation. Most PC-based mobile workstations can't idle below 10 watts, and 15 is common (that's why they get 6 hour battery lives in absolute beet cases ).
I also haven't managed to get it over 50 watts for any sustained period, other than with Cinebench. DxO Photo Lab's famously piggish DeepPrime XD noise reduction/debayering? 30 watts or so, with a 2-3 second boost to 80 watts at the end of the 30 second procedure. Any laptop PC workstation would take longer, and it would peg both CPU and GPU to the wall at well over 100 watts, possibly 200. The draw would be so high that it would refuse to run on battery. It would run at half speed, and it would have an hour or so of battery life. This thing will give me close to all-day battery life while EDITING 100 MP PHOTOS. That is, of course, assuming that I'm actually editing (it would probably only run about 3 hours in a single batch export on battery).
The M4 Max is almost certainly going to be a mild upgrade to this (like M2 Max over M1 Max), not what this is over M1/M2 Max, nor what M1 Max was over Intel. They're not going to manage to squeeze in 50% extra P-cores, PLUS get a 15% improvement in per-core performance, PLUS add efficiency, all in the same generation again for a while. If we're lucky, M4 will be a particularly good upgrade for the base and Pro chips - the spectacular one in the M3 range is the Max (and presumably the Ultra we've yet to see). Maybe it'll just be 10-15% across the board like M2.
At least for photographic use, it's beginning to beg the question of what more we need. I'm manipulating large numbers of 100 MP medium format image files, and it's well past the line of fast enough. Every operation that should be real-time is, even on a stitched panorama. Export times are fast enough not to worry about. 64 GB of RAM sees no pressure, even on that big pano stitch (I was watching that especially closely, because I didn't get 128 GB, I was still inside the return window when I stitched that pano, and I wanted to make sure I wasn't wrong). Sensors aren't increasing in resolution especially fast, and I'm extremely confident this thing will handle the next generation (~160 MP), and probably the one after that (~220 MP), even going by high-res medium format sensors (yes, I know about Phase One's 150 MP chip that uses a larger version of medium format, but that's around $50,000, so it's not relevant to most of us).
Somewhere in that range (maybe below 1 nm, but not FAR below), there is a brick wall known as the Standard Model of physics. Since measuring chip features is more than a bit arcane, I don't know how big "3nm" features really are. The atomic radius of silicon is around 0.15 nm (again, depending on exactly what you measure). We're looking at VERY small numbers of atoms, and that means that you start worrying about things like Heisenberg's Uncertainty Principle. Quantum physics is weird, but it is also, as far as we know, correct.
I also have the MBP 2012 and am in the same situation as you are. I wanted to get a new MBP for the past 3 years but always felt like I could hold off and wait for the newly released chipsets to get worked out when the first generation came out. Now I know that I need to replace my laptop and it looks like the M4 MBP is going to be the one.
I'd agree with you if Apple didn't already start down this path with the M3 Pro being very differently laid out vs the M1 and M2. I think we're going to see more divergence exactly like they did with GPU cores, especially as, at least for the next couple of years, "AI" has the marketing power (and weight with investors) and Neural Engine compute will be a major differentiator with graphs that show it as "the best ML computer you can buy"*. We'll see probably in Jan/Feb 2025.
Apple tends to start iteratively with smaller updates and I think began to lay the foundation with M3 Pro. The MBP is not low volume either, compared to the air yes but the MBP is the second-best selling Mac based on the last data I saw, far exceeding iMac, mini, Pro, studio, etc. It's not a low-volume product like the Vision Pro or current Mac Pro.
*on their hand selected tests that are optimized.
Well, it is to keep people updating. Competitors are introducing new sets of processors every year and Apple clearly doesn't want to disadvantage Macs. If Apple believes that people who are really serious about software should make their own hardware, then Apple Silicon is exactly at the heart of hardware. Constantly updating processors brings the benefits of more performance and efficiency. Besides, Apple has adequate resources to devote into silicon design, and economies-of-scale advantages to follow the latest and most advanced manufacture node from TSMC. Anyway, at least it is better than a new color update or new "fancy and environment-friendly" material or new "anodization seal to reduce fingerprints" every year.
Well duh. The m3 MacBook Pro is a done product, so the MacBook team will be working on... *drumroll*... its successor.
It's no longer that simple. We don't even know for sure that the simple doubling model for most things will even hold between M3 Max and M3 Ultra (I suspect THAT will, because an Ultra has literally been two Max dies)
At the M1 generation, many more things doubled directly than didn't (occasionally affected by cutting off a core or two.
M1 Pro was a double M1 in almost every way (except that it not only didn't double e-cores, it actually lost a pair). The neural engine also doesn't ever seem to get doubled, except that the Ultra gets two.
M1 Max took the Pro and doubled GPU and RAM capacity (by doubling memory channels) again, but didn't touch the CPU cores.
M1 Ultra is a literal double Max.
By the M3 generation:
The Pro was knocked back significantly by comparison to the base model, except with extra e-cores. Instead of getting double the CPU P-cores, it gets either 1.25x or 1.5x. The higher end Pro model still gets double the memory bandwidth of the base model, but the cut-down model gets 1.5x. The Pro gets a relatively random multiplier (1.4x or 1.8x) in GPU power, instead of 2x.
The Max also has some differences. It gets more CPU cores than it's ever had - it's a doubling of the Pro's P-cores, rather than a doubling of the BASE model (since there are two P-core variations in the Pro, the two Max models each double one of them). Its memory bandwidth is still double the Pro's, but the lower-end Max doubles the lower-end Pro (just like the CPU cores). The GPU cores don't relate neatly to the Pro, but they are 3x and 4x the base model's complement. The e-cores are exactly the same as the base model and LESS than the Pro. The closest approximation to a full-blown M3 Max is a double full-blown M3 PRO, EXCEPT that the Max has four MORE GPU cores than it "should", has the same neural engine (not doubled), and actually has two fewer e-cores.
Given that the Pro itself is not really a double base model, it's no longer simple expansion. Yes, Apple still uses the same basic cores and just reassorts them, all the way from the iPhone to the Ultra - but those reassortments have gotten more complicated.
I honestly wish that would happen lol. Give us a high performance mode that just obliterates everything…
Possibly a new screen, I mean why not if it drops in.
Mac redesign cycle is 5 years, so I would expect a new M6 MacBook Pro.
So when is the touch screen coming? Meanwhile I'm happy with my 16 inch 96Gb, 4Tb M2 Max despite having to offload files as 4Tb is insufficient. 2Tb micro SSDs are invaluable!