M2 vs. M3 Buyer's Guide: How Much Better Really Is M3?

[Razorpit]

The main reason to purchase an M3, it’s going to be around a little longer when Apple decides to start artificially deprecating support for older M1’s and M2’s in OS updates.

 

[il_teo77]

My MBP M1 Pro is working like a charm. I can only bring it to sweat while doing render work in ArchiCAD. And I didn't do nearly enough to justify an M1 Max, and I still don't do enough to justify an upgrade. Plus the app was rewritten for Apple Silicon so I immediately gained speed compared to the Rosetta 2 version. The switch to Apple Silicon was so impressive that, for the ones that made the jump right away, it will take some time to justify an upgrade.
Let's hope Apple doesn't switch back to portless designs, introduce some butterfly keyboard or something...

 

[Razorpit]

Apple Silicon Macs are likely to be supported for years and years.. the fact they are still releasing new models with 8 GB base RAM means even the first M1 Macs with 8 GB should do fine for years and years to come. Apple have created themselves a huge number of customers on new hardware with 8 GB RAM... who will be expecting at least 4 or 5 years run with what they just bought.

Irrespective of what they are actually running on it, in terms of third party software.

They’ll come up with something to artificially deprecate them. Remember how screaming fast Intel’s were at the time compared to the PowerPC’s? Yet this timeline happened:
- January ’06 first Intel machine (32-bit) announced.
- June ‘06 64-bit Intel processors announced.
- August ‘09 MacOS 10.6 releases requiring 64-bit processors.

If anything Apple is even more fiscally ruthless now. I know things are not the same with these processors, but I have every confidence they’ll come up with something.

 

[ifxf]

I wonder over the next few years how many security issues researchers will discover in the Neural engine.

 

[ifxf]

You can do photo and video on an iPhone and on a iPad. There is no “chip” for photo editing.

Besides, a M3 Max is better at everything than a M3 Pro, so there is no reason to pic a M3 Pro over a M3 Max at all.

Coding is just creating text files, you can do this on a toaster even. There is no dedicated M3 Pro chip for coding.

Not true, part of coding is compiling and debugging, both can take advantage of extra RAM for caching frequently used files.

 

[fuchsdh]

Who do you see the M3 Pro as being the ideal choice of chip for now? I understand why the chip is positioned close to the base chip and far from the Max, because it apparently doesn't cost much more to make than the base chip.

Depends what you consider "far" and "near" here. Yeah, the paper specs on stuff like core counts and RAM are certainly closer to the M3 than the M3 Max, but that's arguably a function of a lot more users needing "better than M3" than users who need "nearly an M3 Max". Far more people only need up to 36GB than they do 128GB, or use up to 2 or 4TB of storage versus 8TB. And as with most tech, you pay correspondingly more for increasingly marginal gains when you're spending on professional hardware.

They’ll come up with something to artificially deprecate them. Remember how screaming fast Intel’s were at the time compared to the PowerPC’s? Yet this timeline happened:
- January ’06 first Intel machine (32-bit) announced.
- June ‘06 64-bit Intel processors announced.
- August ‘09 MacOS 10.6 releases requiring 64-bit processors.

If anything Apple is even more fiscally ruthless now. I know things are not the same with these processors, but I have every confidence they’ll come up with something.

Arguably this has already happened, in that the base M1 Macs don't support MFI-certified assistive hearing devices being paired directly in Sonoma, while the M1x variants and newer can.

But speaking to your point, I think that it's likely the MX chips get a longer support window than Intel and a lot of specific eatures won't come to the old models, just like they have been recently with Intel support of certain features.

 

[ric22]

Depends what you consider "far" and "near" here. Yeah, the paper specs on stuff like core counts and RAM are certainly closer to the M3 than the M3 Max, but that's arguably a function of a lot more users needing "better than M3" than users who need "nearly an M3 Max". Far more people only need up to 36GB than they do 128GB, or use up to 2 or 4TB of storage versus 8TB. And as with most tech, you pay correspondingly more for increasingly marginal gains when you're spending on professional hardware.

Go on then, list a use case for someone that needs more CPU/GPU power than an M3 but needs less power than is provided by an M3 Max?

👀🤷🏼‍♂️

And no listing artificially created constraints like ports and RAM and SSD size limits, please. Let's just focus on the compute power and not on the extended options opened up in the purchasing menu when you buy a pricier chip.

 

[Elusi]

Using that logic, the M3 is the first 3nm design with a lot of new changes and features in the chip, that undoubtedly will be refined and improved quite a bit in the next iteration as Apples engineers work out the kinks. M4 is likely to be a more refined and performant in the same way as M2 was a refined and performant M1. The M3 is also a bit of a dead end in the respect to its manufacturing process which has very poor yields and will be dropped as soon as an alternative is ready to go. Many of the M3's are either binned or designs that are compromised in some ways to get the most of each wafer. It's very much a version one product all over again, in design and manufacturing and if you are worried about that kind of thing, you would definitely stay away from M3 and hope the M4 fixes a lot of that. Honestly the M2 is the best buy in respect of the fact it's the best of the 5nm process. The M3 will most certainly not be the best of the 3nm process by a long shot and I suspect in time will be seen as a poor first outing all things considered.

Software support follows changes in architecture, not process node. Binned chips, new process, none of that usually matters. There is likely going to be a new iteration of the Metal API within a couple of years where the M3 is more compliant than the M1 or M2. That could lead to the M1/M2 being dropped by future OS's sooner than the M3. But again, it's just soft musing over "future-proofing", going by what has happened in the past.

I would disagree that you're going by my logic with that response tbh.

 

[Jack Burton]

A shame more outlets don't cover the difference RT cores make for 3d rendering. Cinebench 2024 supports the RT cores in M3 generation. M3 Max made a leap because of this!

 

[karranz]

what a m**s

 

[WC7]

Using that logic, the M3 is the first 3nm design with a lot of new changes and features in the chip, that undoubtedly will be refined and improved quite a bit in the next iteration as Apples engineers work out the kinks. M4 is likely to be a more refined and performant in the same way as M2 was a refined and performant M1. The M3 is also a bit of a dead end in the respect to its manufacturing process which has very poor yields and will be dropped as soon as an alternative is ready to go. Many of the M3's are either binned or designs that are compromised in some ways to get the most of each wafer. It's very much a version one product all over again, in design and manufacturing and if you are worried about that kind of thing, you would definitely stay away from M3 and hope the M4 fixes a lot of that. Honestly the M2 is the best buy in respect of the fact it's the best of the 5nm process. The M3 will most certainly not be the best of the 3nm process by a long shot and I suspect in time will be seen as a poor first outing all things considered.

odd even cadence

 

[theluggage]

Go on then, list a use case for someone that needs more CPU/GPU power than an M3 but needs less power than is provided by an M3 Max?

👀🤷🏼‍♂️

And no listing artificially created constraints like ports and RAM and SSD size limits, please. Let's just focus on the compute power and not on the extended options opened up in the purchasing menu when you buy a pricier chip.

In theory, one big difference should be battery life - the M3 pro has increased the efficiencyerformance core ratio so it should be a sensible compromise for people who actually want to use it on the road. I don't know if any good pro vs. max battery life tests have appeared yet.

Also, it depends what you mean by "artificially created constraints" but I'm pretty sure that the number of RAM configurations, ports and supported displays are down to the number of TB4 controllers, displayport controllers and memory controller channels physically implemented on the chip. So you really can't dismiss things like needing a third display as reasons for wanting the M3 pro over the M3.

 

[fuchsdh]

Go on then, list a use case for someone that needs more CPU/GPU power than an M3 but needs less power than is provided by an M3 Max?

👀🤷🏼‍♂️

And no listing artificially created constraints like ports and RAM and SSD size limits, please. Let's just focus on the compute power and not on the extended options opened up in the purchasing menu when you buy a pricier chip.

Your focus on "needs" is a bit off. A base M3 can competently handle most tasks thrown at it; I'm not aware of any software that just won't run. It's a matter of doing something faster and speeding productivity. I really don't see what your point is trying to create this fictitious person who is only served by an M2 Pro and will not be served by an M3 Pro (where the regressions are slight, and irrelevant in the sense that if you're someone upgrading every single year money clearly has no point to you and who the hell cares? Next year will be faster by some marginal amount too, and so on and so forth.)

And the attempt to hand wave the other differences as "artificially created constraints" is missing the point. You can be upset Apple doesn't make the exact precise chip that prioritizes exactly what you want, but you can also be upset that you can't get an M3 Pro machine for $5. It doesn't change the fact that Apple has structured their chips the way they have; as you go up in price, you're paying not only for cores but greater capabilities with ports, RAM, etc. The use case for the M3 Pro is the same as it ever was. Do you want more than what an M3 gives you? Then pay more.

 

[ric22]

Your focus on "needs" is a bit off. A base M3 can competently handle most tasks thrown at it; I'm not aware of any software that just won't run. It's a matter of doing something faster and speeding productivity. I really don't see what your point is trying to create this fictitious person who is only served by an M2 Pro and will not be served by an M3 Pro (where the regressions are slight, and irrelevant in the sense that if you're someone upgrading every single year money clearly has no point to you and who the hell cares? Next year will be faster by some marginal amount too, and so on and so forth.)

And the attempt to hand wave the other differences as "artificially created constraints" is missing the point. You can be upset Apple doesn't make the exact precise chip that prioritizes exactly what you want, but you can also be upset that you can't get an M3 Pro machine for $5. It doesn't change the fact that Apple has structured their chips the way they have; as you go up in price, you're paying not only for cores but greater capabilities with ports, RAM, etc. The use case for the M3 Pro is the same as it ever was. Do you want more than what an M3 gives you? Then pay more.

What?? Are are you confusing me with someone else?? Read my post again- I never mentioned comparisons between M3 Pro and M2 Pro.

Reading failure aside, why did you need to be rude in your reply? Upset I can't buy an M3 Pro for $5? How could you misconstrue what I said so badly that you'd write that?

 

[Viale Montanane]

Where are the performance vs power stats and graphs?

 

[DrWojtek]

M2 pro is much better than an M3 base.


I just got a new 16 GB ram 2 TB M2 pro mini. Was going to wait for an M3 pro mini but from what I'm reading it would be a very small increase and not worth it especially when they gimping the M3 pro

Depends. If you shelled out for the 12 core M2 Pro you’re right. If you’re talking about the default 10-core M2 Pro though, absolutely not. The M3 is faster in single core and practically ties in multi-core.

Which is why Im waiting for the M3 Mini with 24 GB RAM. Will last me at least five years. Hell, my 2018 MBP 15” is still my only computer and it does fine. The M3 mini will probably last past 2030.

 

[ric22]

So you really can't dismiss things like needing a third display as reasons for wanting the M3 pro over the M3.

Pretty sure I said it was the only reason I could think of earlier...

The reason I said the problems were artificial was because they're using a tablet chip (M3) in laptops, rather than their base chip without such limitations (M1/2/3 Pro).

 

[Viale Montanane]

The future is RISC-V. Apple is looking at it. It will level the playing field.

 

[DrWojtek]

Pretty sure I said it was the only reason I could think of earlier...

The reason I said the problems were artificial was because they're using a tablet chip (M3) in laptops, rather than their base chip without such limitations (M1/2/3 Pro).

How is the M3 a tablet chip? There is no useful tablet to buy, they are all a bad compromise from a laptop with a physical keyboard.

 

[seek3r]

The future is RISC-V. Apple is looking at it. It will level the playing field.

Highly doubt they’ll do a big move anytime soon, they just switched over fully to ARM and RISC-V still has teething pains and the ISA isnt as mature or full as ARM’s not to mention *vastly* less support on the software side.

Even more importantly, since apple controls the silicon design, doesnt use the ARM reference designs, but exposes the ARM ISA, theoretically they can pull in portions of RISC-V if there’s a compelling reason or even swap out completely under the hood and continue to expose the ARM ISA if they really wanted to. Apple’s got it’s own arch functionally already, nothing stopping them from borrowing from RISC-V now without changing anything on the system level or userland. I think AS will continue to evolve under the hood while continuing to expose the ARM ISA.

That *may* mean some things from RISC-V end up in the silicon, though I honestly dont think so anytime soon, and I highly doubt anything above that is shifting off ARM for at least another decade or two.

 

[KPOM]

They’ll come up with something to artificially deprecate them. Remember how screaming fast Intel’s were at the time compared to the PowerPC’s? Yet this timeline happened:
- January ’06 first Intel machine (32-bit) announced.
- June ‘06 64-bit Intel processors announced.
- August ‘09 MacOS 10.6 releases requiring 64-bit processors.

If anything Apple is even more fiscally ruthless now. I know things are not the same with these processors, but I have every confidence they’ll come up with something.

Though it’s already been 3 years and Apple is still selling the M1 Air. We’ll see if the next version of macOS includes any M3-specific features, but it will definitely still support the M1.

 

[KPOM]

Highly doubt they’ll do a big move anytime soon, they just switched over fully to ARM and RISC-V still has teething pains and the ISA isnt as mature or full as ARM’s not to mention *vastly* less support on the software side.

That *may* mean some things from RISC-V end up in the silicon, though I honestly dont think so anytime soon, and I highly doubt anything above that is shifting off ARM for at least another decade or two.

ARM’s IPO filing said that Apple has extended their license through at least 2040.

 

[seek3r]

Though it’s already been 3 years and Apple is still selling the M1 Air. We’ll see if the next version of macOS includes any M3-specific features, but it will definitely still support the M1.

My bet is the M1s get 2012 MBP level support, 8 years of running on the latest OS and 2 or 3 years of security patches beyond. They know there was a huge buy of M1s at release, and they’ve been selling M1s consistently since, kinda the same way the 2012s were. If it’s consistent with the 2012 machines on how long the legs are the 2027 MacOS release should be their final fully supported release, getting dropped in 2028 and security patches through 2030 or 2031.

I wonder if internally Apple has some sort of designation for “LTS” products, the watch series 3 and the iphone 6s come to mind as well

 

[ric22]

How is the M3 a tablet chip? There is no useful tablet to buy, they are all a bad compromise from a laptop with a physical keyboard.

If you don't think they'll put the M3 in an iPad in the next twelve months I'll take that wager. Want to put $100 on it?

 

[DrWojtek]

If you don't think they'll put the M3 in an iPad in the next twelve months I'll take that wager. Want to put $100 on it?

I think they will, too, but it is way overpowered for the iPad. The iPad is a large iPhone. Good for media consumption. Not so good for anything else.