"The first Apple M3 Pro chip benchmarks
have leaked!
It’s actually slower than the M2 Max based
on margin of error.
M3 Pro: 15,173
M2 Max: 15,242"
An alternate, and in my opinion more accurate and relevant, way of phrasing this would be:
"Very cool! The new *mid-level* chips perform as well
(within <1%) as the old *performance-tier* chips, so
you get more performance for less money - or more
perf for the same money."
In what world is this not a good thing? Pretty solid annual incremental upgrade.
Of course I understand why people are underwhelmed with the incremental M2 and M3 gains, after the groundbreaking M1. Now that Apple have got all of that groundbreakingness out of the way, yeah, sure, it's going to be a while before something really groundbreaking again -- but that usually happens with entirely new architectures, and new architectures don't just fall out of the sky on an annual basis. And yeah, maybe we all hoped for more from the 3nm process.
Look at the annual incremental performance increases seen from AMD and Intel. The M3's progress compares favorably. I read some 14th-gen Intel benchmarks and they were mostly indistinguishable from 13th-gen benchmarks.
People are picking on that lede tweet somewhat unfairly. Numbers in the rest of the article say that the same benchmark run between the two "pro" variants improves only 6%. And that's actually quite disappointing for a chip that's supposed to be a on a new semiconductor node. Not a lot of people make a laptop purchase decision over 6%.
> Look at the annual incremental performance increases seen from AMD and Intel and I think these gains compare pretty favorably. I read some 14th-gen Intel benchmarks and they were pretty indistinguishable from 13-gen benchmarks.
Exactly, because Raptor Lake and Alder Lake were fabbed on the same process (10nm++/Intel7, the name changed over the period for marketing reasons). That wasn't supposed to happen here, it was supposed to be new stuff.
> People are picking on that lede tweet somewhat unfairly. Numbers in the rest of the article say that the same benchmark run between the two "pro" variants improves only 6%. And that's actually quite disappointing for a chip that's supposed to be a on a new semiconductor node. Not a lot of people make a laptop purchase decision over 6%.
The M3 Pro has been neutered - the normal M2 Pro* was 8 Performance + 4 Efficiency cores (same as the Max) whereas the M3 Pro is just 6 P + 6 E cores.
If you want the full complement of CPU cores on the M3, you have to get the Max variant.
*There was a special 'low' end M2 pro that only had 6 + 4 cores.
Neutered? What a strange way to describe a chip that is about faster than its predecessor in most benchmarks. You're strangely focusing on an implementation detail rather than actual performance.
If a car company replaced a 4.0 liter internal combustion engine with a 3.8 liter engine that outperformed its predecessor, would you say that they "neutered it" because hey, you're getting 0.2 less liters of displacement?
For me to call something "crippled" or "neutered" or some such it would have to have actual functionality removed, or a meaningful reduction in actual performance. This is the opposite of that.
If you want to call the M3 Pro an underwhelming upgrade relative to the M2 Pro, that's your right and I don't really disagree with you, but I also think it compares pretty favorably to the annual incremental upgrades from Intel and others.
The M3 Pro has been restricted to prevent it from competing with the Max on CPU performance. Whereas the M2 Pro and M2 Max were essential identical on CPU performance (with the exception of the 10-core M2 Pro which was only standard on the low-end 14" model).
To me that feels like neutering.
Doesn't mean it's a bad chip/machine but clearly the product marketing people made the call here.
To neuter an animal is to remove their testicles, eliminating their ability and drive to reproduce and generally making them less aggressive.
That's not what they did to the M3 Pro. What they did was, they improved it a bit relative to the M2 Pro. But not as much as they improved the models that bracket it.
I realize I'm complaining about needlessly hyperbolic tech-related smack talk on the frigging internet, which is sort of like complaining about moistness in the ocean. But still, lol @ describing a modest upgrade as "neutering."
You know what was neutered? The Apple IIgs. Those 65c816 CPUs could go up to like 16mhz, easy peasy. But they stuck a 3.57mhz 65c816 in there and never upgraded it so that the IIgs wouldn't encroach on the Mac. Now that was a real hatchet job.
Sure it might be faster but in context it doesn’t feel like a significant improvement. The base M3 and the M3 Max have fairly large gains over their predecessors, but the M3 Pro doesn’t have nearly as large of a gain (this is all mentioned in the article). It stands out and it’s pretty clear that Apple wants people to “upgrade” to the Max which comes with a +$800 margin.
I’m holding judgement until I see real world performance benchmarks vs synthetic but I fully understand everyone’s reservations with the M3 Pro.
Sure it might be faster but in context it doesn’t feel
like a significant improvement.
So they can't just make things faster in an objective way. They have to make you "feel" a certain thing.
It stands out and it’s pretty clear that Apple wants
people to “upgrade” to the Max which comes with a +$800 margin.
Few thoughts there.
1. Well, they certainly won't mind if you buy the Max. No argument there.
2. As for Apple's intentions I think they were extremely honest during the initial presentation itself. The market for these machines are primarily M1 and Intel Mac users. Very few people upgrade their laptops every single year.
3. The M2 lineup was a bit weird, right? If I am remembering correctly there was not a clear case for the M2 Max versus the M2 Pro. I think the M3 lineup is a bit of a correction there. While I certainly wish every chip got like, a million times faster... the product tiers here seem less confusing.
> It stands out and it’s pretty clear that Apple wants people to “upgrade” to the Max which comes with a +$800 margin.
Eh? Someone who was previous going to buy an M2 Pro would very likely buy an M3 Pro, not an M3 Max.
This is likely Apple managing yield. If they'd stuck with the same core count, the M3 Pro would be faster, sure, but it would probably also be more expensive.
> (10nm++/Intel7, the name changed over the period for marketing reasons)
People like to pick on this change a lot but these names have been marketing for decades, and other fabs (including TSMC and GloFo) named nodes more advanced than they were purely to look better (iirc - GF 14nm was much closer to Intel's 22nm than to 14nm, Samsung 8nm performed way worse than TSMC N7 etc.) - meanwhile "Intel 10ESF" aka "Intel 7" is actually very similar in most metrics to TSMC N7, though is more amenable to higher clocks.
> Not a lot of people make a laptop purchase decision over 6%.
How many people make laptop purchase decisions over performance any more? I’d think the GPU boost and things like AV1 are going to matter more, which is probably why Apple focused on them.
Yup. And good bump in single core performance, which you'll notice for most use cases. So really mid level part beats top tier part from last generation.
The difference between the M2 Pro and M2 Max was quite subtle, especially on the CPU department. Same exact CPU cores except the Pro had a binned variant with two P cores disabled. The biggest difference between the two was on the GPU front. If they're testing the CPU, then the M2 Pro and M2 Max should have similar performance (assuming non-binned M2 Pro).
> The M2 was launched in June 2022, but the Pro and Max launched in January 2023. They're not even a year old and have already been relegated to the trash heap of silicon history. It stands to reason the M2 chips didn't sell very well, so Apple wanted to get the M3 out ASAP to reverse that trend.
That theory would make a lot more sense if the M3 was providing a big performance boost.
Instead buyers of the M3 Pro get 25% less memory bandwidth than the M2 Pro, and two of the performance cores swapped for efficiency cores. Not much to reinvigorate sales by motivating M2 Pro users to upgrade other than a higher number.
The M1 Pro came in binned 6+2 and I’ll bet most buyers took that configuration to save money or upgrade to Max if they want more performance and RAM. Apple could be providing a good upgrade for what people really bought, and with 6+6 and more single core score it is a real upgrade. Is it worth it? There’s no other improvements than the chip anyway.
> Not much to reinvigorate sales by motivating M2 Pro users to upgrade other than a higher number.
Do you, in 2023, know _anyone_ who replaces their laptop every year? Realistically, virtually no-one does this; the primary market for this is actually people who are still on Intel chips.
Exactly! I am about to upgrade from M1 Air to the new M3 pro, and although the gains will not be as huge as from Intel - ARM, I still expect a significant upgrade for a ~3 year old laptop.
I feel like the M3 has been a bit of a misfire, launch-wise.
In between all this and the fact that they're all shipping with Ventura and cannot be upgraded to Sonoma yet to me says that for all the claims of "Apple may not be first, but it will always do it best/right" are evaporating.
Seems like this is a bit of a rush.
And to all the people working the "They're comparing it to the M1, because that's who they're targeting for upgrades"?
Bullshit.
Try saying that about Intel, if they start benchmarking 14th gen chips against 11/12th. "Because they're targeting those users, not 13th gen owners".
Think about how that sounds. Think about how you'd laugh if they tried it. Think about how Apple is spinning that, and you're buying into it. Hmm.
Just not a good look. I can imagine there'd be howls of laughter from some corners if Microsoft released a new Surface tablet and said "BTW, it will come with Windows 10 and there's a bug that won't let you upgrade to Windows 11 until we fix it..."
I wonder what kind of usage metrics Apple has that led it to reduce (8 down to 6) the number of performance cores and increase (4 up to 6) the number of efficiency cores on the M3 Pro? Or the reduction (25%) in memory bandwidth?
Maybe they're trying to get people to buy the Max version? The spec differences are interesting:
I think differentiating the pro and max is a big part of it, but it also seems like they're trying to spread out the benefits of the switch to 3nm over time. I would expect performance cores and bandwidth to go back up next year to maintain steady performance increases year over year.
Way outside of my field of knowledge but would it be around prioritizing battery life above all else on the pro? Would also nudge people who _need_ the higher power features to not stick with Pro too, obviously.
That confuses me because shouldn't those people just get an M3? and then those who get the M3 Pro want better performance. I don't understand the 'pro' naming scheme means if they're not serious about making it pro.
What is Pro to you? I ask, because it can be a huge range of professional tasks and many won’t need the higher performance of the Max, but will benefit from the efficiency of the Pro.
The Pro is still a significant jump over the base M3, both in performance and capability across every metric. It just happens to be a mid range SKU and for some reason people seem to think in binary high/low terms.
Isn't it basically the same reason loads of things will have sizing options along the lines of medium/large/massive instead of small/medium/large.
At this stage I'd say "Pro" in terms of Macbooks is is more likely to be for people who want something more powerful or a bit fancier than an Air than people who need something more powerful, so there needs to be another level above that
Every CPU lineup has an intermediate; Apple actually has fewer than most, and I'd expect they'll probably add one or two more SKUs as time passes to deal with binning more efficiently.
Reducing memory bandwidth is particularly bad given that most AI work loads are memory bound.
With the memory bandwidth Apple's fully integrated GPU with shared memory is actually pretty good for smaller AI stuff like running and fine tuning llama2-sized models. It gives you a decently fast GPU option with a lot of RAM. NVIDIA cards with similar RAM are a decent fraction of the price of a whole Pro or Max M chip laptop.
The neural engine in Apple chips is really for small models for doing things like Face ID, local inference for Photos.app, etc. It's there to run those small models with very little power consumption. It's not that great for huge models. The GPU is better for those, and they have been adding cores and performance there.
But yes, most large models are pretty heavily memory bound.
Both really, since models are pretty much 100% "hot" RAM. The bigger the model the more bandwidth it demands too since evaluating it means streaming the entire thing through the processor.
It's called market segmentation, and Apple is very good at it, but in the early days of M1 (and M2, which is just a warmed-over M1) they didn't have the engineering time to make 4 tiers (Base, Pro, Max, Ultra) and had to do a half-assed job of chopping off part of the Max to make the Pro. With M3, the Pro is its own design, with carefully designed limitations to force you to the Max.
Also, I see the discussion about pro vs max being focused on performance, but what about thermals or battery life? Perhaps many professional users would be better off using pro and not max?
> Perhaps many professional users would be better off using pro and not max?
That would be entirely dependent on whether they benefit from the extra performance or RAM that you can get with Max.
I don't think that there's a significant difference in thermals or battery life on identical workloads. My M2 Max gets me through a full working day with 30-50% charge still left. It's the only laptop I've ever had I can be completely confident about should I not bring a charger with me.
Even so, my regret over buying an M2 Pro just before the M3 was released decreases each day. In fact, it might start going negative such that I’m happy to have bought the M2.
To be honest, I was in the same boat with M1 -> M2. Not that the M2 wasn’t an improvement, and as pointed out this article compares M2 Max to M3 Pro, it’s just that M1 was so much better than my i7 based MBP that the gains made by M2 weren’t significant enough to worry.
I have a work issued M1 Max, which has an incredible amount of computational power.
I wanted a personal computer, and purchased an identical machine, M1 Max, 64Gig, whatever. I realized 4 days later that it wasn't possible for me to push my personal machine anywhere near this machines limits. I returned it, and bought an extremely cheap ($240?) MacBook Pro, i5 - 2017.
It hasn't slowed me down, and my strategy is to use macOS and not sign in with my Apple ID. No Mac App Store, iCloud, etc. I use App tamer and little snitch to eke out any performance gains possible... it's an excellent machine, and I'm so happy with it.
I want to thank Apple for introducing this express route in the rat race, which has lead to an extremely friendly second market of intel, and 1st generation M's.
I implore you to ask yourself if you really need the next generation, or even the previous one before purchasing
Sorry but this is the worst machine Apple has made in possibly around 20 years. Feels heat constrained in it's own chassis, fan turns on at the drop of a hat, battery not big enough for sustained work and started durning off at 20% on older units when you open a video call, charger it ships with not powerful enough to jump start from empty without a 5 minute wait, keyboard feels gross makes too much (ugly crunchy sounding) noise and keys jam from everyday dust.
M1 was heaven after that shambles and I can never go back to hot, noisy and low battery laptops after this.
In my young years I had little understanding of the value of time, and would spend an inordinate amount of time on eking out that last little bit of perfection (whatever "perfect" looked like to me, anyway).
Bragging rights fits in here as well for a certain cohort. "Oh, you got 5.551Ghz, not bad but I'm at 5.5511Ghz".
Another classic example of this is uptime bragging for some linux fans, as a way to both one-up each other and to one-up the Windows community, which (more so in the past) had pretty awful uptimes due to various bugs and the need to reboot after making config changes. (We used to joke: "You have moved your mouse, please reboot Windows for this change to take effect.")
I think most people in this kind of community though just enjoy tinkering and maximizing performance/whatever.
The heat, battery life, speakers, microphone, display, connectivity, and more have improved in that time as well. I'd rather have a low end new Mac than a high end old Mac.
And now your platform is going to stop getting OS builds potentially as soon as next year because Apple switched to a new architecture. There’s a reason that system cost less than an iPad.
Apple was basically begging Intel users to upgrade to Apple Silicon at the most recent keynote. I think Sonoma is the last Intel build. Apple drops a ton of technical baggage if they can stop supporting Intel.
The machine you should have purchased is the M1 MacBook Air. You can find a used one in the $500-600 range or a new one in the $700’s and it will smoke the 2017 Intel system while returning double the battery life and making zero noise with its fanless design. It’s comfortable on a lap and doesn’t get hot.
Look, it’s bad that technology is disposable, but if you buy a Mac you are getting ~8-10 years of feature updates and an additional ~2 years of extended support security updates. Maybe after that you’ve got OpenCore Legacy, but that won’t help you if Apple doesn’t release binaries on your processor architecture.
If you buy a PC you’ve got a system that is better designed for alternative operating systems and long-term support. But when you buy a Mac that’s just part of the deal, you’re in the platform. I consider my Mac system to be more of a leased tool and I’ve accepted its limited lifespan as part of the compromise of that platform.
If you buy Apple’s previous architecture you’re getting a much worse experience, and there’s history to back up that fact. What you did was equivalent to buying an iBook G4 in 2008. Your system is going to either be a brick or a jankier-than-average Linux box in 2-3 years (i.e., a ThinkPad makes a better Linux box than a MacBook).
Finally, if you don’t even use iCloud or an Apple ID I’m not sure what the appeal of a Mac is in the first place, especially one that isn’t on a particularly unique or advantageous architecture. I’ve got a Mac because I prefer an iPhone and I’m kind of stuck with it if I want to send texts from my computer. If you’re not even using anything involving an Apple ID you might as well be on Linux or Windows.
I don't think buying a new Macbook (or any computer) sooner than after 5-7 years of use is pretty meaningless for most cases.
But going for a 2017 one is simply strange considering the differences in all areas. I switched from 2019 Pro to M2 Air and it's simply wonderfull. It's twiceas powerfull and completely silent and cold.
2019s Pro was good, but it's not nearly as fast for photo editing and hot as hell even if you just code in nvim.
Honestly, you made an incredibly irrational choice. You went from what is likely a $2500 machine to a $250 machine that is more than 10x worse, but you could have gone from a $2500 machine to a $1200 machine that is less than 2x worse by buying a new M1 MBA or similar instead.
The fact you seem proud of this choice and are holding it out as some sort of virtue as if the difference between an Intel Mac from 2017 and an M1 Mac is inconsequential clout chasing is the cherry on top. The efficiency gains alone in the higher performance and lower power usage make running a non-M1 Mac an environmental loss in comparison, considering that as you so clearly point out, Mac's have extremely long usage lifespans.
Seems to be based on the perspective that the focus is indeed on absolute performance gains rather than performance per power draw.
Based on the range of devices the M chips will be used for both immediately and in future it's not at all obvious that should be true. Think about the vision pro. Name of the game there is workable performance / battery tradeoff, not maxing out raw performance.
Same is true for most users of macbooks frankly, including (lots of) engineers using pros.
It wouldn't be surprising to me that an acceptable outcome of the shift to the new node was maintaining performance of M2 with gains in battery life. More performance will inevitably come with M4 etc anyway.
I think you nailed it. This is more than plausible, given Apple's penchant for valuing the packaging of technology (its application) over the sheer horsepower of cutting-edge tech. More performance per power draw would 100% fit with that and enable a lot of applications; Vision, Apple Watch, etc.
It's amazing how often tech media works themselves over.
Right. Apple machines have never been top performers. The M* machines are just good compromise machines when you need to be mobile and are energy constrained. Or sometimes, if you spend $6000+, they can have competitive absolute performance specs, re: available RAM and RAM bandwidth for tasks like LLM.
One of the interesting things I've noticed is that the transistor count for the M3 Pro hasn't gone up much.
For example, the regular went from M1 16B, M2 20B, M3 25B; The Max went from M1 57B, M2 67B, M3 92B. On the other hand, the Pro went from M1 33.7B, M2 40B, M3 37B. For the regular, transistor count is up 56% over the M1 and for the Max it's up 61%. However, for the Pro it's only up 10% over the M1 (and slightly down from the M2).
In terms of the cores, the M1 Pro was either 6+2 or 8+2 Performance+Efficiency cores. The M2 Pro became 6+4 or 8+4 so you still had 6 or 8 performance cores. The M3 Pro is either 5+6 or 6+6 so fewer performance cores.
It just seems like the regular has kept the same 4+4 and gone up a bunch in transistors and the Max has gone from 8+2 to 8+4 to 10+4 or 12+4 (a decent step up in cores and transistors). The Pro, on the other hand, hasn't been given the same treatment. It's still a great processor, but it feels like it isn't quite the value that it was. The M1 Pro offered such huge performance with up to 8 performance cores. The M3 Pro feels like a bit of a weak spot in Apple's lineup where it'd be worth deciding between the M3 Max or the regular M3 depending on your needs.
Still, I love Apple Silicon. It's amazing having a laptop that just feels so good to use that I don't think about raw performance numbers. Things just work without pain or heat/fans. I know that's hard to quantify, but my M1 MacBook Pro has felt so different from all the computers I've used before. I was using a 28W Intel 2020 MacBook Pro and my M1 MacBook Pro just felt completely different in terms of performance. I guess part of that is the raw performance with an M1 getting single-core performance that is 85% faster and multi-core performance that's 2.8x faster, but it's also that Apple did it at power/thermal level that means I never hear a fan and my laptop never feels warm. Maybe a big part of it is just that Apple Silicon was such a huge improvement that I'm still in awe of it. Maybe part of it is that I haven't seen what Intel has done since 2020 (and they've been catching up after years of mediocrity). Still, for the first time I feel like I'm waiting on non-CPU stuff for a reason to upgrade (like an OLED or microLED display or something).
Last year's benchmarking software benchmarks last years top-tier chips on par with this year's mid-tier chips... Go figure. What's the story here?
Not saying it's wrong, but it's always best to wait for benchmarks of real-world scenarios covering what you're actually trying to do, especially when this launch was so GPU-focused on Transcoding, Ray Tracing, and ML.
Live by benchmark results and slick performance marketing, die by benchmark results and slick performance marketing. The market is ruthless; if you aren't delivering more of what you promised made you great last year, you're failing.
Really the underlying news here, which the article doesn't really treat with, is that TSMC's 3nm node is sort of a bust. It's not looking like a full node upgrade at all (something that's been reasonably clear for a few months based on the leaked pitch/density numbers). They still have a comfortable lead over Intel, but scaling limitations come for us all eventually.
At this point, Intel's roadmap seems unreasonably optimistic. They haven't launched processors with Intel 4 yet. They're set to be released December 14th, but we aren't seeing Intel launch a full line of processors with Intel 4.
If Intel 4 is coming in December 2023 in a very limited way, are we really expecting 20A in 2024?
I'd love for Intel's roadmap to be real, but it just doesn't seem likely. I guess it also depends on what you mean by "in 2024". For example, Intel 4 came to us "in Q3 2022"[1], but we aren't seeing any processors until 17 months later. If Intel 20A comes to us "in Q3 2024", but we don't see any processors out until December 2025, that's a big difference. It's an especially big difference if it means that it's in processors in December 2025 in a very limited way.
TSMC's 3nm tech is shipping in Apple Silicon CPUs at high volume.
Intel seems to say "Intel 20A Q3 2024", but that isn't when we get CPUs with that tech. Maybe it's when their teams can create limited volume samples? Realistically, it seems like there's 18+ months between Intel's roadmap and the CPUs getting into something you can buy.
Again, I'd love Intel's roadmap to be accurate, but if Intel 4 started "in Q3 2022" and we're getting limited laptops with it in December 2023 or January 2024, that seems like Intel saying that they've gotten to a point well before they can actually get that into our hands.
> I'd say gains from M2 Max to M3 Max are pretty reasonable.
How are they reasonable to you? As I see it, they just added 50% more (performance) cores to the M3 Max compared to the M2 Max (12 versus 8, respectively) and achieved 50% higher multi-core performance: https://browser.geekbench.com/v6/cpu/compare/3434546?baselin.... No rocket science here.
If anything, we should be asking “why only +50% for multi-core performance when single core performance increased by ~20%?”.
Low performance gains for a mid-level chip are totally a good thing if it's significantly more efficient. (I haven't seen any numbers for M3 Pro yet, but here's hoping.)
Of course I understand why people are underwhelmed with the incremental M2 and M3 gains, after the groundbreaking M1. Now that Apple have got all of that groundbreakingness out of the way, yeah, sure, it's going to be a while before something really groundbreaking again -- but that usually happens with entirely new architectures, and new architectures don't just fall out of the sky on an annual basis. And yeah, maybe we all hoped for more from the 3nm process.
Look at the annual incremental performance increases seen from AMD and Intel. The M3's progress compares favorably. I read some 14th-gen Intel benchmarks and they were mostly indistinguishable from 13th-gen benchmarks.