I can’t believe that we’re already talking about M4s, seems like the other day when we were expecting the first ARM Mac chips… time flies.
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's M4 MacBook Pro Lineup: What to Expect
- Thread starter MacRumors
- Start date
- Sort by reaction score
Are they not going to update MacStudio with M3 in June? I want to replace my 2015 iMac asap. I don't know what to get. M2 Max Mac Studio ?
Some signs point to them skipping this one. For example, it seems the M3 Max doesn’t allow for expansion to an M3 Ultra, so perhaps the M3 Ultra won’t exist at all, and instead they’ll go straight to M4 Ultra next year.
Even if all that were true, you can still improve the core design or switch to a different ISA.
Sounds like there will be some significant AI features added this year that won't work on lesser neural engines.
Does sound like that, doesn’t it? At this point though, with my M2 IPP which I have no intention to replace, I would be content with a massive upgrade to Siri, Spotlight, and lookup. I don’t need Dall E type features although they do sound awesone.
The way they are releasing the M series chips is a bit weird, almost overlapping the Ultra models with the next iteration of the chips. And that other rumour this week about Gurman saying that some of the pro desktops won't be getting the M3 adds even more to this weirdness. Maybe they are assuming that pro buyers don't care if they buy a chip from a previous generation. I'm planning to buy a Mac Studio with M3 Max or Ultra this summer, but I see it's not guaranteed that it will be released, and, no, I'm not buying neither a M2 nor an iMac.
HobeSoundDarryl
macrumors G5
Overcoming the “unfixable hack vector” may be- in part- what this apparent rush to M4 is about.
Legal minds among us: if there is absolute certainty this hardware flaw exists in all M1-3 silicon, is continuing to sell it anyway an accumulating class action waiting to happen? Up until they know about the flaw, they have some legal defense, “we didn’t know” room. As soon as a Corp. knows but keeps selling anyway is one of the oldest class action catalysts in the book.
So my guess is that M4 will indeed close that “unfixable” vector and the rush to it is at least partially to get off the “knew but kept selling anyway” legal risk train.
Now, I’ve seen some speculation that they can minimize the “unfixable” actual risk with some software changes that apparently meaningfully slows down macOS. If that’s true, another class legal risk builds… which is slowing down systems but still marketing them on “incredible, magical” speed.
If both apply, perhaps the apparent rush to M4 everything is not just natural flow... or N3E cost savings... or even “A.I.”... but good old legal exposure management. Whatever the catalyst, bring on the next generation. Tech progress is good.
Last edited:
HobeSoundDarryl
macrumors G5
Due to this “problem” coupled with speculation that highest end Macs are most profitable, there is logical speculation that in M4, the barely established order of M-base to M-ultra will flip so that M4 ultra arrives first… and thus can be “king” of the power (and number painted on chip) position for the entire stretch of time until the next M-ultra is released.
Imagine: M4 Ultra rolls out first in a Studio & Pro. Those who burn for latest number on a chip may burn for either Studio or Pro vs. waiting a few months for Mx PRO & MAX, so Apple makes more money. Those who favor cheapest with latest number painted on chip wait the longer for Mbase... AND those who wait for cheapest and- presumably- least profitable will then be hard bumping up against the M++ rumors of the next M-ultra about to hit. Exactly as it is now with Ultra hitting last, having only a few months as "king" of Mx power before M++ MAX arrives with comparable benefits/speed. The window for Apple to make the most profit per unit sold is relatively small that way and they don't even get the easy push a "higher number pained on chip" sales to the must-have "latest & greatest" crowd. In fact, the latter is just the opposite because the psychology is that when Ultra finally arrives, the next M++ chips are about to launch... so WAIT instead of buy now.
Since it appears a new ultra will no longer be 2 MAXes linked together, I take this speculation as at least plausible… as anything “more profitable” seems like it would be prioritized by modern Apple Inc. If the new Ultra will be a custom-designed chip from scratch, it does not have to wait for MAX to be complete (to offer MAX in duo form). Release when ready... even if that becomes FIRST of a new Mx generation.
All speculative of course, but plausible to me… if nothing else for the 💰💰💰
Last edited:
I max my M3Max out doing AI based post edit video denoise. More neural engines will help with some AI video like in edit masking, but man, we just need faster and more GPUs. The M3Max is 40% faster than a M2Max. But still only noise processing at 3.8 fps. Which means sending that beauty work out to a server farm In the back room.
Language models and other AI will also need more GPU to be responsive.
That said, the M3Max is a beast. I can edit multiple 4k,6k and 8k raw video on the same timeline without proxies. Nothing bogs down in the edit process unless I do something crazy like a post production rack focus which uses the neural engines and lots of GPU.
Video editors are always gonna need more computer than the manufacturers can deliver.
All they need to do is give 16gb to the base model. It will end up being the best selling pro ever.
Well, we'll see if that rumour pans out. What would be good is if they could keep the whole range in step so the Mini, iMac and Studio don't keep getting left behind.
Yes, a 3-year upgrade cycle would be adequate for many existing users - and the M1 Max is still a serious piece of kit - personally, these days, I "plan for 3, hope for 5" and I believe a number of business users lease for 3 years. However, Apple also need to have something appealing for potential customers whenever they get the 3/5/whatever-year itch, not to mention attracting new customers to the platform - and for the consumer market that usually means something not 2-and-a-half years old. If I want a computer that might last 5+ years I'd tend to buy something just after it gets updated (maybe wait a couple of months to see if there are teething problems). Sales are going to tail off once a product gets old.
Plus, I guess there's a deep-pocketed hard core who upgrade as soon as a new model appears.
Yes, I think its the lack of a M3 Ultra delaying the Studio/Mini range. Apple have changed the balance of the regular/Pro/Max lineup and dropped the concept of the Pro, Max and Ultra all being based on the same die. In relative terms they've "upgraded" the Max and "downgraded" the Pro (and, as far as I can tell, bumped up the price of the Max). They're still better than their respective M2 namesakes, but M3 Pro is a bit of a disappointing upgrade from M2 Pro whereas M3 Max is pushing more towards M2 Ultra territory.
I'm wondering if the Mini/Studio range will get a bit of a re-shuffle. Maybe they could make a M4 Max Mini (I'm sure anything that would fit in a MacBook Pro would fit in a Mini) and just offer a M4 "Ultra"/whatever superchip in the Studio (= 2025 Mac Pro) and Mac Pro (= 2025 Mac Pro With Slots). That's just spitballing, though.
Faster/more cores is always good, but more and more computing work is being shifted to specialised processing units like GPUs, media engines and neural engines, so that may be where the bulk of future advances occur. Hence the extra media engine on the M1/M2 Max, hardware ray tracing on the M3, the rumoured neural engine upgrades on the M4...
The whole 'symmetrical multiprocessing with multiple general-purpose CPUs' concept was partly because that was much easier to develop/adapt conventional software for than more radical 'massively parallel' or neural network-based computing (Transputer and all that...). Modern graphics rendering and machine learning workloads need the more exotic hardware, though. All of those "ARM powered" or "x86 powered" supercomputers tend to be just using the regular CPUs to control custom vector processors etc.
Looking forward to see what they come up with for M4. How much die space is Apple willing to dedicate to AI acceleration?
Mainly hoping for another graphics jump though. M3's big cores already go toe to toe with my desktop's R7-5800X, I'm hoping that M4 Max can match the RTX 3080.
Mainly hoping for another graphics jump though. M3's big cores already go toe to toe with my desktop's R7-5800X, I'm hoping that M4 Max can match the RTX 3080.
Except that batteries can not be made smaller. A 12" M4 would have a shorter run time before it needs to be charged.
Most of the time on Intel it was yearly until the last few years when Intel slowed down on delivery. We just weren’t paying as close attention to the chip upgrades and the chip lines didn’t have as strong a branding. Now, recency bias makes it seem like this pace is new.
Other companies’ computers are also upgrading each year. Actual customers are not rebuying each year. It’s an escalator that keeps going up but you just get in when you need to get to the next floor.
It may, but I’m not too worried about the recent vulnerability announcement. Do you remember the same kind of vulnerabilities on Intel processors a few years ago? Did we all have to trash our computers and get new ones for that reason? No, it had very little real-world impact.
no joke, this is probably the feature I’m most excited about
Kind of hoped we might see a new OLED panel as I'm not a fan of the Mini-LED displays due to the bloom/halo effect is has on light text on darker backgrounds. Plus having seen the contract, refresh rate, colour clarity offered by OLED everything else just seems second rate in comparison.
Personally I'm holding out for a newer OLED display equipped Macbook before I upgrade my Air M2
Personally I'm holding out for a newer OLED display equipped Macbook before I upgrade my Air M2
Last edited:
That's not so unusual. Intel's high-end chips are often two or three years after related low-end ones.
No, it's just that the volume of the Mac Studio is going to be way lower no matter what, so they don't prioritize it as much.
People say this, but my M1 Pro has an 8p/2e setup, the M2 Pro has 8p/4e, and now the M3 Pro has 6p/6e. Sure, having fewer p-cores is a bummer, but to me, I'd end up with thrice the e-cores! Given that the battery life on my 14-inch 10-core M1 Pro MBP has been merely OK, I imagine that could be quite an improvement.
(Plus, the M1/M2/M3's e-cores have improved at a faster pace than the p-cores, if I'm not mistaken.)