Delayed shipping times for current models sometimes means an update is imminent.
See full article...
See full article...
Signs point to a sooner-rather-than-later M5 MacBook Pro refresh
- Thread starter JournalBot
- Start date
Leaping off this: I primarily use a Mac for native application development — i.e. mostly Xcode and Instruments — and although it used to be the case that there were real and obvious advantages to moving up the range, nowadays a MacBook Air is largely indistinguishable from a higher-end Mac with the same RAM footprint. Incremental builds take a negligible amount of time, and even on an M1 everything feels speedy enough.
I'd be curious what sorts of development are still pushing people further up the price range. In my imagination it's a lot of web-centric stuff with server containers, or possibly cross-development things where maybe you want a couple of OSes at once, but I suspect I'm being ignorant. What have been the experiences of others around here?
Upvote
3
(4
/
-1)
I suppose my point is more that an up-spec'ed Mac Studio is going to be just as good for software development as any Mac Pro.
As far as I'm aware, the only benefit of a Mac Pro vs. a Mac Studio is that the Pro has expansion slots.
But what kind of expansion card would help with development?
Last edited:
Upvote
0
(0
/
0)
I'm an audio professional who uses PCIe cards. When the trash can Mac Pro came out I went through a lot of hassle and expense to move my cards from inside an older Mac Pro to a Thunderbolt external chassis. When the trash can Mac Pro became obsolete I saw no reason to go back to Mac Pro internal solution when I could use the same chassis with a maxed out Mac Studio Ultra that has more power than I'll ever use.
Upvote
14
(14
/
0)
It's great compared to Intel, but pretty bad by historical standards. Within my lifetime of following PC specs, for a short while performance would double within 18 months, then it became a doubling within 24 months and that held roughly steady for many, many years (Moore's law).
Now we consider a doubling in 4+ years to be pretty good, even great!
Upvote
-10
(1
/
-11)
The thing is, Moore's law was never actually a law. It was an expectation, that became reality by people working to match its predictions.
We are approaching hard physical limits based upon the actual, physical size of the silicon atom (an atomic radius of 111 picometres, and a Van der Waals radius of 210 picometres; 1000 picometres equals one nanometre.) The closer we approach those limits, the harder it becomes to make more progress, and there'll come a point where we can't make more progress at packing more transistors into a smaller space. At that point, Moore's law comes to a screaming halt.
If you are basing your expectations on historical improvements, you're ignoring that fundamental fact. And that's before we get into the insanely high levels of precision engineering required for extreme ultraviolet etching, which is required for these tiny nodes TSMC is putting out.
Frankly, the fact that we're still managing die shrinks is verging on miraculous.
Upvote
16
(16
/
0)
AIUI, the “nm” part of a process name (e.g. “3nm”) is as much a marketing term, meant primarily to differentiate from the previous process, as it is a descriptor of any physical property. Am I wrongthinking here?
Upvote
3
(3
/
0)
It depends on your workload and the absolute numbers too though. If you got something that used to take ten minutes and now takes five, that sounds like a decent improvement. Ten hours to five hours would be a big win.. but if your biggest workload used to take ten seconds, cutting that to five doesn’t really seem like a large benefit.
Upvote
5
(5
/
0)
I haven't seen a dev use a Mac Pro in probably 10 years. The Mac Pro exists for A/V pros who need PCI slots. Everyone else should be using an Ultra at most.
Upvote
4
(4
/
0)
If you are a developer of software for those expansion cards. That's about all I can think of. Even if you're doing AI, you're better off with a cluster of Studios connected via Thunderbolt.
Upvote
3
(3
/
0)
Try telling that to my poor little M2 14" Macbook Pro that threw a fit when I tried loading up an UnrealEngine project on it the other evening to troubleshoot something that came to me while sat on the sofa (TBH I think the issue was more likely lack of RAM tho)
Upvote
2
(2
/
0)
My M1 Pro (14”) is a boss for actually creating stuff in Blender - the UI etc. is smooth as silk - but it’s the rendering that is still firmly in “go and make a cup of tea while it runs” territory.
Upvote
3
(3
/
0)
It doesn’t matter. The processors all do work at the same pace.
It's a fallacy to believe a faster machine = more work output.
It's a fallacy to believe a faster machine = more work output.
Upvote
-10
(0
/
-10)
Well that's just the reality of all of the low hanging fruit already being taken and node shrinks being harder to come by and ramping up in cost to get there at that. In single core performance in particular we were told long ago that fast gains were done before Apple got to the party and their continual pushing of this has been quite impressive. We have to take it in the context of 2026 and not 1996.
Upvote
8
(8
/
0)
No, you're not. For example, TSMC's "3nm" processes have a gate pitch of 45-48nm. But the fact remains that each step in process does represent a reduction in size; even if the marketing name doesn't accurately reflect the actual size, transistors are still getting smaller, and there is still a hard physical limit that we're approaching.
Upvote
4
(4
/
0)
Latency—their being connected over TB5 will never compare to the cards communicating over PCI-E 5, much less talking to each other P2P which Apple could decide to do. If you’re serving inference results to lots of employees this is actually not that big a deal but relatively monolithic workloads suffer.
Speed—again, the amount of time a cluster of Studios would take to respond to large queries while running a large model at high quant is quite high in comparison to this theoretical Mac Pro box.
Time is money.
Upvote
1
(1
/
0)
These are both the same reason: the only reason your hypothetical Mac Pro would be faster than a cluster of Mac Studios is because of the interconnect latency.
How much does interconnect latency affect the performance of these models?
LLM data centers use clusters. How does their interconnect latency compare to TB5?
Upvote
0
(0
/
0)
Except for the times when they reached a limit in the particular metric, and changed the definition of the "Law" to try and keep it relevant.,
Upvote
-3
(0
/
-3)
No, the speed part comes from the tensor hardware that the GPUs have (and that new M5 hardware will also have but the Studio currently does not)
TB5 does 80gbps, PCIe 5 does 120 bidirectional and current gen NVlink is 1.8tbps.
Older Nvlink is probably the fairest comparison because of cost but even then, that did 300gbps.
Upvote
0
(0
/
0)
So you're hypothesizing a world where the Mac Pro gets an M5 refresh before the Mac Studio?
No.
Upvote
-1
(0
/
-1)
I don't believe Moore ever did that. He observed, back in 1965, that the number of transistors in an integrated circuit was doubling every year; in 1975 he predicted that that trend would continue until about 1980, at which point he expected the rate of doubling to drop to every two years.
Others took that prediction and used it as a guideline for planning and semiconductor R&D, which is how we ended up with the prediction being shown to be "right". Moore himself has stated that the projection can't continue forever.
Regardless of all this, I stand by my statement: "Moore's law" was never a law in any meaningful sense. It was always a projection that came to be viewed as a law because it was so close to reality - but reality was itself shaped by that projection. There's nothing inherent in science that makes it a law in the ways that, for example, the laws of thermodynamics are.
Upvote
7
(7
/
0)
Yeah, Moore’s law was a complete bamboozlement from beginning to end — and I say this as someone who’s spent most of his career in the semiconductor industry. We all deluded ourselves into believing in it, we all were terrified that someone else would beat us to the punch, and so… we all made it real. The improvements have frankly continued far longer than I ever believed possible.
Upvote
2
(2
/
0)
I'm waiting for them to bring back user-replaceable SSDs, RAM, and batteries. And to fix that f@%!ng notch.
...we both may be waiting a while.
Upvote
-6
(1
/
-7)
Upvote
4
(4
/
0)
To be fair, the ADB era was the last time that Apple ever made a decent mouse, so maybe they should bring that back.
Upvote
-1
(3
/
-4)
Still waiting for them to let me connect my Vampire taps for my ThickNet network...
Upvote
0
(1
/
-1)
Other than having to endlessly clean lint off the ball and rollers, the ADB mouse was excellent. (A problem with all mice of the era, except for the laser mice that required those mirrored mouse pads.)
It is weird how all Apple mice since then have been ergonomic disasters. Especially considering the care they put into their keyboards.
Upvote
1
(2
/
-1)
Agreed.
Wait what? Apple keyboards have also been shitty un-adjustable crappily-switched fragile disasters forever. I think the last Apple keyboard with anything even resembling decent ergonomics was the CrumbCatcher 5000* that came with the old white polycarbonate iMac.
*This is probably not its official Apple name.
Upvote
-1
(0
/
-1)
There is a chance that an exploit working only on very old hardware won’t get exploited anymore. Too few victims. And especially too few wealthy victims.
Upvote
0
(0
/
0)
I don’t think that effect is so strong for Apple. For example, I have a very nice eight core M1. And it will be replaced with a refurbished M6. Thats the plan, and apple can preannounce what they like, that is what I will buy. It’s only for first time buyers.
And if your Mac breaks today, you need a new one today. So you buy what’s available today.
Upvote
0
(2
/
-2)
Well, aside from the butterfly switch durability/reliability debacle, it has been understood for the last ~20+ years that Apple makes the best laptop keyboards, in terms of how it feels to type on them. That's always a high point that's highlighted in comparison reviews (well, keyboard + touchpad, but that includes keyboard).
(And even with the butterfly switches, they were making an effort to even out the actuation force across each key even more than it already was.)
And their desktop keyboards have been similar to their laptop keyboards for a long time, so if you like that kind of keyboard, it's quite nice to type on.
https://www.rtings.com/keyboard/reviews/apple/magic-keyboard-with-touch-id-and-numeric-keypad
From Rtings: "The Apple Magic Keyboard with Touch ID is very good for office use. Typing feels great because the keys are light and offer good tactile feedback, and you shouldn't feel fatigued during long typing sessions, thanks to the low profile."
Rtings also has some choice comments re: how s**tty their mouse is.
Upvote
0
(0
/
0)
OK, so Apple wins the “tastiest turd” competition.
I’ve got a Magic Keyboard. It’s trash.
I’ve got a Magic Keyboard. It’s trash.
Upvote
0
(0
/
0)
Thanks for this link, I had read it before when the introduced PCC. They refer to custom-built hardware in it but no reference to custom chips so it may as well be only some custome motherboard and support chips paired with regular M-processor, and not higher-end ones. If you have another source to back the existence of this higer-end chips I would be happy to check it out.
Upvote
0
(0
/
0)
Never! That would mean be locked to a desk, and not being able to work at the beach before my surf sessions
(Or work on the bed once I come back destroyed by the waves)
Upvote
0
(0
/
0)
Try building a react native project. That thing is just stupid, but it’s what some of us have to deal with..
For example, I recently discovered that their new arch has a script that regenerates a bunch of C sources every rebuild, which completely breaks Xcode’s incremental builds. I was able to avoid that by patching RN code manually, but I’m still annoyed by that.
and don’t even get me started on Android Studio and it’s emulators
Upvote
1
(1
/
0)