Apple has announced what should be some exceedingly impressive chips today to follow the M1.
They seem to be both based around the same cores used in the M1, but:
- The M1 Pro has 8 performance, 2 efficiency cores, a 16 core GPU, and up to 32GB RAM.
- The M1 Max has 8 performance, 2 efficiency cores, a 32 core GPU, and up to 64GB of RAM - with double the bandwidth of the other versions, claimed to be 400GB/s. That’s slightly insane.
I’m… looking forward to benchmarks, though at this point, I’ll just be watching from the sidelines.
If only someone made an ARM SBC that didn’t entirely suck?
A bit of a more technical dive:
One thing I missed is that the Pro has twice the already insane memory bandwidth of the M1. The Max has 4x.
Let’s reflect on MBW numbers on the M1 compared to some other systems…
Those aren’t top of the line, but neither are they slouches…
sniffs My PineBook Pro only manages about 5000 on the MCBLOCK test. So does my ODroid N2+. The Pi4 comes in around 2500. A linear scaling would put the M1 Max at around 160,000. Bonkers if it can do it! I expect a lot of that bandwidth will get consumed by the GPU, but, still.
And remember, this is still the “Oh, yeah, we should do a laptop…” chip.
For those who missed it, laptops now have a notch, and have gone back to ports on the side. SD, HDMI, Headphone, and Magsafe (3… boo…). Plus some Thunderbolt 4 ports you can charge from too. They look legit.
Not that I’m going to be buying one. Ever, at this point.
Wait it can charge on both mag3 and thunder4? Interesting.
Work may get me one eventually.
That’s my impression of it, which… good! The two aren’t really the same, and IMO, the loss of Magsafe was a huge loss in terms of practical function of the laptop. TB charging is great if you’re desk-docked, but if, say… you’re in a living room, using it on your lap, with kids around? Magsafe means you won’t destroy the laptop when a kid trips over the cord. Having two small children, I can confirm that they trip over cords.
And while I recognize a few people are annoyed at the loss of the touchbar and the increase in weight, I think this is a very, very solid, post-Ive laptop, and exactly what Apple should be doing.
There’s only so much space to experiment in a form factor before you’ve covered the bounds, and when you’re constrained by “The keyboard needs to fit human hands,” and you’ve pretty much pushed the display out so far there are no meaningful bezels left… just iterate on it. And then put in the stuff that you lost while experimenting with winning industrial design awards. An SD card reader and HDMI port are useful.
Not that I will be getting one, pretty much ever at this point.
… though it is occasionally interesting to use my work iPhone and remember just what a smooth interface feels like. sigh Need to stop doing that.
The profile view definitely reminds me of my first PowerBook G4. And that’s not bad.
You’re right, it really does look like a TiBook! Now it just needs a flip down port protector with all the ports on the back!
This, I so want to replace my circa 2014 NUCs with something more powerful and more energy efficient. An M1 (|Pro|Max) would be perfect, but it sounds like running Linux on the Mac Mini is still sketchy (if nothing else, you need an external drive & network adapter).
Hopefully the rest of the world will catch up in making good ARM chips.
There are plenty of good ARM chips. In phones. I guess the licensing costs for SBC use are prohibitive or something. I don’t know the details, just that there’s very, very little out there.
I still have my M1 Mini, I’ve considered stuffing Linux on it to see how it is, and may do that at some point, but Linux on a 5k display, software rendered, sounds like a bad time of it.
Isn’t a fair bit of it, that the top tier chip manufacturers aren’t willing to open-source there driver/HAL layers to an extent that they can easily be fully utilised with linux? Especially in the GPU/NPU/TPU space.
I’m not sure it matters. The Linux driver interfaces, GPU interfaces, etc, are entirely unlike Apple’s interfaces.
I’m not sure that even a fully open sourced driver set for OS X would be of that much use for Linux, not that we’ll ever see that. Apple doesn’t appear to be hindering Linux dev on Apple Silicon, so much as they simply don’t care one way or another. It runs OS X, that’s the target, that’s all they care about.
Though I should probably try one of the Linuxes at one point on it. I’ve not gotten around to selling it and it’s not worth anything anymore.
M1 Max. GeekBench 5.
My house desktop, 8700K, is 1357/6959.
For comparison, the M1 Mini, which is blisteringly fast, is 1746/7711. So the Max is about the same, single threaded, with another 65% on top when fully loaded. And insane memory bandwidth that almost certainly doesn’t show up in Geekbench tests.
I wouldn’t mind swapping out for a M1 max mini.
The M1 Pro and M1 Max change the narrative completely – these designs feel like truly SoCs that have been made with power users in mind, with Apple increasing the performance metrics in all vectors. We expected large performance jumps, but we didn’t expect the some of the monstrous increases that the new chips are able to achieve.
On the CPU side, doubling up on the performance cores is an evident way to increase performance – the competition also does so with some of their designs. How Apple does it differently, is that it not only scaled the CPU cores, but everything surrounding them. It’s not just 4 additional performance cores, it’s a whole new performance cluster with its own L2. On the memory side, Apple has scaled its memory subsystem to never before seen dimensions, and this allows the M1 Pro & Max to achieve performance figures that simply weren’t even considered possible in a laptop chip. The chips here aren’t only able to outclass any competitor laptop design, but also competes against the best desktop systems out there, you’d have to bring out server-class hardware to get ahead of the M1 Max – it’s just generally absurd.
To get Anandtech to call something “absurd”…
Super duper deep dive into the architecture: https://drive.google.com/file/d/1WrMYCZMnhsGP4o3H33ioAUKL_bjuJSPt/view
This is like Fog-level analysis! loads and reads