• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Apple Announces Next-Generation M1 Pro and M1 Max Chips

OP
sweetchaos

sweetchaos

Major Contributor
The Curator
Joined
Nov 29, 2019
Messages
3,872
Likes
11,549
Location
BC, Canada
I can't help feeling like the pro or max would just be overkill
No such thing. ;)

Plus, I suspect Apple is working on faster chips for Mac Pro’s (rumoured for 2022 release), for Xtreme(TM) performance.

Macrumors write:
“The other Mac Pro models will use high-end Apple silicon chip options with 20 or 40 computing cores, made up of 6 high-performance or 32 high-performance cores and four or eight high-efficiency cores. These upgraded chips are also expected to include 64 or 128 core GPUs, and at the top of the line, the graphics chips would be several times faster than the graphics modules Apple uses from Nvidia and AMD.”
 

elvisizer

Active Member
Joined
Sep 16, 2018
Messages
257
Likes
197
Plus, I suspect Apple is working on faster chips for Mac Pro’s (rumoured for 2022 release), for Xtreme(TM) performance.
guaranteed they are. also pretty much a given that we'll have an m2 next year, then m2max equivalent after that . . . . apple controls the release cycle now so it'll be like the iphones- new chips every year!
 

jhaider

Major Contributor
Forum Donor
Joined
Jun 5, 2016
Messages
2,822
Likes
4,514
So far I think the lede has been omitted or buried - Apple reverses unforced error, finally revives MagSafe on current MBPs. Everything else is just window dressing.

To put a finer point on it, no MagSafe is why I haven't considered replacing my 2015 Retina MacBook Pro until now. I'd have ordered the new one Monday if my logic board hadn't fried in August and cost $500 to repair. I chose repair over "upgrade" solely because MagSafe. But when Apple stocks refurbs with enough RAM and storage of this one, I'll upgrade. No issues arose from upgrading our media server from a 2014 Mac mini to an M1 Mac mini. Don't care about virtualization - when I'm required to use third tier platforms (i.e. to run audio software such as VituixCAD, London Architect, Audio Architect, etc.) I bought a separate machine for that to conserve storage space on the MBP for music!
 
OP
sweetchaos

sweetchaos

Major Contributor
The Curator
Joined
Nov 29, 2019
Messages
3,872
Likes
11,549
Location
BC, Canada
There's a difference between models with this feature that Apple didn't really mention.

While both models can be charged over Thunderbolt or MagSafe, only the 14-inch MacBook Pro can fast charge over the USB-C/Thunderbolt 4 ports as well as ‌MagSafe‌.

By contrast, the 16-inch MacBook Pro is limited to fast charging over the ‌MagSafe‌ port.
Source
 

elvisizer

Active Member
Joined
Sep 16, 2018
Messages
257
Likes
197
wow that's a bizarre one! edit: oh wait actually it makes sense- usb-c PD tops out at 100 watts, the 16" needs 140 watts to fast charge.
 

blueone

Major Contributor
Forum Donor
Joined
May 11, 2019
Messages
1,179
Likes
1,494
Location
USA
It just occurred to me while reading another post, I misread yours. It is my understanding that the DRAM in the M1 variants are off the shelf DRAM chips that are in the SoC package, but are separate from the SoC. This has two big advantages for Apple, both of which accrue to users, but one directly and one indirectly. The first is that an in-package interconnect can be a lot faster than DDR4, as DDR3/4/5 is intended for trace, and is half duplex. I don't know for sure, but I suspect the M1's DRAM interface is full duplex. The better in-package DRAM interconnect probably gives the M1 a significant performance advantage over CPUs with external DIMMs, but I am just guessing.

The second advantage is that Apple can buy DRAM chips directly from the DRAM foundries (e.g. Samsung, Micron, etc.) and not have to pay the mark-up the DIMM manufacturers charge, or the added cost of the DIMM boards, connectors, and added motherboard layers for the DDR traces.

Ha! I was wrong. The M1 is using the LPDDR5 interface at 6400 MT (mega transfers) per second. I gave Apple too much credit, but the M1 processors are still pretty nifty.

 
OP
sweetchaos

sweetchaos

Major Contributor
The Curator
Joined
Nov 29, 2019
Messages
3,872
Likes
11,549
Location
BC, Canada
While it’s possible to spec-out a 14-inch MacBook Pro with the M1 Max chipset, MacRumors is reporting that only the 16-inch MacBook Pro with M1 Max can take advantage of a “High Power Mode” to be available in macOS Monterey.

So unless you must have the smaller form factor of the 14" model and/or you don’t care about performance (why are you getting the M1 Max?!), it would be foolish to spec-out an M1 Max configuration without going to the 16" model.

There is no meaningful price difference between the two sizes (about US$200 for a high-spec model), so it is almost entirely about choosing form factor.
 

ScofieldKid

Member
Joined
Feb 23, 2019
Messages
86
Likes
24
Location
Northwest, USA
So, the move from M1 to M1 Pro gets you... 1) double the die size (transistor count), 2) go from 4x4 to 6x8 on performance-cores to efficiency cores, 3) double the GPU cores, and 4) move from 64-bit to 128-bit memory and 5) up the memory clock from 2133 to 3200. So yeah, that probably gets you going.

Only other big note I would make, if you are tied to a bunch of licenses that are stuck at 32-bit (Big Sur), then you have to factor that in. It looks like only the older one (the M1 Macbook Air, for example) is supported as a Big Sur box.
 

TunaBug

Member
Forum Donor
Joined
Aug 19, 2021
Messages
81
Likes
100
Location
Seattle-ish
We have function keys and magsafe back, but I'm still waiting on user replaceable RAM & SSD.

I often wondered if so much of Apple's form wayyyyyyy before function of the past decade had to do with Jobs not being around to constrain Ive. Function seems to have been making a comeback since Ive departed.
 

blueone

Major Contributor
Forum Donor
Joined
May 11, 2019
Messages
1,179
Likes
1,494
Location
USA
So, the move from M1 to M1 Pro gets you... 1) double the die size (transistor count), 2) go from 4x4 to 6x8 on performance-cores to efficiency cores, 3) double the GPU cores, and 4) move from 64-bit to 128-bit memory and 5) up the memory clock from 2133 to 3200. So yeah, that probably gets you going.
The Anandtech article I referenced in a prior post says the M1 Pro and Max use LPDDR5-6400. Looking at a top of the line client CPU from Intel:

Intel® Core™ i9-11980HK Processor (24M Cache, up to 5.00 GHz)

The spec on the Intel 65 watt thing says its maximum memory bandwidth is 51.2GB/sec. The M1 Pro is 4x that, and the M1 Max is at 8x. The AMD Ryzen 5980HX is a little better, at about 68.2GB/sec (using LPDDR4-4266 DIMMs), but still wimpy compared to the M1 Pro or Max.

AMD Ryzen 9 5980HX: performance and specs | NanoReview

Apple's M1 looks very impressive by comparison on this important metric for lowering latency.
 

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
It just occurred to me while reading another post, I misread yours. It is my understanding that the DRAM in the M1 variants are off the shelf DRAM chips that are in the SoC package, but are separate from the SoC. This has two big advantages for Apple, both of which accrue to users, but one directly and one indirectly. The first is that an in-package interconnect can be a lot faster than DDR4, as DDR3/4/5 is intended for trace, and is half duplex. I don't know for sure, but I suspect the M1's DRAM interface is full duplex. The better in-package DRAM interconnect probably gives the M1 a significant performance advantage over CPUs with external DIMMs, but I am just guessing.

The second advantage is that Apple can buy DRAM chips directly from the DRAM foundries (e.g. Samsung, Micron, etc.) and not have to pay the mark-up the DIMM manufacturers charge, or the added cost of the DIMM boards, connectors, and added motherboard layers for the DDR traces.
There’s only one chip on one silicon chip. Apple doesn’t buy any silicon from anyone else.
 

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
We have function keys and magsafe back, but I'm still waiting on user replaceable RAM & SSD.
It will never happen for the foreseeable future. Apple will use a SoC. that’s their USP and the reason of the high performance.
 

blueone

Major Contributor
Forum Donor
Joined
May 11, 2019
Messages
1,179
Likes
1,494
Location
USA
There’s only one chip on one silicon chip. Apple doesn’t buy any silicon from anyone else.

Not correct. Apple is using LPDDR5-6400 memory chips in the SoC multi-chip package for the Pro and the Max.
 

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK

elvisizer

Active Member
Joined
Sep 16, 2018
Messages
257
Likes
197
So, the move from M1 to M1 Pro gets you... 1) double the die size (transistor count), 2) go from 4x4 to 6x8 on performance-cores to efficiency cores, 3) double the GPU cores, and 4) move from 64-bit to 128-bit memory and 5) up the memory clock from 2133 to 3200. So yeah, that probably gets you going.

Only other big note I would make, if you are tied to a bunch of licenses that are stuck at 32-bit (Big Sur), then you have to factor that in. It looks like only the older one (the M1 Macbook Air, for example) is supported as a Big Sur box.
Big Sur already doesn't support 32 bit apps, they were dropped all the way back on catalina already

Starting with macOS Catalina, 32-bit apps are no longer compatible with macOS.
 

blueone

Major Contributor
Forum Donor
Joined
May 11, 2019
Messages
1,179
Likes
1,494
Location
USA
Every circuit is deposited on just one silicon wafer. That’s what SoC means. System on [one] chip. It’s not a multi-chip package.


Oh really? The M1 Max SoC has about 57 billion transistors in it. It supports configurations up to 64GB of DRAM. So, tell us, how does the math work with the DRAM on the SoC?

Edit: Just for nit-picking terminology correctness: An individual chip that goes into a package is called a "die". A wafer refers to the circular silicon discs (usually 300mm or 400mm in diameter) that the individual dies are etched into by lithography. The dies are then cut from the wafers, and are mounted in packages. If multiple chips are put into a package it used to be called just a multi-chip package, but now so many dies are sometimes put into a single package that the dies are called "chiplets" (AMD's terminology) or the overall strategy is called "tiled dies" in a package (Intel terminology). Usually chiplets or tiled-dies refer to functionally distributing the circuitry from one SoC to multiple dies, either to take advantage of different fabrication processes for different functional blocks, or to make the dies smaller so it is easier for the fabs to achieve higher production yields (a larger number of fully functional chips from a wafer). The chiplet and tiled die strategies also include intra-package interconnect standards for inter-die communication, like Intel's EMIB (https://www.intel.com/content/www/us/en/silicon-innovations/6-pillars/emib.html).
 
Last edited:

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,494
It really surprises me how Apple can be so ahead of the curve sometimes but still have jackshit completely bonkers practices regarding repair-ability.

Sure 5nm transistors wold put them on top of the game, but my device becoming e-waste once a connector becomes loose inside is not worth it.

What's exactly bonkers about restricting repairs in such a manner that they reap benefits of by having a portion of that after-sales pie? They also don't care about e-waste (you could never be a company of their caliber and legitimately care about something of this nature from a true sense - you'd have to be a total moron otherwise, but if you actually were a moron, there's no way your company could be this big in the first place). They have a goal of carbon neutral by 2030 (whatever that even means considering there is no international certification nor standard agreed upon for such a notion). Which is fine, but their stupid little stunts of not providing chargers, earphones, while selling things like wireless charger and things of that nature don't make logical sense for various reasons. In the same way their claims about piracy are constantly being contradicted with their behavior in other territories (and most famously being a part of the PRISM as Snowden & Friends have shown).

Anything they're "ahead" in, is squarely in line with the well established fiduciary responsibility to shareholders. And that's making money. If they can get bonus points beside that goal for PR's sake along the way for anything, they will. And that's simply because they got the "making money" part down enough.
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,423
Likes
7,940
Location
Brussels, Belgium
What's exactly bonkers about restricting repairs in such a manner that they reap benefits of by having a portion of that after-sales pie?

when was the last time something you have from Apple broke?

My iPhone was stuck in an apple logo reboot loop. there is only one apple store in my entire country so it took two weeks to get an appointment with the 'Genius bar'.

The guy over there determined without even touching the phone that it was a software problem, and that he has to wipe out my data and flash the phone. I declined, he said i should search Google for 'not approved by apple' methods to get my data back.

Guess what? it was a loose battery connector.

if they can repair the device they made i don't care how tall they build these walls around their garden, but when the 'Genius' bar consists of morons who are not authorized to fix the phone then you lost all my respect points. We're all entitled to our own opinions.
 

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
Oh really? The M1 Max SoC has about 57 billion transistors in it. It supports configurations up to 64GB of DRAM. So, tell us, how does the math work with the DRAM on the SoC?

Edit: Just for nit-picking terminology correctness: An individual chip that goes into a package is called a "die". A wafer refers to the circular silicon discs (usually 300mm or 400mm in diameter) that the individual dies are etched into by lithography. The dies are then cut from the wafers, and are mounted in packages. If multiple chips are put into a package it used to be called just a multi-chip package, but now so many dies are sometimes put into a single package that the dies are called "chiplets" (AMD's terminology) or the overall strategy is called "tiled dies" in a package (Intel terminology). Usually chiplets or tiled-dies refer to functionally distributing the circuitry from one SoC to multiple dies, either to take advantage of different fabrication processes for different functional blocks, or to make the dies smaller so it is easier for the fabs to achieve higher production yields (a larger number of fully functional chips from a wafer). The chiplet and tiled die strategies also include intra-package interconnect standards for inter-die communication, like Intel's EMIB (https://www.intel.com/content/www/us/en/silicon-innovations/6-pillars/emib.html).
I stand corrected. I learned something today. Thank you.
 
Top Bottom