Apple Announces New Line of MacBooks and Mac Minis Powered by M1

[dragontamer5788]

That's an extremely fat decoder, and a very fat backend. I'm curious how this will perform. This is the 2nd fattest chip I've ever seen, after Power9's SMT8 cores (12-wide decoder on Power9 SMT8).

EDIT: Honestly, this thing is almost certainly the king of instructions-per-clock. Nothing else on the market comes close to how wide that frontend is. I feel pretty confident making that prediction even before benchmarks come out.

Its still clocked pretty low though.

 

[SL2]

How can it be OSx when OSx is x86 and ARM is not x86

Doesn't matter, it's the same with Windows, Linux, Chromium, etc, they work with more than one platform. Application software support got nothing to do with it.

 

[biffzinker]

How can it be OSx when OSx is x86 and ARM is not x86, they can call it OSx and yea I guess it now is and will be going forward but it is in essence iOS and it can't run native x86 applications like macs have done for the last 15 years, so there's a whole plethora of applications that will need to be rewritten or run in x86 emulation

It’s simple really, iPhone/iPad both are running the same OS. Both devices are running with a stripped down version of MacOS X. Hard to say that still applies with the current iOS/iPadOS 14.

 

[Vya Domus]

They are on the same package. Did we watch the same presentation?

Package != chip. Are we using the same terminology established for the past couple of decades ?

Is this is a single chip ?

 

[silentbogo]

since 2009?

So, you are ignoring 8 years it existed before that?

 

[Dredi]

RAM isn't integrated

HBM, so not integrated to the chip, but integrated to the package. At least according to the finnish product page. First mass market CPU product with ram speeds in the hundreds of GBs per second.

My guess is one or two 8GB HBM2e stacks next to the M1 chip depending on the model.

 

[Darksaber]

So the M1 maxes out at 16GB memory and has no eGPU support and no Bootcamp support - the latter two were expected.

It is clear that Apple offers the M1 to the entry-level segment of their products. Mac Mini, MB Air and entry-level, 2-port MBP 13. The simple fact, that the on-die DRAM maxes out at 16 means that there needs to be another SOC chip to allow for more. On top of that, this is shared memory - while discreet GPU based systems don't have that shortcoming.

I am guessing that there will be an even faster version - more CPU cores, more GPU cores, more memory - an M2 - that goes into the higher-end Mac/Mac Pro/MB Pro 13s and the 16-inch units in 2021. I dare say, that we may also see a 13-inch MBP phased out at that point, replaced by 14-inch versions. Currently, the Macbook Air and MBP are way too similar with the exception of the display quality, so there needs to be further differentiation. Also, even if they did not talk about it, the Air is passive while the MBP is not - so are they clocked at different speeds to keep within power envelopes?

Financially it would also make sense. With a single silicone SKU, they no longer have the ability to differentiate different price points within a product stack. So if they want to be able to charge 2000+ USD for a Macbook Pro, they NEED to have a faster and thus more expensive SKU. In the end, that means more money in their pocket as they don't have to pay Intel anything anymore too.

I will buy the last-gen Intel MBP13 once they phase them out completely, or go for a brand spanking new high-end M2 based one at that point.

Just my 2 cents.

 

[Steevo]

I want my I pad to have a keyboard?

 

[Dredi]

I want my I pad to have a keyboard?

More like ”I want my pc to be faster and have a better battery life without sacrificing size, weight or noise level.”

 

[Initialised]

How can it be OSx when OSx is x86 and ARM is not x86, they can call it OSx and yea I guess it now is and will be going forward but it is in essence iOS and it can't run native x86 applications like macs have done for the last 15 years, so there's a whole plethora of applications that will need to be rewritten or run in x86 emulation

It's not OS X, it's MacOS 11 - did you miss the Spinal Tap reference in the June announcement?

 

[xman2007]

It’s simple really, iPhone/iPad both are running the same OS. Both devices are running with a stripped down version of MacOS X. Hard to say that still applies with the current iOS/iPadOS 14.

Of course it will, maybe with a new shiny front end GUI, but in essence it will be iOS as in ARM based apps and not x86, good luck thinking you can just buy a shiny new ARM based mac and run all your older mac apps on it, it won't, they will be ARM based apps which are only available on mobile until now

 

[RandallFlagg]

"ARM" and "x86" are mostly irrelevant to the developer. Compilers mostly take care of that now, high level languages like Objective -C / Swift / Java / C# really only expose underlying hardware via libraries / classes.

This transition will be a *lot* easier than PowerPC to x86 for developers. Development tools were much more primitive 10+ years ago.

 

[bulldozer]

Find it interesting that it seems like in most cases the Apple silicone versions of devices are replacing the lower end models if their claims are so spot on.

Also Mac mini for instance went from 4 usb ports to 2 usb ports. So for some of the cheaper production uses those are lost displays. Yes I actually have workflow to use more than a few displays through Presentation software. New Mac Mini is now useless for that use for me. Also Macbook Pro 13 again only two ports. So it is like it is replacing the 8th gen Intel version. Not the newer 10th gen. Seems like the IO might have some limitations. Still curious how they will ever compete with an actual Mac Pro tower. Compete with the server software. Server ECC memory etc. Not only that but the actual compute power of GPU's from Nvidia or AMD on the higher end.

Also interesting as it seems eventually SKU's are going to have to be reduced to just having options for more memory or more ssd space. Seems they are simplifying things, but think it might pull them out of the more professional spaces long term and more into the consumer prosumer spaces.

 

[RandallFlagg]

Find it interesting that it seems like in most cases the Apple silicone versions of devices are replacing the lower end models if their claims are so spot on.

Also Mac mini for instance went from 4 usb ports to 2 usb ports. So for some of the cheaper production uses those are lost displays. Yes I actually have workflow to use more than a few displays through Presentation software. New Mac Mini is now useless for that use for me. Also Macbook Pro 13 again only two ports. So it is like it is replacing the 8th gen Intel version. Not the newer 10th gen. Seems like the IO might have some limitations. Still curious how they will ever compete with an actual Mac Pro tower. Compete with the server software. Server ECC memory etc. Not only that but the actual compute power of GPU's from Nvidia or AMD on the higher end.

Also interesting as it seems eventually SKU's are going to have to be reduced to just having options for more memory or more ssd space. Seems they are simplifying things, but think it might pull them out of the more professional spaces long term and more into the consumer prosumer spaces.

I think they are trying to make a market to encourage developers to migrate, for one. Honestly if the M1 lives up to or even near it's claims, you can have a macbook air that performs like a 10900 in compute with graphics similar to that of a 1050Ti for $999. Expansion isn't a big deal with Thunderbolt 4, except the lack of eGPU.

If it lives up to the hype, by this time 2021 they should be in a good position to move up the line. Comparisons to difficulty migrating from PPC to x86 are really very weak, I was into Macs at the time and it was pretty fluid for one, but also the tools are 10X better now. In fact, I suspect that many applications can be compiled and migrated in a matter of days if not hours - especially if it is written in Swift, which is a lot closer to java / C# than Objective-C. Most of the time would likely be spent on testing and validation.

I think the transition will be far faster than the PPC->x86. If it lives up to the hype, that is.

 

[biffzinker]

Of course it will, maybe with a new shiny front end GUI, but in essence it will be iOS as in ARM based apps and not x86, good luck thinking you can just buy a shiny new ARM based mac and run all your older mac apps on it, it won't, they will be ARM based apps which are only available on mobile until now

Apple’s got that covered with Rosetta 2. The only limitation for the JIT is VM software such as Parallels, and VMware, and Kernel extensions.

Rosetta 2 will let x86_64 apps run on Macs with ARM-based M1 chips - Liliputing

Rosetta 2 will let x86_64 apps run on Macs with ARM-based M1 chips

liliputing.com

 

[bulldozer]

I think they are trying to make a market to encourage developers to migrate, for one. Honestly if the M1 lives up to or even near it's claims, you can have a macbook air that performs like a 10900 in compute with graphics similar to that of a 1050Ti for $999. Expansion isn't a big deal with Thunderbolt 4, except the lack of eGPU.

If it lives up to the hype, by this time 2021 they should be in a good position to move up the line. Comparisons to difficulty migrating from PPC to x86 are really very weak, I was into Macs at the time and it was pretty fluid for one, but also the tools are 10X better now. In fact, I suspect that many applications can be compiled and migrated in a matter of days if not hours - especially if it is written in Swift, which is a lot closer to java / C# than Objective-C. Most of the time would likely be spent on testing and validation.

I think the transition will be far faster than the PPC->x86. If it lives up to the hype, that is.

Have a hard time seeing that. If a Macbook Air was at that level then they are going to have a hard time selling higher end devices. MAC OS does have a limitation with some things unless they fixed it with Big Sur. Each output can only use a single display. This isn't the case with Windows. Between that and a hardware switch I am guessing offering a more extensive IO might have caused some issues. Apple doesn't take loses. I have a hard time imagining that they are putting in all this R&D and making something that is going to be able to eventually compete at the Enterprise level yet they are keeping prices the same and even in the case of the Mac Mini reducing the price $100. On the Enterprise level or you even see it in emulation honestly is the Intel Instruction sets at times being used to a large degree for certain workflows. Curious how those work out for Apple. The presentation made reference to things like photoshop multiple times but then they said it won't be native until 2021. No mention of Adobe AE or Premiere. On the higher end of things there are some softwares that are so expensive people just rent out workstation time instead of even purchasing the software. I can't imagine that Apple is going to get into these places.

They also based on their ridiculous non information graphs made reference multiple times to 10w laptops. Which basically is what is in ultrabooks. Not what is found in higher end laptops with dedicated GPU's. I think the gpu numbers at best are making reference to macbook pro intel 645. in context a 1050ti is about 400 percent faster in benchmarks compared to a 645. The 1050ti is over 4 years old and was a low end part when it came out from nvidia.

I think for what Apple is doing it will work. I just don't see them competing long term in the high end. Maybe AMD will stall out but they increased in IPC 20 percent this last gen leap. Apple says they increased 16 percent from a13 to a14.

While Apple might bring some costs down by doing it in house they are now fronting the cost of r&d and fabrication for their chips instead of just purchasing them. They are also having to maintain somewhat of a relationship with other companies including Intel for thunderbolt and having controllers for things like that. Arm seems to be moving pretty swiftly lately but just like all tech I am sure they will hit a wall at some point. Either Arm will front the cost to develop some of that as they grow or Apple will to work around it. AMD had to redesign from the ground up and that is what cost money. Still curious if ARM licensing will be effected long term by the Nvidia purchase which is still an unknown. Regardless if Apple is making their own chips or not. They are still Arm.

It doesn't seem like it took Apple 2 or more years to roll out to Intel from PPC. Just seems kind of long and drawn out and at the point risky as well. Again just not sure I see a world that what they are doing translates into the server and enterprise areas. They will bring gaming in the form of mobile games to their devices but lose things like Steam and real PC gaming it seems. And with all of this transition it seems rather short sighted to not include LTE for mobile or touchscreens to the devices as you integrate mobile apps. Something Windows laptops and tablets have been doing for years in both Intel and Arm based products.

 

[mechtech]

So is this based off of an ARM cpu or something completely new and unique?

I'm wondering if Nvidia has or when they have cleared (all done?) legal hurdles and own ARM, if they are going to get a Windows optimized version for ARM so they can start competing against Intel and AMD?

 

[Zareek]

$699 for the Mac Mini based on M1. The profit margins on these must be absolute insanity!!! I figured with the ARM chips they would shift to a more reasonable pricing structure passing the savings onto their customers. That wouldn't be very Apple of them now would it. They are going to continue to pretend they are selling the absolute best hardware for reasonable prices. I will believe in their performance claims when they can show the M1 running the same software accomplishing the same task as a top end x86 machine, mobile doesn't count. SPECint and Geekbench mean nothing in the real world.

 

[RandallFlagg]

So is this based off of an ARM cpu or something completely new and unique?

I'm wondering if Nvidia has or when they have cleared (all done?) legal hurdles and own ARM, if they are going to get a Windows optimized version for ARM so they can start competing against Intel and AMD?

Kind of, Apple was one of the founding partners of ARM. Originally it was Acorn, Apple, someone else I don't recall. They got involved when they were doing their first PDA, the Newton. By virtue of being a founding member, Apple has rights to use ARM designs - there are some restrictions I believe but that is the gist of it.

That said, Apple deviated from standard ARM architecture on the iPhone 4S if I recall correctly, and at this point it really isn't ARM anymore even though it uses ARM microcode instruction sets.

$699 for the Mac Mini based on M1. The profit margins on these must be absolute insanity!!! I figured with the ARM chips they would shift to a more reasonable pricing structure passing the savings onto their customers. That wouldn't be very Apple of them now would it. They are going to continue to pretend they are selling the absolute best hardware for reasonable prices. I will believe in their performance claims when they can show the M1 running the same software accomplishing the same task as a top end x86 machine, mobile doesn't count. SPECint and Geekbench mean nothing in the real world.

Actually I suspect Apple is going to lose money initially on these, especially if you consider the R&D invested. This first year is probably the bait, holiday 2021 and beyond will be the hook and I bet prices go up at that point. Sorta like the original iPhone cost $499.

 

[dicobalt]

I didn't know Trump wrote Apple's promotional material.

 

[TheLostSwede]

HBM, so not integrated to the chip, but integrated to the package. At least according to the finnish product page. First mass market CPU product with ram speeds in the hundreds of GBs per second.

My guess is one or two 8GB HBM2e stacks next to the M1 chip depending on the model.

It's not HBM, where do people get this from? It's custom packaged LPDDR4x.

So the M1 maxes out at 16GB memory and has no eGPU support and no Bootcamp support - the latter two were expected.

It is clear that Apple offers the M1 to the entry-level segment of their products. Mac Mini, MB Air and entry-level, 2-port MBP 13. The simple fact, that the on-die DRAM maxes out at 16 means that there needs to be another SOC chip to allow for more. On top of that, this is shared memory - while discreet GPU based systems don't have that shortcoming.

I am guessing that there will be an even faster version - more CPU cores, more GPU cores, more memory - an M2 - that goes into the higher-end Mac/Mac Pro/MB Pro 13s and the 16-inch units in 2021. I dare say, that we may also see a 13-inch MBP phased out at that point, replaced by 14-inch versions. Currently, the Macbook Air and MBP are way too similar with the exception of the display quality, so there needs to be further differentiation. Also, even if they did not talk about it, the Air is passive while the MBP is not - so are they clocked at different speeds to keep within power envelopes?

Financially it would also make sense. With a single silicone SKU, they no longer have the ability to differentiate different price points within a product stack. So if they want to be able to charge 2000+ USD for a Macbook Pro, they NEED to have a faster and thus more expensive SKU. In the end, that means more money in their pocket as they don't have to pay Intel anything anymore too.

I will buy the last-gen Intel MBP13 once they phase them out completely, or go for a brand spanking new high-end M2 based one at that point.

Just my 2 cents.

But they did differentiate the SKUs, in a sneaky way. The bar model Air has 7 GPU cores, whereas all other devices have 8.

 

[Dredi]

It's not HBM, where do people get this from? It's custom packaged LPDDR4x.

From apple.

Translation of the relevant sentences is ”M1 chip enables up to 16 gigabytes of super fast unified memory. Low latency HBM-memory is integrated to a single component, which allows apps to share data efficiently between the processor, the GPU and the neural engine.”.

It is possible that the translated websites were not properly QA:d and that the memory is not HBM.

 

[Vya Domus]

It is possible that the translated websites were not properly QA:d and that the memory is not HBM.

That's 100% percent the case because on the US version of the site it just says "high‑bandwidth, low‑latency memory".

They showed a picture with the board showing the DRAM, so for the love of God can we stop with this pointless debate ?

 

[TheLostSwede]

From apple.
View attachment 175212
Translation of the relevant sentences is ”M1 chip enables up to 16 gigabytes of super fast unified memory. Low latency HBM-memory is integrated to a single component, which allows apps to share data efficiently between the processor, the GPU and the neural engine.”.

It is possible that the translated websites were not properly QA:d and that the memory is not HBM.

Please find that text on the English site, as I can't. Apple's reality distortion filter doesn't help either, as they clearly don't want to post any kinds of real spec.

Mac

The most powerful Mac laptops and desktops ever. Supercharged by Apple silicon. MacBook Air, MacBook Pro, iMac, Mac mini, Mac Studio, and Mac Pro.

www.apple.com

 

[Dredi]

Please find that text on the English site, as I can't. Apple's reality distortion filter doesn't help either, as they clearly don't want to post any kinds of real spec.

Mac

The most powerful Mac laptops and desktops ever. Supercharged by Apple silicon. MacBook Air, MacBook Pro, iMac, Mac mini, Mac Studio, and Mac Pro.

www.apple.com

Anandtech seems to be 100% sure that this is just a (very weird) translation error.