Watch the M1 Apple Silicon MacBook Air Blaze Through Opening Every Default App

[DocNo]

I'm curious what is letting the ARM chip opens apps this fast. The SSD isn't much different than a 16" MBP, and the cpu is only a bit better at single threaded speeds.

Integration. The RAM is ON the SOC - shorter physical distance = speed. Wider data path. ARM can also have wider pipelines than x86 due to x86 architectural limitations (mainly from backwards compatibility). Tight optimization with Apples compliers since Apple also makes the hardware.

Compared to x86, Apple's ARM implementation is a "clean slate". Especially with their focus on 64 bit only.

Another way to put it - they get more work done in the same amount of time vs. x86 because they have a clean design not bogged down by 40 years of backwards compatibility baggage, and they also control the entire system now so they can leave out crap they don't want and optimize the heck out of small things that will have a huge performance impact in work loads their customers do, whereas getting commodity/least common denominator parts vendors like Intel to do such work is pretty hard when your only 10% of their customer base.

TANSTAAFL!

It's not just individual parts like CPU cores or SSDs, but the entire symphony of hardware and software working in concert. Your only as fast as your slowest link and all that...

 

[DocNo]

I think you'll be perfectly fine with 8 GB of RAM.

Forever limiting yourself for $200 seems pretty silly to me. I can get heavy into swap town just with Safari these days

 

[mike...]

You are completely wrong, “they knew pros would jump on them as soon as released and then get mad” how would you know this? Please elaborate
Also I guarantee you don’t need 32gb or even 16gb of memory.

Please, oh wise and wonderful oracle of memory, tell me how much I don't need

 

[djjeff]

Integration. The RAM is ON the SOC - shorter physical distance = speed. Wider data path. ARM can also have wider pipelines than x86 due to x86 architectural limitations (mainly from backwards compatibility). Tight optimization with Apples compliers since Apple also makes the hardware.

Compared to x86, Apple's ARM implementation is a "clean slate". Especially with their focus on 64 bit only.

Another way to put it - they get more work done in the same amount of time vs. x86 because they have a clean design not bogged down by 40 years of backwards compatibility baggage, and they also control the entire system now so they can leave out crap they don't want and optimize the heck out of small things that will have a huge performance impact in work loads their customers do, whereas getting commodity/least common denominator parts vendors like Intel to do such work is pretty hard when your only 10% of their customer base.

TANSTAAFL!

It's not just individual parts like CPU cores or SSDs, but the entire symphony of hardware and software working in concert. Your only as fast as your slowest link and all that...

I am in agreement with most of what you wrote but puzzled by the architecture limitations you say are inherent to x86 (shouldn't we really be discussing x64 as there are limits in x86 which are not applicable in x64). Can you elaborate on these architectural limitations?

I think one of the main benefits of the M1 is, as you stated, everything is on a single chip and tightly integrated. Performance can be increased in such a design (an example would be the performance improvements observed when the memory controller was moved onto the processor).

 

[djjeff]

Forever limiting yourself for $200 seems pretty silly to me. I can get heavy into swap town just with Safari these days

If one doesn't require more than 8GB then it's $200 wasted. Keep in mind that on the lowest cost MBA this $200 represents 20% of the system cost. Maybe wise to save the money for an updated system when they're released and the OP needs a new one.

 

[DocNo]

To everyone saying they want more than 16GB of RAM... bear in mind how efficient the M1 chip is. So it's not really a case of having more RAM, it's more a case of how it is managed and the M1 seems to be excellent at Memory management.

The OS does memory management and dictates memory requirements.

Big Sur is NOT iOS. iOS manages memory in a completely different manner. There are some efficiencies with ARM that might add up to some minor reductions in memory requirements, but I would expect Intel and AS Macs to have similar RAM requirements.

And more and more with Safari I'm routinely getting into swap with Big Sur on my 8GB 2015 MBA so yeah, even doing "boring" stuff can consume large amounts of RAM.

Since your machine is fixed and not updatable I would never buy a machine without at least 16GB of RAM, even if I had to wait to get the extra $200 saved up. Yes, the new SSDs are a lot faster but why wear them out faster with a lot of swapping? I think people tend to forget and underestimate the wear rates in SSDs. High write loads can destroy consumer SSDs quickly (always amused when people want to run SSDs in network video recorders for camera systems - hope you have stock in SSD drive manufacturing companies).

 

[v3rlon]

This is nonsense. There's only one chip because Apple have only switched their lowest-performing devices. You'll have more choice once they've fully transitioned away from Intel.



More uninformed nonsense. If you have a large amount of data to load in to RAM you need a large amount of RAM for it, otherwise the OS will start to use swap. macOS has not changed in this regard. Nothing about the M1 prevents it.



Yes of course they do. If you have any kind of professional workflow you'll be switching between multiple running apps all the time.



Maybe you should have watched the video before you commented? macOS has its own memory warning and that's what was shown in the video:

View attachment 1671965



This shows a fundamental misunderstanding of RAM. In my analogy, the RAM is the bucket. In your analogy, RAM is a fuel. That makes no sense.

Think of it another way, you have a 10L bucket full of flour. You're saying Apple has invented some way to let you bake 20 cakes instead of 10 after you fill up their magic bucket with flour. It's not possible. Sure you may be able to pack more flour in by compressing it a little (macOS has had memory compression since Mavericks) but you can't make 16GB of RAM be as effective as 32GB RAM. The improvements the M1 brings are mostly to do with speed.

If you have 10 litres of stuff to store, you need a 10 litre bucket. If you have 32GB of data to load in to RAM, you need 32GB of RAM. If you don't, the operating system will swap which is analogous to your stuff overflowing the bucket and spilling out all over the floor making it harder to manage. Due to the way computers work you can only use flour directly from the bucket. So before you can use the flour from the floor, you first have to empty the bucket a little bit then go around scooping up all the spilled flour. This is obviously much slower and less efficient than if you simply bought a bigger bucket (i.e. more RAM).

No, you misunderstand my picking on your awful analogy, but the part in the post before that should have been clear.
If MacOS M1 takes up less space in RAM than MacOS x86, there is more room left over in the bucket for other applications.

To abuse your next analogy, if you have a bag of flour big enough to bake 20 cakes, but for some reason need to use almost of that to dust the cake pan and other surfaces, plus some more that you just can't get out of the bag because it sticks to the sides of said bag, and some more lost as waste, you can only bake 10 cakes. If Apple comes along with some non-stick magic teflon so you no longer need to dust the cake pan and improves the bag design, you can get a lot closer to that theoretical limit of 20 cakes.
BUT, if you absolutely need to bake 25 cakes, you are going to need two bags of flour.

I never said it solves every problem. I CLEARLY stated that it didn't. I DID say that it was possible to improve the situation, just as MacOS does over Windows. For applications where you usually need more than 16GB, you might be able to get by with less (say you needed 18GB, now you are okay). Where you need a LOT more than 16GB (say 64GB), you still need more than 16GB.

 

[DocNo]

I am in agreement with most of what you wrote but puzzled by the architecture limitations you say are inherent to x86 (shouldn't we really be discussing x64 as there are limits in x86 which are not applicable in x64). Can you elaborate on these architectural limitations?

Anandtech has a pretty good first look at the M1 and towards the front they talk about a few of them. The biggest is the pipeline width; Intel and AMD are pretty much stuck at how wide they can make the pipeline due to the way several instructions work.

x86 is still relevant because there is no pure x64 only CPU out there. If you think people wailing about Apple ditching Intel is epic, just watch what would happen if Intel or AMD tried to introduce a clean x64 only CPU.

Legacy support doesn't come for free. It's a blessing and curse for Windows. If you really need it, it's good that it's there as a choice. On the other hand I'm happy that Apple is more willing to "break things" so we can get real advancements like Apple Silicon. This is "only" a scaled up tablet chip. I can't wait to see what they do with a design aimed expressly at the desktop with higher power and thermal ceilings.

And before someone points out there is some embedded only x64 variant out there - when you can show it booting Windows and running the usual mix of productivity software and games that people want, then you can call me. Apple will get the majority of Mac software ARM native before there would ever be a clean 64 bit x86 version of Windows and all the assorted third party software. Backwards compatibility is a drug the x86 crowd is grossly addicted to

 

[mike...]

Since your machine is fixed and not updatable I would never buy a machine without at least 16GB of RAM, even if I had to wait to get the extra $200 saved up. Yes, the new SSDs are a lot faster but why wear them out faster with a lot of swapping? I think people tend to forget and underestimate the wear rates in SSDs. High write loads can destroy consumer SSDs quickly (always amused when people want to run SSDs in network video recorders for camera systems - hope you have stock in SSD drive manufacturing companies).

This is an important point that a lot of people forget. If you only end up swapping just a little bit more than the memory you have available, in all likelihood the SSD will outlast your computer. However, if you do have demanding apps then you could easily kill an SSD that's not user-replaceable. I have first-hand experience of this using my 16GB MBP. I've already killed one SSD. On a typical day I write 200-300GB to disk mostly because of swap, and my worst ever day was just over a terabyte! Apple are shipping these machines with 256GB SSDs as the default, and something like a Samsung 870 Evo 256GB is warrantied at only 150TBW.

 

[DocNo]

Where did you hear this?

The color and they are still selling a more powerful Intel Mini are obvious giveaways. Yeah, it's speculation but its about as low risk speculation as it gets.

 

[djjeff]

No, you misunderstand my picking on your awful analogy, but the part in the post before that should have been clear.
If MacOS M1 takes up less space in RAM than MacOS x86, there is more room left over in the bucket for other applications.

Historically RISC based applications consumed more RAM than CISC based applications. The concept of RISC was to have a smaller number of instructions which performed less work per instruction but performed that work quickly. As a result, on average, more instructions were required to perform the same amount of work as a CISC processor.

In the early days of computing, when memory was expensive, a CISC design made sense. By using more complex instructions but reducing the number of them required to perform a specific task one needed less memory. Over time RISC became more like CISC and CISC became more like RISC. At least Intel implementations use a RISC type of core whereas the instruction decoding logic has do deal with the CISC instruction set.

I can't speak to code sizes for Apple Silicon implementations of ARM as it's too new. If it uses fewer instructions than x64 that would be impressive. However I doubt it would use half as much. Of course all this ignores data set size requirements which are, for the most part, implementation independent.

 

[djjeff]

Anandtech has a pretty good first look at the M1 and towards the front they talk about a few of them. The biggest is the pipeline width; Intel and AMD are pretty much stuck at how wide they can make the pipeline due to the way several instructions work.

x86 is still relevant because there is no pure x64 only CPU out there. If you think people wailing about Apple ditching Intel is epic, just watch what would happen if Intel or AMD tried to introduce a clean x64 only CPU.

Legacy support doesn't come for free. It's a blessing and curse for Windows. If you really need it, it's good that it's there as a choice. On the other hand I'm happy that Apple is more willing to "break things" so we can get real advancements like Apple Silicon. This is "only" a scaled up tablet chip. I can't wait to see what they do with a design aimed expressly at the desktop with higher power and thermal ceilings.

And before someone points out there is some embedded only x64 variant out there - when you can show it booting Windows and running the usual mix of productivity software and games that people want, then you can call me. Apple will get the majority of Mac software ARM native before there would ever be a clean 64 bit x86 version of Windows and all the assorted third party software. Backwards compatibility is a drug the x86 crowd is grossly addicted to

I'll give the Anandtech article a looksee. As for Intel / AMD breaking from x64 they already tried that with Itanium. While the Itanium had issues with its VLIW architecture it was the poor x86 performance which kept most people away from adopting it.

EDIT: Just read the Anandtech article you mentioned (at least I think it's the article you mentioned) and it said there were architectural constraints that the x86 decoding blocks would take substantial increases to increase but I didn't see that they couldn't be increased. This is one benefit of an orthogonal ISA.

 

[DocNo]

I know you think you're being clever, but macOS has had built-in memory compression since Mavericks.

And it's still not magic. I'm routinely venturing into swap with just Safari on Big Sur with my 2015 MBA. I'm hoping it's something they can resolve but I reiterate - it's silly to permanently handicap yourself for $200 for something so vital to performance. Drop down to the 7 core GPU to make it a $150 upgrade to 16GB if every dollar counts.

 

[mike...]

And it's still not magic. I'm routinely venturing into swap with just Safari on Big Sur with my 2015 MBA. I'm hoping it's something they can resolve but I reiterate - it's silly to permanently handicap yourself for $200 for something so vital to performance. Drop down to the 7 core GPU to make it a $150 upgrade to 16GB if every dollar counts.

Yeah I agree. $200 now is cheaper than having to upgrade early if you made a mistake or your machine's SSD died from constant writes, which is more likely on the cheaper machines as smaller SSDs usually have lower write endurance. If somebody is on the fence about 8GB vs 16GB they should just go for 16GB. The worst that can happen is they get a better machine!

 

[DocNo]

I also ask because I would love to see what the M1 could do in a development sense. Rather than "OMG, you're such a hater", I would love to see if a developer with a iOS / Swift background could talk about it out of experience. So far, I haven't seen very much of it, and it's probably because it's a MacBook Air - it's not for the development world only from the name itself. Guess I would have to touch base with the Mac mini crew.

People have - specifically the ATP boys. If you are doing light duty with xCode only, it blows away Marco's iMac Pro.

However do anything memory intensive like opening a few simulators and you are going to bounce off that 16GB RAM ceiling pretty quickly.

These are amazing machines, but they are ENTRY LEVEL - if you routinely need more than 16GB of RAM today (and as a developer I would think that is pretty common) these are not the machines for you. Your machines are coming

 

[DocNo]

Whats the take-away here ? If you find yourself saying stuff like "nobody will every need more than 8kb of RAM.....", you dont' know what you are talking about.

Indeed.

Not ONE TIME in using computers for over 30 years now have I ever felt the need to complain about having too much RAM.

Now, having too little? More times than care to remember. Including the 2015 MBA I'm typing on now

 

[DocNo]

That does not invalidate the claims posted here. The scenarios I have tested comparing an iPad Pro to my 2019 i9 iMac with 64GB of RAM.

Your iPad is running an ENTIRELY different OS with an ENTIRELY different memory management model.

You are NOT comparing similar things. ARM is not magical. It does not change the laws of physics. Apple Silicon Macs will be FAR closer to Intel Macs for memory requirement comparisons. ARM has some efficiencies that may make a few hundred megabyte differences but it's not going to be gigabytes.

And even if the new SSDs swap fast enough that you don't notice, flash memory in SSDs has a fixed lifespan. A fixed life span that is orders of magnitude lower than RAM chips. Have fun repairing that permanently attached SSD you prematurely wear out from excessive swapping all to save $200. $40 a year if you keep it for five years. Less if you keep it longer.

If you routinely swap with 8GB on your Mac today, it would be NUTS to get an 8GB M1 Mac just because you think ARM has some sort of mystical RAM fairy.

The only reason my current '15 MBA only has 8GB is it was the max that was offered at the time. Safari has become such a pig in Big Sur I can't wait to upgrade to a machine that can hold 16GB of RAM.

 

[DocNo]

I don't get this whole "x amount per year" thing. Are we buying the machines or are we renting the machines?

Total cost of ownership? Return on investment? Value over time?

$200 for an additional 8 GB of RAM - special low power RAM on top of that (also imposes the 16GB limit BTW).

That insane battery life? That higher cost, low power DDR4 is part of the equation that delivers the insane battery life.

That may not be as important to you - if so than this probably isn't the machine for you either. Nothing wrong with that - machines that try to be all things to everyone are also very often of little value to everyone too.

SSD costs - I am more with you there. Thankfully I comfortably fit with 512GB sized drive for how I use my MBA and thats a reasonable upgrade. The higher capacity ones are relatively nuts though so I'm glad I don't even need to be tempted by them

 

[DocNo]

If one doesn't require more than 8GB then it's $200 wasted. Keep in mind that on the lowest cost MBA this $200 represents 20% of the system cost. Maybe wise to save the money for an updated system when they're released and the OP needs a new one.

I never once have lamented having too much RAM. On more occasions than I care to recall I have lamented not having enough RAM. If you are going to be a typical Mac user and keep your machine for five years, $200 now is pretty cheap amortized over five years.

It doesn't take much to fill 8GB these days - as I have said many times, in particular Safari on Big Sur is using a lot more memory than it did under Catalina. I suppose if you never use sites like YouTube on the Internet it won't be an issue.

You have one chance to pick your RAM; after that you are stuck.

I'll put it this way - I'd rather have the RAM than Apple Care if I could only afford one

 

[DocNo]

I'll give the Anandtech article a looksee. As for Intel / AMD breaking from x64 they already tried that with Itanium. While the Itanium had issues with its VLIW architecture it was the poor x86 performance which kept most people away from adopting it.

EDIT: Just read the Anandtech article you mentioned (at least I think it's the article you mentioned) and it said there were architectural constraints that the x86 decoding blocks would take substantial increases to increase but I didn't see that they couldn't be increased. This is one benefit of an orthogonal ISA.

Itanium's failure was due to shenanigans with HP more than anything else.

That and the "backwards compatibility at all costs" mindset prevalent in the PC space. USB was available from Intel for YEARS but only broke into the PC space as a standard feature due to.... the original candy colored iMacs and all the fun candy colored USB accessories that got made chasing after a piece of the iMac popularity wave. Otherwise PC's would still probably be screwing with serial and parallel connections.

Indeed I had DIMMs on my Mac for almost a decade before they got popular in PCs - plugging in 7 DIPs for a RAM upgrade (8 if you were fancy and went for ECC RAM) went on for a long time. I still regret getting rid of my first 386dx board that 4MB of DIP and 4MB of DIMM RAM. So ugly it had it's own beauty Ran OS/2 like a top too.

As for the didn't say x86 couldn't be increased - you don't think Intel would be all over it already if it was even remotely possible? Especially after being as stalled out in their process improvements as they have been for almost 10 years now? I mean it's possible they were just holding back until Apple up and spanked them - but we might be venturing into infinite probability drive territory with that speculation

 

[aesc80]

People have - specifically the ATP boys. If you are doing light duty with xCode only, it blows away Marco's iMac Pro.

However do anything memory intensive like opening a few simulators and you are going to bounce off that 16GB RAM ceiling pretty quickly.

These are amazing machines, but they are ENTRY LEVEL - if you routinely need more than 16GB of RAM today (and as a developer I would think that is pretty common) these are not the machines for you. Your machines are coming

See, THIS is a voice of reason! You have a number of people here to quickly say "YoU dOn'T nEeD mOrE tHaN 16GB eVeR AnYmOrE", like all of a sudden we've become memory communists up in here!

Agreed! It's entry level and it's definitely not in its scope. However, its perfectly a good ask to see how well it would perform at a professional level since this is an entirely new architecture. Sure, I can open 10 different versions of Calendar on this machine ... its just not impressive to someone like myself that uses more than the pedestrian stuff here. If 16GB of memory with an M1 chip is all it takes, that would be dope!

So thank you so much for the sound response. Sound criticism and a desire to ask for more is not hate, and it especially doesn't make you any less of an Apple fan. It just means you want more knowledge rather than handing over your cash because an Apple insignia is on it.

 

[cupcakes2000]

The original poster said “what happens when you use them all at once and really need them”. These are ALL lightweight programs, you just switch to it and use it. Additionally, there’s no one using Keynote and Numbers and GarageBand simultaneously. Maybe Safari could be downloading something in the background, maybe Podcasts could be playing, but those are all lightweight tasks. There’s not an intensive app in the bunch. The only apps shown are the ones that the vast majority of macOS users will actually use.

Maybe or maybe not, I’m not able to know how others use their stuff. However - I was responding to your below quote, where you said ‘Does anyone really use more than one application at a time?’

Yes they do. I do daily.

Does anyone really use more than one application at a time? Having windows “available” isn’t the same as actually typing into one window while touching up a drawing in another while meticulously fine tuning the color in another while scrolling through a webpage in yet another. I’m not even sure the UI handles multiple targets.

 

[DocNo]

See, THIS is a voice of reason! You have a number of people here to quickly say "YoU dOn'T nEeD mOrE tHaN 16GB eVeR AnYmOrE", like all of a sudden we've become memory communists up in here!

It's beyond nonsensical.

Never ONCE have I complained about having too much RAM. Or CPU speed. Or disk space (especially with SSDs!). But I routinely run into limitations with all of the above.

If we were talking 32GB vs 128GB or something that would be one thing, but sadly these days 8GB is pretty easy to blow past. Especially with Big Sur and Safari right now (and these machines are stuck at Big Sur - no going back to anything earlier).

 

[djjeff]

I never once have lamented having too much RAM. On more occasions than I care to recall I have lamented not having enough RAM. If you are going to be a typical Mac user and keep your machine for five years, $200 now is pretty cheap amortized over five years.

It doesn't take much to fill 8GB these days - as I have said many times, in particular Safari on Big Sur is using a lot more memory than it did under Catalina. I suppose if you never use sites like YouTube on the Internet it won't be an issue.

You have one chance to pick your RAM; after that you are stuck.

I'll put it this way - I'd rather have the RAM than Apple Care if I could only afford one

I understand where you're coming from and I dislike the trend away from post purchase RAM upgrades. However I think one of the reasons we see so many comments about the M1 magically reducing RAM requirements is because those making such claims over estimated their RAM requirements and purchased more than they really needed.

The Windows PC I am responding with only has 8GB of RAM and it's serving its purpose just fine. I could upgrade it to 16GB but for what I use it for I doubt I would see any appreciable benefit.

If I were buying one of these new systems (and I just might) I would strongly consider the 16GB version.

 

[djjeff]

As for the didn't say x86 couldn't be increased - you don't think Intel would be all over it already if it was even remotely possible? Especially after being as stalled out in their process improvements as they have been for almost 10 years now? I mean it's possible they were just holding back until Apple up and spanked them - but we might be venturing into infinite probability drive territory with that speculation

I really don't know what challenges Intel is facing. However I do think they are hamstrung with the backwards compatibility requirement. The PC I am responding on can still run software dating back to the 80s. Good or bad that's the reality.

That said I am waiting to see how the Apple Silicon (AS) compares in higher end systems. I fully expected their entry (low end?) systems would be very capable compared to x64. I am curious if they can achieve the same with higher end systems. Regardless I expect them to be very competitive.

What is of interest to me is the pace at which Apple will be able to maintain large increases in performance. This is like being in the 90's all over again. Will AS eventually see diminishing performance gains in a similar manner being attributed to x64?