Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Maximara

macrumors 68000
Jun 16, 2008
1,707
908
I see the fiction continues....

Apple has already demoed full Office (and no - LibreOffice is trash by comparison) running natively on Apple Silicon (not ARM - Apple's home designed SOCs only use the ARM ISA; they are NOT Cortex). They likewise demoed Creative Suite - which means all CS products are already ready. They also demoed Autodesk Maya to show how fast Rosetta 2 is (and it is fast).

Also, the latest market shares have Apple much higher than 6% of the market. It is more like 17% market share for MacOS worldwide and 27% in the US:


Even the more conservative netmarketshare statistics has MacOS worldwide marketshare at 9.55% Though if you really want a real trip down the rabbit hole come over to the Could x86-64 emulation be possible on ARM Macs? thread.

If you are running 3D CAD software the ARM Mac will likely slow to a crawl because it is having to emulate an Intel CPU but if running Photoshop, Adobe has re-written its software to take advantage of the new hardware.

Ok. Let's be clear here. Rosetta 2 does not emulate anything; Rosetta 2 translates

"Rosetta 2 essentially “translates” instructions that were written for Intel processors into commands that Apple’s chips can understand. (...) Rosetta 2 can convert an application right at installation time, effectively creating an ARM-optimized version of the app before you’ve opened it." - Rosetta 2 is Apple’s key to making the ARM transition less painful

"Rosetta is a translation process that allows users to run apps that contain x86_64 instructions on Apple silicon. (...) If an executable contains only Intel instructions, macOS automatically launches Rosetta and begins the translation process. When translation finishes, the system launches the translated executable in place of the original. - About the Rosetta Translation Environment

For the TL;DR croud translation is NOT emulation.
 
Last edited:

throAU

macrumors G3
Feb 13, 2012
9,198
7,346
Perth, Western Australia
The critical question is "Faster for what?"

I'm sure the new Apple Silicone will make some operations much faster and others much slower. But really what do 99% of Mac users do? They type email and watch Youtube. A 5 minute Youtube video will still take 5 minutes to play. and that email will still take just as long to read.

If you are running 3D CAD software the ARM Mac will likely slow to a crawl because it is having to emulate an Intel CPU but if running Photoshop, Adobe has re-written its software to take advantage of the new hardware.

So 99% of users will not notice the change as they don't push their computers. For the 1% who do, I expect wildly different results. Overall it will be a win for Apple even if they do lose 0.5% of their customers.

Nah...

I don't think a lot of people posting here realise just how freaking far behind intel is, and how despite AMD being massively ahead of intel (and don't get me wrong, AMD are doing great things CPU wise), they're still quite a way back from where they could be due to maintaining compatibility with 4 decades of x86 quirks for software compatibility reasons.

ARM when originally released back in the 80s was WAY a head of anything else on the market. Like 2-4x as fast as the fastest desktop PC of the day in terms of CPU. As fate/circumstance/home computer market would have it, they ended up being marginalised in home computers and only really saw use in the education sector with the original Acorn Archimedes and RiscPC but IBM compatibility was what everyone at the time wanted. The rest of the home computer market (Acorn, Commodore, Atari, etc.) died, and the ARM based home/business computer died with it.

Fast forward 30 years and ARM is used in everything outside of PCs and Macs and intel have hit several walls - brain drain losing their most competent people to elsewhere, massive fabrication failures (and intel fab was previously a home court advantage for intel which enabled even a relatively mediocre design to perform well due to clock speed advantage from better manufacturing). Right now intel are staring down the barrel of no 7nm process until 2022-2023 at which point TSMC will be on 3nm, and intel are also scrabbling to fix 20 years of performance vs. security hacks. They're WAY behind on fab, and WAY behind on design capability as a result. They simply can't build a competitive product on their own manufacturing tech and ship it in volume.

Right now intel are behind and their roadmap (what they are proposing even) is bad. And they haven't successfully executed on their roadmap since 2014. They've repeatedly slipped. They delayed 14nm, they delayed 10nm and they've delayed 7nm. Their core is not that different since Skylake which itself is only a slight evolution of Sandy bridge from 2011!

The ARM chips will blow intel away (already do, even if we aren't comparing apples to apples; the Apple processors handily beating intel even being limited to FAR less power consumption and thermal headroom by virtue of being inside fan-less phones and tablets), not because of magic or because or some unbelievable progress Apple has made, but purely because intel and the x86 market has stumbled so badly and stagnated for near on a decade that they have been caught up and passed. You simply can't stagnate for a decade like intel has and expect to remain at the front. Apple has had what would have previously been accepted as "expected" performance gains, generation on generation.

Ryzen is great relatively speaking but AMD stumbled for 4-5 years themselves with bulldozer. the 2010-2016 era was essentially close to ZERO progress by either AMD or intel. AMD because they bet the farm on a design that didn't pan out and nearly went broke; intel because they grew complacent with no competition in their markets. Ryzen if compared to what we should have seen in terms of progress from both AMD and intel is where we should have been in 2014 or thereabouts, if both companies didn't have epic failures in execution in the same decade.

As to pro users - expect more from things like the afterburner card and other bespoke apple hardware for particular tasks. Look up what FPGAs are and understand that afterburner is a big powerful FPGA (Both why it is so expensive, and also why it is so powerful) - essentially hardware that can be reconfigured for optimal performance on specific tasks pretty much as quickly as loading a different app. And it then does those tasks in HARDWARE. Right now it does pro-res. But that's not the only thing it can do, if programmed.

Performance is the least of our worries with Apple Silicon and apple getting of x86 in general. As above, the A13 is outperforming intel parts in 10-20% of the power draw. And that's last year's hardware in a mobile phone form factor.

The future looks great. The new parts will enable entirely new ways of interacting with your computer. They're going to be a game changer. Think less about "is my laptop going to be any faster" and more "what will new apps be able to do with 5-10x or more performance in the next 5 years?"
 
Last edited:

johngwheeler

macrumors 6502a
Dec 30, 2010
639
211
I come from a land down-under...
Fast forward 30 years and ARM is used in everything outside of PCs and Macs and intel have hit several walls

...

The ARM chips will blow intel away (already do, even if we aren't comparing apples to apples; the Apple processors handily beating intel even being limited to FAR less power consumption and thermal headroom by virtue of being inside fan-less phones and tablets), ...

Performance is the least of our worries with Apple Silicon and apple getting of x86 in general. As above, the A13 is outperforming intel parts in 10-20% of the power draw. And that's last year's hardware in a mobile phone form factor.

Well, not quite...Intel and AMD are used in at least 98% of servers / data centers, and that is a very large part of the CPU market (35-40%). ARM is growing in this segement, but is still a very small player. Increased use of ARM in personal computers will definitely increase the market share of ARM-based CPUs, and will probably have a knock-on effect in increasing server usage too.

My optimism about possible real-world performance is more measured than yours. Yes, Apple-Silicon and ARM CPUs have very good performance/watt, but I'm not sure how well these will scale up with increased clock-speeds even with active cooling.

We can't just assume that because an A14 uses 20% the power of a 28W Intel Tiger Lake, that Apple will be able to bump it up to 28W and get 5x the speed. TDP is linearly proportional to frequency and to the square of the voltage. You can't easily increase clock speed beyond 4-5GHz, and I don't know if there is any advantage in increasing voltage beyond what it already is (which keeps power consumption low).

So I imagine this leaves adding more cores, which will not lead to linear improvements for applications that can't use the cores.

Another limitation is power/area - as the lithography shrinks, the amount of heat per unit of area increases, so a 5nm chip will have a tougher job removing heat compared to a 10nm chip, given the same TDP.

I think we will see diminishing returns. Apple Silicon will be faster than current Intel 10th/11th gen, but maybe of the order of 50-100%, and it will get harder each year.
 

Joelist

macrumors 6502
Jan 28, 2014
463
373
Illinois
We REALLY need to stop conflating Apple Silicon and ARM - they are not the same thing. literally the only similarity is Apple using the ISA - the microarchitecture (which is far and away the main driver of processor performance and efficiency) is totally different. In fact, the only thing out there it somewhat resembles is the Intel Core 2 microarchitecture. This is why it runs rings around the real ARM (Cortex based) SOCs. Big wide pipes with out of order executuion and super accurate branch prediction.

And this microarchitecture is why Apple already knows their ASi platform will scale up well for the Mac - it already has the needed beef and power. The A14 has 2 High Performance and 4 Low Power cores - I expect the Mac SOC family to be more like 6-8 High Performance and 6 Low Power with the attendant increases in cache.
 

Maximara

macrumors 68000
Jun 16, 2008
1,707
908
I think we will see diminishing returns. Apple Silicon will be faster than current Intel 10th/11th gen, but maybe of the order of 50-100%, and it will get harder each year.

Moore's Law died a while ago so the shrinkage wasn't going to continue. Also somewhere along the line Quantum Tunneling is going to happen and that will be it for making chips smaller. The only option will be what Intel is doing now - throw more cores on the chip.
 

throAU

macrumors G3
Feb 13, 2012
9,198
7,346
Perth, Western Australia
Well, not quite...Intel and AMD are used in at least 98% of servers / data centers, and that is a very large part of the CPU market (35-40%). ARM is growing in this segement, but is still a very small player. Increased use of ARM in personal computers will definitely increase the market share of ARM-based CPUs, and will probably have a knock-on effect in increasing server usage too.

You're talking about PC based servers, which I covered with my "PCs" statement above.

There's also a whole heap of ARM based servers, Amazon run them. Sparc, pa-risc, etc. also do exist.

I'm talking about end devices. there are BILLIONS of ARM devices out there from cars, to fridges to watches to tablets to washing machines, etc.
[automerge]1600841454[/automerge]
Moore's Law died a while ago so the shrinkage wasn't going to continue. Also somewhere along the line Quantum Tunneling is going to happen and that will be it for making chips smaller. The only option will be what Intel is doing now - throw more cores on the chip.

Moores law is based on transistor count (not performance) and is alive and well - over at TSMC.

If intel can throw cores at the problem, so can Apple. And they're on a better process so they can fit more in the same thermal/power envelope.
 

throAU

macrumors G3
Feb 13, 2012
9,198
7,346
Perth, Western Australia
We REALLY need to stop conflating Apple Silicon and ARM

Not really, they're ARM compatible, based on an ARM licensed core, and (most importantly) Arm is a lot easier to type than "Apple Silicon". "ARM based Mac" is fairly unambiguous because there's no competitor arm based Mac running anything different.

Unless you want me to abbreviate to AS. All Apple Silicon is, is marketing for their extensions, and we've been talking about ARM Macs long before the marketing department got involved.

*shrug*

Same way I write x86 to cover both intel and AMD. neither of those chips bear much resemblance in terms of design to the original 8086 but they're still compatible and still part of the same family tree.
 

johngwheeler

macrumors 6502a
Dec 30, 2010
639
211
I come from a land down-under...
Moore's Law died a while ago so the shrinkage wasn't going to continue. Also somewhere along the line Quantum Tunneling is going to happen and that will be it for making chips smaller. The only option will be what Intel is doing now - throw more cores on the chip.

I'm surprised that TSMC thinks they will be able to produce 3nm chips at volume by 2022, but if they do, it sounds like it will be very hard to go smaller without technical advances with different semi-conductor materials or exotic things like quantum computing (which sound like they are decades away from practical application).

So, yes, the easiest solution at the moment is to add more cores, which is a potential strength for ARM-based designs that have already scaled to 128 cores. Now the software has to find a way to use these cores efficiently - which may well be the greater challenge.
 

leman

macrumors Core
Oct 14, 2008
19,518
19,666
I don't think a lot of people posting here realise just how freaking far behind intel is, and how despite AMD being massively ahead of intel (and don't get me wrong, AMD are doing great things CPU wise), they're still quite a way back from where they could be due to maintaining compatibility with 4 decades of x86 quirks for software compatibility reasons.

I don't know if AMD is that much ahead of Intel to be honest. They had an efficiency advantage (which had mostly disappeared with Tiger Lake), and they always lagged behind in single-threaded performance. Once Intel brings out 8-core Willow Cove designs, they should outperform AMD's offerings. AMD had an advantage in integrated graphics performance, but Intel has closed that one as well.

Of course, all depends on how well Zen 3 will turn out. Finger's crossed that AMD's latest can compete with both Intel and Nvidia. More real competition on the market means better customer experience.

based on an ARM licensed core

AFAIK, Apple doesn't use ARM IP in their designs. Which is one of the reasons they perform that well. But new ARM Neoverse should be a formidable competitor.
 
  • Like
Reactions: Joelist

Maximara

macrumors 68000
Jun 16, 2008
1,707
908
Moores law is based on transistor count (not performance) and is alive and well - over at TSMC.

Based on what I was told about the way nm are used I double checked and that is NOT correct.

Transistor density (MTr/mm2): 7 nm TSMC N7FF with 96.5; Intel 10 nm at 100.76. Last time I checked 96.5 is less than 100.76, So you don't have Moore ( :p ) transistors with the TSMC process regardless of you using transistor count or transistor density

A key component of Moore's Law is that a certain increase in transistor count is to happen within a certain period of time.

In fact, the original 1965 Moore's Law was "the number of transistors per silicon chip doubles every year." When that slowed Moore revised it, in 1975, to the more pessimistic "the number of transistors per silicon chip doubles every two years." - Encyclopaedia Britannica

"Moore's Law, by the strictest definition of doubling chip densities every two years, isn't happening anymore," - CES 2019: Moore's Law is dead, says Nvidia's CEO

Then there is this variant via David House: computer chip performance would roughly double every 18 months (with no increase in power consumption).

The original Moore's Law is indeed dead and has been dead for a long time

If intel can throw cores at the problem, so can Apple. And they're on a better process so they can fit more in the same thermal/power envelope.
Definitely.
 
Last edited:

Tafkaeken

macrumors member
Oct 6, 2018
81
62
Slightly OT, but I heard something like that “there are as many transistors in one iPhone that it was in the world 1998”. Does anybody know the exact fun fact here?
 

throAU

macrumors G3
Feb 13, 2012
9,198
7,346
Perth, Western Australia
Once Intel brings out 8-core Willow Cove designs, they should outperform AMD's offerings.

Sure. It's been 4-5 years since 10nm parts were originally promised to be shipping in volume and they still aren't yet. Intel 10nm is a joke and isn't shipping in volume any time soon for parts bigger than 4 cores. Never mind 7nm, which was due in 2020 and has been pushed back to 2023 at this stage, at which point TSMC will be on 3nm (or at least 5nm+).

You say intel have caught up to AMD, but the difference is AMD was shipping Ryzen 2 parts at the start of the year, and you still can't buy a tiger lake machine right now. They're vapour, intel can't actually build them in volume yet. Also, beating AMD's integrated Vega 11 GPU from 3 years ago? GG intel. Really. I guess.

Never mind things with more than 8 cores.

And yes, as per my post above, the only reason AMD looks so great vs. intel at the moment is because intel is doing so badly - AND they have a great fab with TSMC, not because AMD is amazingly spectacular. But within the x86 space, AMD is winning right now.

Against Apple's parts and ARM in general? They're both behind, but intel is further back.
 
  • Like
Reactions: Maximara

leman

macrumors Core
Oct 14, 2008
19,518
19,666
Sure. It's been 4-5 years since 10nm parts were originally promised to be shipping in volume and they still aren't yet. Intel 10nm is a joke and isn't shipping in volume any time soon for parts bigger than 4 cores. Never mind 7nm, which was due in 2020 and has been pushed back to 2023 at this stage, at which point TSMC will be on 3nm (or at least 5nm+).

Ah, I see what you mean. Yeah, until Intel is able to ship Tiger Lake in volume, their (arguable) technical superiority doesn't mean much...
 

Joelist

macrumors 6502
Jan 28, 2014
463
373
Illinois
I don't know if AMD is that much ahead of Intel to be honest. They had an efficiency advantage (which had mostly disappeared with Tiger Lake), and they always lagged behind in single-threaded performance. Once Intel brings out 8-core Willow Cove designs, they should outperform AMD's offerings. AMD had an advantage in integrated graphics performance, but Intel has closed that one as well.

Of course, all depends on how well Zen 3 will turn out. Finger's crossed that AMD's latest can compete with both Intel and Nvidia. More real competition on the market means better customer experience.



AFAIK, Apple doesn't use ARM IP in their designs. Which is one of the reasons they perform that well. But new ARM Neoverse should be a formidable competitor.

Thanks! I don't know why people out there still think Apple Silicon uses ARM designs. It has been common knowledge since the A6 that Apple uses ONLY the instruction set and nothing else, and that the core designs and the rest of the microarchitecture is 100% Apple designed and created.
 
  • Like
Reactions: Maximara

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
Not really, they're ARM compatible, based on an ARM licensed core, and (most importantly) Arm is a lot easier to type than "Apple Silicon". "ARM based Mac" is fairly unambiguous because there's no competitor arm based Mac running anything different.

Unless you want me to abbreviate to AS. All Apple Silicon is, is marketing for their extensions, and we've been talking about ARM Macs long before the marketing department got involved.
Apple doesn’t use ARM licensed cores. They have an architecture license which means that they can design their own architecture around an ARM compatible instruction set—potentially with extensions. This is really important for their competitiveness. Even Samsung is using some ARM designed cores in their SoCs and Qualcomm seems to have switched over to only using ARM cores. Part of this is that the recent ARM cores look very competitive to Apple’s recent designs. We will see how that plays out on the A14 and whatever Apple comes up with for Macs (probably the same CPU cores but more cores than the A14 with other changes like more cache and more GPU cores).
 

dmccloud

macrumors 68040
Sep 7, 2009
3,138
1,899
Anchorage, AK
We REALLY need to stop conflating Apple Silicon and ARM - they are not the same thing. literally the only similarity is Apple using the ISA - the microarchitecture (which is far and away the main driver of processor performance and efficiency) is totally different. In fact, the only thing out there it somewhat resembles is the Intel Core 2 microarchitecture. This is why it runs rings around the real ARM (Cortex based) SOCs. Big wide pipes with out of order executuion and super accurate branch prediction.

And this microarchitecture is why Apple already knows their ASi platform will scale up well for the Mac - it already has the needed beef and power. The A14 has 2 High Performance and 4 Low Power cores - I expect the Mac SOC family to be more like 6-8 High Performance and 6 Low Power with the attendant increases in cache.

Actually, Apple is using the ARM instruction set (not the microarchitecture), just as AMD uses the x86 instruction set for its Ryzen and Threadripper CPUs. That's why people use the terms "Apple Silicon" and "ARM" interchangeably - one term refers to the processors themselves, the other applies to the instruction set the processors use. You also somehow managed to simultaneously claim that Apple was using the ARM microarchitecture at the same time you admitted they were not running the Cortex microarchitecture.
 
  • Like
Reactions: Maximara

dmccloud

macrumors 68040
Sep 7, 2009
3,138
1,899
Anchorage, AK
You say intel have caught up to AMD, but the difference is AMD was shipping Ryzen 2 parts at the start of the year, and you still can't buy a tiger lake machine right now. They're vapour, intel can't actually build them in volume yet. Also, beating AMD's integrated Vega 11 GPU from 3 years ago? GG intel. Really. I guess.

What's worse about Intel's dog and pony show from a couple of weeks ago is that they only compared the integrated performance of their 11th gen processors (which will only replace Y and U-series mobile SKUs, nothing is even in the pipeline for either gaming notebooks or desktops at the moment) to that 3 year old Vega GPU and a workstation class MX350 and only as it pertained to gaming. Most of that announcement was a self high-five and a rebranding/reimaging of their processor lineup, and Intel formally introducing Project Athena to the public under a different name. Regarding Ryzen 2, did you mean 3rd gen Ryzen parts?
 
  • Like
Reactions: Maximara

Joelist

macrumors 6502
Jan 28, 2014
463
373
Illinois
Actually, Apple is using the ARM instruction set (not the microarchitecture), just as AMD uses the x86 instruction set for its Ryzen and Threadripper CPUs. That's why people use the terms "Apple Silicon" and "ARM" interchangeably - one term refers to the processors themselves, the other applies to the instruction set the processors use. You also somehow managed to simultaneously claim that Apple was using the ARM microarchitecture at the same time you admitted they were not running the Cortex microarchitecture.

What!?! I was the one who pointed out that they do not run Cortex; they use their own. This is why I think the terms ARM and Apple Silicon should not be interchanged, as the term ARM implies Cortex. Also the ISA is nowhere near the driver of performance or efficiencty that the microarchitecture is.
 

dmccloud

macrumors 68040
Sep 7, 2009
3,138
1,899
Anchorage, AK
What!?! I was the one who pointed out that they do not run Cortex; they use their own. This is why I think the terms ARM and Apple Silicon should not be interchanged, as the term ARM implies Cortex. Also the ISA is nowhere near the driver of performance or efficiencty that the microarchitecture is.

ARM is interchangeable with x86 for all intents and purposes, because both terms refer to the instruction sets. This term can also refer to Cortex (A-profile, R-profile, or M-profile), SecurCore, Ethos, or the new Neoverse cores, since they are all built upon the same ISA:

hhttps://www.arm.com/products/silicon-ip-cpu
 

Joelist

macrumors 6502
Jan 28, 2014
463
373
Illinois
ARM is interchangeable with x86 for all intents and purposes, because both terms refer to the instruction sets. This term can also refer to Cortex (A-profile, R-profile, or M-profile), SecurCore, Ethos, or the new Neoverse cores, since they are all built upon the same ISA:

hhttps://www.arm.com/products/silicon-ip-cpu

I think you misunderstood my post then - I agree that x86 and ARM are both ISAs. But ARM and Apple Silicon are definitely NOT interchangeable.
 

throAU

macrumors G3
Feb 13, 2012
9,198
7,346
Perth, Western Australia
Apple doesn’t use ARM licensed cores. They have an architecture license which means that they can design their own architecture around an ARM compatible instruction set—potentially with extensions.

Yes, I'm aware of that. They use the ARM ISA and thus, as far as I am concerned, they're ARM (Based). They've been discussed as ARM for the past 5 years of rumours, and despite Apple marketing, they're still going to be descended from ARM (and called "ARM based Macs") for their product life.

Like I said above, people still call intel and AMD processors "x86" despite the 8086 they are named after being from the late 1970s-early 80s and having had many major architecture shifts and additions made to it.

They're still part of the ARM family tree; they're ARM. ARM don't actually make processors, everybody licenses a core and extends it, Apple are no different.
 
Last edited:
  • Like
Reactions: Maximara

bill-p

macrumors 68030
Jul 23, 2011
2,929
1,589
I think you misunderstood my post then - I agree that x86 and ARM are both ISAs. But ARM and Apple Silicon are definitely NOT interchangeable.

You're basically saying Qualcomm's Snapdragon is also not ARM because their processors don't use Cortex core architecture either.
 
  • Like
Reactions: throAU

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
You're basically saying Qualcomm's Snapdragon is also not ARM because their processors don't use Cortex core architecture either.
Their newest ones actually do. They’ve stopped designing their own architectures recently.
 
Last edited:

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
Yes, I'm aware of that. They use the ARM ISA and thus, as far as I am concerned, they're ARM (Based). They've been discussed as ARM for the past 5 years of rumours, and despite Apple marketing, they're still going to be descended from ARM (and called "ARM based Macs" for their product life.

Like I said above, people still call intel and AMD processors "x86" despite the 8086 they are named after being from the late 1970s-early 80s and having had many major architecture shifts and additions made to it.

They're still part of the ARM family tree; they're ARM. ARM don't actually make processors, everybody licenses a core and extends it, Apple are no different.
This just isn’t true. You are wrong, sorry. Apple is not extending an existing ARM coretex design. They have their own independent architecture, not licensing an ARM core. Saying the same thing in different ways doesn’t make it any more true.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.