Will Intel and the others catch up to match M1

[Bearbrickster]

Now that we know the M1 chip is giving Apple a big advantage to the Intel/windows combo. When will Intel/windows start going the Arm route?

This entry level M1 is beating some of Intel/AMD’s desktop chip. When the next iteration of the M1 chip comes out for the 16mbp, i mac and mac pro, I think its safe to say it will run circles round Intel and AMD’s most expensive chips.

So when will Intel or AMD or Samsung fab make an ARM based CPU for the windows platform? 3 years? 5 years?

Samsung fab is/will be making 5nm ARM based chips for their phones now. Can samsung work together with windows to make an ARM chip for the windows PC platform?

 

[revs]

There are ARM based Windows machines. The performance is poor.

ARM isnt what makes these fast - it's Apples implementation of ARM, their chips.

Apple have been making super fast ARM chips in their iOS devices for years, and nobody has caught up to them yet. I'm sure at some point they will, at some point I'm sure we will be back at the 'why arent apple using x86 chips, these ARM chips are slow' stage that we had with the G5 ? (but I hope not!)

 

[Serban55]

Very hard, but lets hope so...for now Intel has no chance...Amd its the closest
But to catch up, that means Apple to fire or let go to their chip brain guy...or, for Apple to stagnate for 2 years

 

[Gnattu]

Apple is not selling chips to 3rd parties, so no direct competition. AMD is already doing a great job and the performance is not bad at all for Zen3. Intel definitely needs to catch up now for sure, but most pressure is not from Apple.

 

[ian87w]

In the near future, on mobile space, probably no. And intel doesn't really have enough pressure to do anything. Sure, they lost Apple as their client, but Apple is not selling their M chips to anyone else, so everyone else are still intel's clients. It's still an x86 world afterall.

Samsung's ARM is behind Qualcomm, and just barely reached parity with Qualcomm by 2021. And that's just on mobile. Going to implement a mobile chipset into a full blown laptop/desktop chip will take one or two more learning curves.

I'm sure Microsoft is really interested in working with Qualcomm or Samsung. Microsoft, despite their "wintel," is not really good friends with intel.

The problem is Qualcomm doesn't seem to feel any competition/pressure. Since Apple is not licensing their silicon to others, Qualcomm literally have the monopoly of high end and performance consumer ARM SoCs (plus their monopoly in the modem space is not helping either). Just look at their half-assed attempt at wearable chips. Even Google gave up on it thanks to Qualcomm (and Samsung went with their own Tizen instead).

The hope, somewhat, is actually from Huawei with their Kirin processors. But Huawei has their own challenges.

In the near future, Apple M1 chips will be peerless. OEMs will be scrambling, putting whatever Qualcomm chips they can find for the next few years, just to match Apple's battery life claims. Microsoft will do their best, but the consumers will not want an ARM Windows until Microsoft can pull off a full transparent emulation for x86 apps. By then, Windows machines will be at the starting line. But then Apple might have lapped them once, or maybe twice. That is if Apple also keep their momentum going. Sometimes Apple also has their own fumbles, allowing the competition to catch up.

 

[Bearbrickster]

Wandering if windows x86 will ever get close to the performance of the ARM Mac or maybe windows have to go down the ARM route? Windows will lose market share if they cant get close to the mac’s performance. Apple is selling computers with much more performance at the similar price point

 

[Serban55]

Wandering if windows x86 will ever get close to the performance of the ARM Mac or maybe windows have to go down the ARM route? Windows will lose market share if they cant get close to the mac’s performance. Apple is selling computers with much more performance at the similar price point

Microsoft with Windows Already tried the arm route with QUALCOMM...i have for 1 year almost surface pro x ... and its horrible
First time they tried to have the best battery life, and that was true but nothing was working well...and second time, in present they tried to increase the perf but they lost half from that good battery life and still no support or very poor emulation software
but without controlling your own chip and not having the dev on your side/or having a very good emulator....is DOA
x86 will never get close to arm in the near future...x86 will face an wall after they go 3nm...because arm will be far better in perf/w X86 is and old arhitecture
i bet amd will do anything to go arm also
But bottom line is, for as long as Apple has the best chip brain in the industry Johny, others cannot match Apple

 

[Zazoh]

Isn’t it the System on a Chip (SOC) aspect that adds the speed by eliminating the need of going through all the busses?

 

[ian87w]

Wandering if windows x86 will ever get close to the performance of the ARM Mac or maybe windows have to go down the ARM route? Windows will lose market share if they cant get close to the mac’s performance. Apple is selling computers with much more performance at the similar price point

I doubt that Windows will lose much marketshare. Majority of the computers in the world is in the enterprise, and even many are still on legacy Windows 7 or older. And they are all entrenched in the world of Windows servers/Active Directory. Considering Apple doesn't seem to be touching the enterprise market anymore (no more Xserve, etc), I don't think Apple will become the majority anytime soon.

Even in the consumer space, majority of the laptops in the world are the low end laptops, those $500 or under. Apple is not in that space, so these consumers won't suddenly be willing to spend double for a laptop.

But Apple will definitely take a chunk in the premium space, as usual (same thing with smartphone market). People already spending $1500-$2000 on a laptop will cringe looking at the stagnant performance of wintel's offerings, and they might switch to Macs.

 

[Serban55]

Isn’t it the System on a Chip (SOC) aspect that adds the speed by eliminating the need of going through all the busses?

of course, thats also a a thing...but for PC industry...where you have a lot of customers who needs access to their components....is very hard....on mac users..for a decade we are used not to be able to add or remove ram, or dGpu or anything .....but for PC users...i dont know.
Apple is thinking to bring the storage also into the SOC....i dont thing half of the pc users wants everything "irreplaceable "

 

[ArPe]

They can catch up with performance next year but not at such low power consumption. Intel and AMD are using 8 full size mobile cores with hyperthreading to get 7000-8000 multi core Geekbench score. M1 can do it with just 4 full size cores and 4 little cores and no hyperthreading ?

 

[ian87w]

of course, thats also a a thing...but for PC industry...where you have a lot of customers who needs access to their components....is very hard....on mac users..for a decade we are used not to be able to add or remove ram, or dGpu or anything .....but for PC users...i dont know.
Apple is thinking to bring the storage also into the SOC....i dont thing half of the pc users wants everything "irreplaceable "

It depends. Those "PC" users you're talking about are the enthusiasts. The lay PC users couldn't care less. Heck, I've seen plenty of people buying a new PC just because the hard drive on their older PCs broke (the rest of the components were actually fine). Lay consumers don't understand RAM, replacing drives, PCI express, etc etc. This is why Apple is doing well with their all-in-ones, laptops, and iPhones. The only thing is Apple is operating in the premium segment, while a larger portion of the PC industry is actually on the lower end.

 

[DaveFromCampbelltown]

Apple has stuck with Intel for nearly 20 years now. The main reason for moving away is that Intel have not been able to make significant improvements in their CPUs for a few years now. Just as they moved from 68k to PowerPC, then to Intel, now they are moving to ARM.

As for Windows, it needs to be chucked out and rebuilt from scratch, just as Apple did when they moved from System 9 to OS X. It's just that Apple has a two-decade head start on Microsoft, which now has 20 more years of backwards compatibility to work around.

I remember in the mid 1980s Microsoft saying that they would migrate from MS-DOS to some form of Unix. They may finally be getting around to doing so.

 

[Broko Fankone]

Apple has stuck with Intel for nearly 20 years now. The main reason for moving away is that Intel have not been able to make significant improvements in their CPUs for a few years now. Just as they moved from 68k to PowerPC, then to Intel, now they are moving to ARM.

As for Windows, it needs to be chucked out and rebuilt from scratch, just as Apple did when they moved from System 9 to OS X. It's just that Apple has a two-decade head start on Microsoft, which now has 20 more years of backwards compatibility to work around.

I remember in the mid 1980s Microsoft saying that they would migrate from MS-DOS to some form of Unix. They may finally be getting around to doing so.

What exactly makes you think that Windows need to be "chucked out and rebuilt from scratch"? It's the most reliable and stable OS available. Oh, is it macOS? I guess if you don't count the hundreds of thousands of crashes/kernel panics on all models since 2016 then it's reliable, yeah! Funny how my windows 10 laptop had zero of these over two years and you know, never endangered my work to the point where my machine simply shuts down, loses unsaved progress and then takes a minute before it can even start again.

We are yet to see if the M1 machines will be plagued by such instability because 2016-2020 absolutely has been.

 

[vladi]

It doesn't work like that, it's not about x86 vs ARM.

M1 gains in CPU/GPU come from ASIC/FPGA like behavior and most of your speed on these new machines come from proper SSD that Apple has gimped on for a very long time. What SoC ASIC has are special instruction sets embedded into chip such as transcoding of certain video codec, just an example so everyone would understand. x86 doesn't do none of that although it has some broad special instructions and same goes for Qualcomm ARM implementation, it's just too broad in order to make it appealing and open to anyone who wants to use their chip for any purpose. nVidia on the other hand uses ARM on the same principle like Apple does, very specific and very narrow.

What Intel intends to do in the future is hybrid chipsets that contain both x86 and ASIC and FPGA chips depending on a purpose. They've already went to shopping spree buying out world leading research teams for both ASIC and FPGA. How well they can integrate them is up to them.

 

[leman]

Depending on what you mean by "catch up". In terms of absolute performance, M1 is not really much faster than Intel or AMD — it is simply able to deliver the same performance at a fraction of used power. There are claims that Apple only gets ahead here because they are using a more advanced process node, but I just don't see it. The difference between Apple and AMD right now is just 1 node (7nm to 5nm) but it takes Apple less than 5 watts to deliver the same performance as Zen 3 at 20 watts. Let's be honest, neither Intel nor AMD are going to reduce their power consumptions by over 70% just by adopting a better process — there is more going on here.

For us customers, the advantage of Apple Silicon is that we can have high-end performance in a form factor that would be usually reserved for low-end ultra-mobile. It has long been a "performance, battery life, compact form factor — pick two", with Apple Silicon you can have all three.

The x86 world is not going anywhere — it works well enough and there too much infrastructure in place that can't just be abandoned. But it stands to reason that Apple will keep the performance per watt crown for a while. Intel and AMD are not going to cross that huge gap overnight. They will need to basically start from scratch. And Apple has a ten year lead on them.

M1 gains in CPU/GPU come from ASIC/FPGA like behavior and most of your speed on these new machines come from proper SSD that Apple has gimped on for a very long time.

M1 outperforms any other CPU out there in compiler tests. What kind of "special instruction" would allow this kind of acceleration compiling code? Apple CPU microarchitecture is simply better, plain and simple. Yes, hardware transcoders and unified memory helps with productivity, but that is only one part of the equation. Apple currently has the best memory prefetch, best branch prediction, best general-purpose FP computation and by far best ILP behavior of any other microarchitecture on the market.

 

[Mitthrawnuruodo]

For some of the reasons why Apple is killing it at the moment, read the section on "MEMORY, AND THE ADVANTAGES OF A BESPOKE HARDWARE ARCHITECTURE" here: https://daringfireball.net/2020/11/the_m1_macs

It all begins with this:

https://twitter.com/i/web/status/1326238434235568128

 

[DaveFromCampbelltown]

What exactly makes you think that Windows need to be "chucked out and rebuilt from scratch"? It's the most reliable and stable OS available. Oh, is it macOS? I guess if you don't count the hundreds of thousands of crashes/kernel panics on all models since 2016 then it's reliable, yeah! Funny how my windows 10 laptop had zero of these over two years and you know, never endangered my work to the point where my machine simply shuts down, loses unsaved progress and then takes a minute before it can even start again.

We are yet to see if the M1 machines will be plagued by such instability because 2016-2020 absolutely has been.

From another post I made elsewhere in MacRumors --

I just find the Windows ecosystem old, out-of-date, complicated and clunky.

1. The OS is basically insecure, so a whole lot of processes are welded on top of it to try to make it more secure. These make installs and updates take a very long time. For example, on the same hardware, a minor update to Windows on an SSD can take 20 minutes or more, while a minor update on Linux on a HDD is less than 5 minutes and less than 10 minutes for MacOS on a Fusion drive. (I have done all three recently.)
2. The User Interface is unnecessarily complicated. On MacOS or Linux I can easily open up the main root volume in one step, while it takes at least 3 steps for Windows.
3. Networking is borked. I can easily scan my home network for shared drives on Mac and Linux, see them come up and connect to them. On Windows I have to know their IP address. On Mac and Linux I can see and connect to drives using SMB or AFP. Windows sees nothing. As said, I have to know the IP address and connect directly.
4. The Registry is a concept that was out-of-date in 2000. Two decades later we still have it.

Other platforms are making improvements in security and performance. Not Microsoft. Apple has introduced a whole new layer of security with its locked System Volume. Ubuntu has improved performance so much that running it from a HDD is quite acceptable, whereas Windows now needs a good SSD to be at least acceptable.

When Linux on a HDD is faster than Windows on a SSD, it shows there are serious problems. It's not just OS updates. It takes me much longer to process a LaTeX file under Windows than Linux or Mac, even when using the same software implementation of LaTeX. Installing packages like LibreOffice take 3 to 4 times longer on Windows than Linux or Mac, and similarly it takes at least twice as long to install MS Office on Windows than Mac. And this is on the same basic hardware, Intel CPU, 8 Gb RAM, SSD for Windows, HDD for Linux, Fusion drive for Mac.

Windows is not inherently stable. My wife works from home, telecommuting, running virtual sessions. Because it's a 'work' machine, I keep it clean. It has a fresh install of Windows 10, and only MS Office and Citrix installed. Nothing else. No games, no non-MS software, and everything is kept up to date. She still has to shut down and restart a couple of times a week. I recently retired from working in a Government Help desk office. We all ran Windows, it was all locked down, approved software only, and we still all had random seizures during the working week and had to re-start. And we weren't running cheapy hardware. It was all HP, Dell, Lenovo and more recently MS hardware. Linux machines running on the same hardware never skipped a beat.

I have been working with personal, mid- and main-frame operating systems since 1975. A locked-down implementation of Windows, with a limited range of approved software, may be suitable for medium to large enterprises and some Government use if you have a good IT Support team. It is no longer suitable for home use.

 

[Broko Fankone]

Do you want me to film restart speeds of both my MBP and win10 laptop and compare which is faster? If there is a "serious problem" I for sure have not experienced it. Your metrics make no sense considering how there is nothing called "basic SSD". Every SSD has specific read/write speeds. My windows 10 ssd is faster and thus it works and load faster than the one in the 2020 mbp.

 

[ArPe]

From another post I made elsewhere in MacRumors --


When Linux on a HDD is faster than Windows on a SSD, it shows there are serious problems. It's not just OS updates. It takes me much longer to process a LaTeX file under Windows than Linux or Mac, even when using the same software implementation of LaTeX. Installing packages like LibreOffice take 3 to 4 times longer on Windows than Linux or Mac, and similarly it takes at least twice as long to install MS Office on Windows than Mac. And this is on the same basic hardware, Intel CPU, 8 Gb RAM, SSD for Windows, HDD for Linux, Fusion drive for Mac.

I have been working with personal, mid- and main-frame operating systems since 1975. A locked-down implementation of Windows, with a limited range of approved software, is suitable for medium to large enterprises and some Government use. It is no longer suitable for home use.

The OS is basically insecure, so a whole lot of processes are welded on top of it to try to make it more secure. These make installs and updates take a very long time. For example, on the same hardware, a minor update to Windows on an SSD can take 20 minutes or more, while a minor update on Linux on a HDD is less than 5 minutes and less than 10 minutes for MacOS on a Fusion drive. (I have done all three recently.

Nah it’s safe and updates are regular and quick. Most people are more worried about the micro operating system embedded inside Intel chips.

The User Interface is unnecessarily complicated. On MacOS or Linux I can easily open up the main root volume in one step, while it takes at least 3 steps for Windows.

Just set up Windows Explorer so opens at the root. You can put the drives on the task bar. Almost everything can be modified with a couple of clicks.


Networking is borked. I can easily scan my home network for shared drives on Mac and Linux, see them come up and connect to them. On Windows I have to know their IP address. On Mac and Linux I can see and connect to drives using SMB or AFP. Windows sees nothing. As said, I have to know the IP address and connect directly.

It is a bit finicky but it’s faster at SMB than macOS.

The Registry is a concept that was out-of-date in 2000. Two decades later we still have it.

True but IT people use it to modify entries and strings to create custom settings and security for corporate networks.

 

[DaveFromCampbelltown]

Do you want me to film restart speeds of both my MBP and win10 laptop and compare which is faster? If there is a "serious problem" I for sure have not experienced it. Your metrics make no sense considering how there is nothing called "basic SSD". Every SSD has specific read/write speeds. My windows 10 ssd is faster and thus it works and load faster than the one in the 2020 mbp.

iMac
Boot Windows 10 2004 to Seagate Fast SSD - 32 seconds
Boot Mac Catalina 10.15.17 to Fusion drive - 25 seconds
Boot Ubuntu to Seagate Expansion SSD (slower than Seagate Fast) - 22 seconds.

Ubuntu 20.10 booting to my slowest SSD is faster than the other two OSs.
Windows booting to my fastest SSD is slower than the other two OSs.

PC
Boot Windows 10 2004 to Crucial SSD - 26 seconds
Boot Ubuntu 20.10 to Crucial SSD - 18 seconds.

When I use the same hardware, Windows is slower than Ubuntu or MacOS,
When I give Windows the faster SSD, it is still slower.

 

[Broko Fankone]

And how does that relate to the problem or reliability that I discussed? My windows boots in ~10 seconds so I honestly can't see a single problem of any possible kind related to boot speeds with it. And booting ubuntu - who cares, exactly? It is a well-known fact that UNIX systems are lighter than windows. Again, who cares if your hardware is fast and everything boots and loads quickly on windows? I've had zero crashes of the whole system in 2 years and with my mac I've had several absolutely random kernel panics in 7 months that shut down my laptop without warning and without explanation. That is unacceptable. Booting a few seconds slower is acceptable. Random shutdowns are NOT acceptable. And honestly, if M1 macs turn out to have the same problem they will be just as unacceptable for anyone who expects, you know, actual reliability and needs to depend on their system to work and not wipe out their work at random.

If M1 macs prove to have solved this issue entirely, then hats off to them. But this remains to be seen.

 

[Bearbrickster]

Depending on what you mean by "catch up". In terms of absolute performance, M1 is not really much faster than Intel or AMD — it is simply able to deliver the same performance at a fraction of used power. There are claims that Apple only gets ahead here because they are using a more advanced process node, but I just don't see it. The difference between Apple and AMD right now is just 1 node (7nm to 5nm) but it takes Apple less than 5 watts to deliver the same performance as Zen 3 at 20 watts. Let's be honest, neither Intel nor AMD are going to reduce their power consumptions by over 70% just by adopting a better process — there is more going on here.

For us customers, the advantage of Apple Silicon is that we can have high-end performance in a form factor that would be usually reserved for low-end ultra-mobile. It has long been a "performance, battery life, compact form factor — pick two", with Apple Silicon you can have all three.

The x86 world is not going anywhere — it works well enough and there too much infrastructure in place that can't just be abandoned. But it stands to reason that Apple will keep the performance per watt crown for a while. Intel and AMD are not going to cross that huge gap overnight. They will need to basically start from scratch. And Apple has a ten year lead on them.



M1 outperforms any other CPU out there in compiler tests. What kind of "special instruction" would allow this kind of acceleration compiling code? Apple CPU microarchitecture is simply better, plain and simple. Yes, hardware transcoders and unified memory helps with productivity, but that is only one part of the equation. Apple currently has the best memory prefetch, best branch prediction, best general-purpose FP computation and by far best ILP behavior of any other microarchitecture on the market.

Consumers benefit the most with this powerful M1. We are getting higher performance machines with better value. The new 16 mbp or imac might be able to match the performance of 2019 mac pro high spec. That will represent tremendous savings for customers. If that happens, I can see many professional windows user switch to the mac. Apple is getting around 10% (i dont know the exact number) of the computer market? Maybe with this big performance gain and value, apple might double their market share to 20%in the next few years.... that might get windows to finally make some major changes to the software side or even work closer with a chip producer for the hardware side... time will tell....

 

[Mcdevidr]

Intels new 11th gen is pretty speedy too. The 1165g7 beats most of their desktop chips in single and dual core tests and is comparable in even up to quad core (also beats M1 in cinebench). This is at 28watts so performance per watt isn’t same as M1 but still quite strong for a mobile chip.

 

[ian87w]

Intels new 11th gen is pretty speedy too. The 1165g7 beats most of their desktop chips in single and dual core tests and is comparable in even up to quad core (also beats M1 in cinebench). This is at 28watts so performance per watt isn’t same as M1 but still quite strong for a mobile chip.

It will be interesting what Apple is cooking for the 28W TDP M chip.

Meanwhile, many OEMs are still selling laptops with 8th gen intel chips. It’s really annoying