6 ms·

I think it's a legitimate question to ask how long Apple's ARM advantage continues. I could totally see a repeat of Apple's PowerPC experience, where they do th

I think it's a legitimate question to ask how long Apple's ARM advantage continues. I could totally see a repeat of Apple's PowerPC experience, where they do the transition, and just smoke the competition for a generation or two, and then get dragged under by the market effects of the x86 duopoly. The next generation Ryzen series (7000 series) looks like it's going to be absolutely insane. That window may end up being shorter then expected.

ajross 4y ago
It's not market effects from x86, it's just process. We've entered a period where (1) Apple is able to command higher margins for its silicon and thus pay for "The Best" of whatever is available, even monopolizing the best processes and (2) "The Best" process isn't Intel, so it's available for purchase. So Apple bought themselves a win, which makes perfect sense. And they had already chosen ARM for iOS (for historical reasons, really -- the original iPhone picked an off-the-shelf Samsung SOC), so that's what they dropped in the Mac. But indeed, this is probably transitory. Eventually TSMC will either lose its lead or ramp its best processes to the point where other manufacturers can buy it. Already, as we see here, they're only mildly ahead of the rest of the pack.
- enragedcacti 4y ago
  I don't think it's fair to say that it's just process, Apple has invested very heavily in their core design and their perf/w + total perf is still unmatched by companies like AMD or Qualcomm on the same process node. It's certainly not a one-horse race but Apple is definitely has world class CPU design capabilities regardless of node.
  - ajross 4y ago
    The "same process node" isn't correct. Nothing from AMD is available in TSMC 5nm yet, and the Snapdragon 888 is on Samsung's clearly struggling 5nm node. Apple is alone at the top of the process ladder right now, and they're getting pretty much exactly the performance[1] benefit you'd expect from that. Certainly the M2 is not a bad SOC, and there's no reason it shouldn't be and remain competetive with everyone else's offerings as the industry evolves. But I do think there's a certain amount of kool aid being drunk by the Mac community about these chips. They're great. But maybe not insanely so. [1] It's absolutely true that Apple's whole-system power integration story remains at the top of the industry. But that was true even when they weren't on the best process.
    - enragedcacti 4y ago
      > The "same process node" isn't correct. Nothing from AMD is available in TSMC 5nm yet, and the Snapdragon 888 is on Samsung's clearly struggling 5nm node. Sorry you are totally right and my phrasing was poor, what I was trying to say is that Apple's TSMC 7nm chips (A13) are still very competitive with Qualcomm and AMDs offerings on the same node. They beat out AMD (5000 series) on perf/w while tying perf and beat Qualcomm (865+) on perf handily with (iirc) similar wattage. M1 and M2 are certainly benefitting from TSMC 5nm compared to everyone else currently and AMD in particular is a lot closer than I think many Apple fans give credit.
      - JAlexoid 4y ago
        Perfomance/watt is a poor comparison. A low power ARM MCU will kill anything in perf/watt. Performance gains aren't linear.
dangus 4y ago
The difference here is that Apple makes more money by using their own silicon regardless of whether or not they have a technical advantage. Apple's systems may very well end up being slower. From a business standpoint, I am not sure that will matter. Neither AMD nor Intel have shown much indication of catching up to Apple's (well, ARM's) mobile device battery life advantage. Apple doesn't sell systems that go in data centers, they don't need to smoke anything.
- Lio 4y ago
  I would think the real difference this time is the existence of the iPhone and iPad markets to help scale the costs of Apple Silicon development. When Apple was using PowerPC the number of PowerPC units shipped was tiny compared to the PC market. The huge volumes of mobile devices that Apple now ship has got to at least help share R&D costs across product lines.
  - jagger27 4y ago
    I agree. I don't think the broader Arm ecosystem is comparable to the Apple-IBM-Motorola alliance of the Power era. Arm is ubiquitous in ways that Power never was. Even if Apple had to start buying chips from Qualcomm (yuck) they'd be okay.
- smoldesu 4y ago
  This is definitely the key takeaway, here. Macs will always be a different kind of computer for a different class of consumer; people will either love them or hate them, regardless of processor speed or battery life. A Macbook is a Macbook, and if you want the benefits then you have to also live with the downsides. Consumers already know this. They'll buy the new Mac even if it's worse (see: 2016), the new iPhone even if it's slower (see: 2022), and the new Magic Mouse even if it's still broken. It doesn't matter how fast or slow the Mac is, Apple has long proven that they exist in a market segment of their own, for better or worse.
- astrange 4y ago
  The aim of Apple Silicon isn't to be "faster", it's to have a high perf/power ratio ie be fast and have high battery life.
klelatti 4y ago
> Market effects of the x86 duopoly. Essentially this was because PC volumes supported Intels R&D and investment being greater than anyone else could afford. We’re no longer in 2000 though. Apple sold 240 million iPhones last year which is not far off global PC sales. Meanwhile TSMC has overtaken Intel with a broader and growing product base whilst Intel is shrinking.
cmrdporcupine 4y ago
The PowerPC problem for Apple was that IBM was not interested in advancing the architecture in the personal computer market, and Freescale had left the market as well. I don't think there was anything intrinsic to the PowerPC ISA itself that made it a hot and slow chip. It was simply the lack of R&D and commitment to fab on new process nodes, etc. The switch to Intel was necessary because they literally had no quality supplier for chips and at that point x86 was practically the only supplier for desktop computer MPUs. Note that their portable / consumer devices have always been ARM. Like, back to the Newton even. Apple now makes their own chips. They've dealt with that vulnerability. And they're fully onboard the ARM bus in a big way and they've managed to bring their customer base with them without a lot of hassle. Over the last 40 years Apple has been burned multiple times with having their key ISA be dropped by its manufacturer. 6502 became a deadend (quickly.) Motorola dropped the ball on the (relatively successful) 68k, and tried to get people onto the ill-fated 88k and then pulled the eject cord and everyone went PowerPC. That lasted maybe a decade and Motorola/Freescale&IBM started to flee the ship. I think Apple has learned their lesson. They don't want to be vulnerable on this front anymore.
- solarmist 4y ago
  What is an MPU?
  - cmrdporcupine 4y ago
    microprocessor unit vs MCU microcontroller
    - solarmist 4y ago
      How is that different than a CPU?
      - wmf 4y ago
        It's not. MPU = CPU
      - cmrdporcupine 4y ago
olliej 4y ago
The core problem (from an Apple business perspective) is that they were using chips that were a group project by ibm and Motorola, where IBM made money primarily off the server class Power chips, and I’m not sure if Motorola ever made money? That meant that the performance requirements of a laptop were simply never going to be IBM or Motorola’s priority. The last PowerPC Mac was the G5 tower, and that was an amazing case with incredible cooling setup because it needed it. The G5 used a ton of power, and generated insane amounts of heat (seriously those G5 heat sinks were larger than most pc gamer/modder heat sinks if there era, before considering the enormous airflow, and even a flow shroud. Simply put there was no way a G5 was ever going to be in a laptop that was actually portable. That’s what led to Apple migrating to intel - intel at least was interested in consumer hardware, and portable computing as well. So the important thing is that Apple’s own cpus are presumably always going to be aligned with what Apple wants, in a way that they can’t rely on others to do.
pacetherace 4y ago
I feel that we are at or already crossed the point where we don't need much more powerful CPUs. This is especially true for majority of Apple's customer base lies. The only game in town that Intel/AMD can really play now is to make their chips cheaper and more palatable (reduce power consumption, offer better integrated GPUs, etc) for x86 based consumer products or slowly see Apple's share in the laptop market increase more and more.
kristianp 4y ago
Apple is getting smaller lithography from TSMC than AMD chips are, so they'll probably have an advantage in their M3 or M2 pro/max by the time Ryzen 7000 comes out.

ajross 4y ago

It's not market effects from x86, it's just process. We've entered a period where (1) Apple is able to command higher margins for its silicon and thus pay for "The Best" of whatever is available, even monopolizing the best processes and (2) "The Best" process isn't Intel, so it's available for purchase. So Apple bought themselves a win, which makes perfect sense. And they had already chosen ARM for iOS (for historical reasons, really -- the original iPhone picked an off-the-shelf Samsung SOC), so that's what they dropped in the Mac. But indeed, this is probably transitory. Eventually TSMC will either lose its lead or ramp its best processes to the point where other manufacturers can buy it. Already, as we see here, they're only mildly ahead of the rest of the pack.

enragedcacti 4y ago
I don't think it's fair to say that it's just process, Apple has invested very heavily in their core design and their perf/w + total perf is still unmatched by companies like AMD or Qualcomm on the same process node. It's certainly not a one-horse race but Apple is definitely has world class CPU design capabilities regardless of node.
- ajross 4y ago
  The "same process node" isn't correct. Nothing from AMD is available in TSMC 5nm yet, and the Snapdragon 888 is on Samsung's clearly struggling 5nm node. Apple is alone at the top of the process ladder right now, and they're getting pretty much exactly the performance[1] benefit you'd expect from that. Certainly the M2 is not a bad SOC, and there's no reason it shouldn't be and remain competetive with everyone else's offerings as the industry evolves. But I do think there's a certain amount of kool aid being drunk by the Mac community about these chips. They're great. But maybe not insanely so. [1] It's absolutely true that Apple's whole-system power integration story remains at the top of the industry. But that was true even when they weren't on the best process.
  - enragedcacti 4y ago
    > The "same process node" isn't correct. Nothing from AMD is available in TSMC 5nm yet, and the Snapdragon 888 is on Samsung's clearly struggling 5nm node. Sorry you are totally right and my phrasing was poor, what I was trying to say is that Apple's TSMC 7nm chips (A13) are still very competitive with Qualcomm and AMDs offerings on the same node. They beat out AMD (5000 series) on perf/w while tying perf and beat Qualcomm (865+) on perf handily with (iirc) similar wattage. M1 and M2 are certainly benefitting from TSMC 5nm compared to everyone else currently and AMD in particular is a lot closer than I think many Apple fans give credit.
    - JAlexoid 4y ago
      Perfomance/watt is a poor comparison. A low power ARM MCU will kill anything in perf/watt. Performance gains aren't linear.

dangus 4y ago

The difference here is that Apple makes more money by using their own silicon regardless of whether or not they have a technical advantage. Apple's systems may very well end up being slower. From a business standpoint, I am not sure that will matter. Neither AMD nor Intel have shown much indication of catching up to Apple's (well, ARM's) mobile device battery life advantage. Apple doesn't sell systems that go in data centers, they don't need to smoke anything.

Lio 4y ago
I would think the real difference this time is the existence of the iPhone and iPad markets to help scale the costs of Apple Silicon development. When Apple was using PowerPC the number of PowerPC units shipped was tiny compared to the PC market. The huge volumes of mobile devices that Apple now ship has got to at least help share R&D costs across product lines.
- jagger27 4y ago
  I agree. I don't think the broader Arm ecosystem is comparable to the Apple-IBM-Motorola alliance of the Power era. Arm is ubiquitous in ways that Power never was. Even if Apple had to start buying chips from Qualcomm (yuck) they'd be okay.
smoldesu 4y ago
This is definitely the key takeaway, here. Macs will always be a different kind of computer for a different class of consumer; people will either love them or hate them, regardless of processor speed or battery life. A Macbook is a Macbook, and if you want the benefits then you have to also live with the downsides. Consumers already know this. They'll buy the new Mac even if it's worse (see: 2016), the new iPhone even if it's slower (see: 2022), and the new Magic Mouse even if it's still broken. It doesn't matter how fast or slow the Mac is, Apple has long proven that they exist in a market segment of their own, for better or worse.
astrange 4y ago
The aim of Apple Silicon isn't to be "faster", it's to have a high perf/power ratio ie be fast and have high battery life.

klelatti 4y ago

> Market effects of the x86 duopoly. Essentially this was because PC volumes supported Intels R&D and investment being greater than anyone else could afford. We’re no longer in 2000 though. Apple sold 240 million iPhones last year which is not far off global PC sales. Meanwhile TSMC has overtaken Intel with a broader and growing product base whilst Intel is shrinking.

cmrdporcupine 4y ago

The PowerPC problem for Apple was that IBM was not interested in advancing the architecture in the personal computer market, and Freescale had left the market as well. I don't think there was anything intrinsic to the PowerPC ISA itself that made it a hot and slow chip. It was simply the lack of R&D and commitment to fab on new process nodes, etc. The switch to Intel was necessary because they literally had no quality supplier for chips and at that point x86 was practically the only supplier for desktop computer MPUs. Note that their portable / consumer devices have always been ARM. Like, back to the Newton even. Apple now makes their own chips. They've dealt with that vulnerability. And they're fully onboard the ARM bus in a big way and they've managed to bring their customer base with them without a lot of hassle. Over the last 40 years Apple has been burned multiple times with having their key ISA be dropped by its manufacturer. 6502 became a deadend (quickly.) Motorola dropped the ball on the (relatively successful) 68k, and tried to get people onto the ill-fated 88k and then pulled the eject cord and everyone went PowerPC. That lasted maybe a decade and Motorola/Freescale&IBM started to flee the ship. I think Apple has learned their lesson. They don't want to be vulnerable on this front anymore.

solarmist 4y ago
What is an MPU?
- cmrdporcupine 4y ago
  microprocessor unit vs MCU microcontroller
  - solarmist 4y ago
    How is that different than a CPU?
    - wmf 4y ago
      It's not. MPU = CPU
    - cmrdporcupine 4y ago

olliej 4y ago

The core problem (from an Apple business perspective) is that they were using chips that were a group project by ibm and Motorola, where IBM made money primarily off the server class Power chips, and I’m not sure if Motorola ever made money? That meant that the performance requirements of a laptop were simply never going to be IBM or Motorola’s priority. The last PowerPC Mac was the G5 tower, and that was an amazing case with incredible cooling setup because it needed it. The G5 used a ton of power, and generated insane amounts of heat (seriously those G5 heat sinks were larger than most pc gamer/modder heat sinks if there era, before considering the enormous airflow, and even a flow shroud. Simply put there was no way a G5 was ever going to be in a laptop that was actually portable. That’s what led to Apple migrating to intel - intel at least was interested in consumer hardware, and portable computing as well. So the important thing is that Apple’s own cpus are presumably always going to be aligned with what Apple wants, in a way that they can’t rely on others to do.

pacetherace 4y ago

I feel that we are at or already crossed the point where we don't need much more powerful CPUs. This is especially true for majority of Apple's customer base lies. The only game in town that Intel/AMD can really play now is to make their chips cheaper and more palatable (reduce power consumption, offer better integrated GPUs, etc) for x86 based consumer products or slowly see Apple's share in the laptop market increase more and more.

kristianp 4y ago

Apple is getting smaller lithography from TSMC than AMD chips are, so they'll probably have an advantage in their M3 or M2 pro/max by the time Ryzen 7000 comes out.