8 ms·

If you remember back to the mid 2000's when Apple switched from PowerPC to (Intel) x86, the higher end machines still ran with PowerPC chips in them while the m

If you remember back to the mid 2000's when Apple switched from PowerPC to (Intel) x86, the higher end machines still ran with PowerPC chips in them while the mid-low end products had x86 chips. All of the old applications that weren't yet ported to x86 had to use "Rosetta." It wasn't until a little while later, presumably once most of the professional software companies had good working x86 versions of their software - that the high end mac pro systems started seeing Intel chips in them. Microsoft started down that path last year with Windows 10 on ARM, and Intel had some legal issues with that - https://arstechnica.com/information-technology/2017/06/intel-fires-warning-shots-at-microsoft-claims-x86-emulation-is-a-patent-minefield/ https://arstechnica.com/information-technology/2017/06/intel... It will be interesting to see how Apple handles this.

apetrovic 9y ago
There's two problems with that story: first, the famous switch from PowerPC to Intel happened when the desktop was the king, and app developers had strong incentive to recompile programs to the new platform. Right now? Not so much. Second, while Apple's mobile processors are currently the best in ARM world, for "pro" desktop computers Apple needs desktop-grade processor. Making something comparable to i7 isn't such an easy and cheap job.
- collyw 9y ago
  > the famous switch from PowerPC to Intel happened when the desktop was the king, I though one of the major reasons for the switch to Intel was the power consumption and performance on laptops.
  - emodendroket 9y ago
    I'm thinking that in this case "desktop" is cast in opposition to tablets and cell phones and laptop computers can be considered "desktop" devices.
- Lramseyer 9y ago
  > Second, while Apple's mobile processors are currently the best in ARM world, for "pro" desktop computers Apple needs desktop-grade processor. Making something comparable to i7 isn't such an easy and cheap job. Qualcomm, NVIDIA, and Cavium have created very powerful ARM chips for non-mobile application. Apple definitely has the chip architects, the money, and the experience to do such a thing.
  - lev99 9y ago
    I have serious doubts that the marketplace can support 3 desktop/laptop workstation CPU manufacturers. (Amd, Intel, and now Apple)
    - oneplane 9y ago
      I don't think it matters to Apple since they will be using their own chips. Intel will simply sell less, and AMD will not change at all. By the way, don't forget Broadcom, Marvell, Cavium, IBM etc. as they are doing big bucks with CPU's and SoC's as well.
      - lev99 9y ago
        It matters to Apple, CPUs are a very big expense for them, and the performance of their computers are very important to them. I'm specifically worried about the performance workstation/gaming cpu market, which Broadcom Marvell, Cavium, and IBM are not major players in. It's currently AMD and Intel. Apple would need to sell a lot of macs with the new CPUs to make the R&D cost worth it and the new CPUs would have to be very fast or they won't sell macs.
        oneplane 9y ago
        I meant the impact on global CPU markets won't matter to Apple, they don't care if Intel makes less money.
- tambourine_man 9y ago
  Adobe for one is making a boatload of money from CC subscriptions, record numbers, so I wouldn't worry too much about that (I imagine MS is equally happy with Office 365 on the desktop). Besides, there's no Carbon to Cocoa or CodeWarrior to Xcode port, so aside from a very small percentage of optimized assembly code, it shouldn't be nearly as hard.
- megy 9y ago
  >Second, while Apple's mobile processors are currently the best in ARM world, for "pro" desktop computers Apple needs desktop-grade processor. Making something comparable to i7 isn't such an easy and cheap job. Good point, I guess Apple never though of that.
  - freehunter 9y ago
    It's quite impressive how many people in the comments can so easily come up with big show-stoppers that Apple never thought about. Must be why Apple is struggling financially and can't get anyone to buy their products.
    - eklavya 9y ago
      While being sarcastic you might have missed the sarcasm of the parent :)
      - freehunter 9y ago
        I both understood and was adding to the sarcasm.
- cerved 9y ago
  Also worth noting is Windows and x86 desktop dominance at the time likely made porting Mac software to x86 easier
danellis 9y ago
I remember all the way back to the early 90s and the original x86 emulator running on ARM. Admittedly that was before optimizations like JIT, and a lot of research has gone into binary translation since then, but it was a painful experience at the time. I wonder how much performance would be lost these days.
- tom_ 9y ago
  If they’re making their own CPUs, they can presumably add whatever additional functionality and/or instructions are necessary to make emulation or JIT translation work effectively. But even if flat-out performance isn’t as good, what about if it proves cheap and easy to scale to more cores? Like, what if you get twice as many cores as the Intel equivalent has threads, albeit at 2/3 the speed (maybe worse or better for some workloads) but no hyperthreading-style contention? This wouldn’t be useless. And GPU-accelerated functionality will be mostly unaffected.
tootie 9y ago
Couldn't they achieve that today by running legacy apps in a hypervisor?
- BillinghamJ 9y ago
  The issue is implementing the features for the hypervisor’s emulation system without Intel-licensed hardware or software.
  - andreiw 9y ago
    lol wut? VMware Workstation, Connectix/MS VirtualPC, Oracle VirtualBox, qemu, bochs...what Intel-licensed hardware or software? How about FX!32 or it’s latest AArch64 cousin seen on Windows 10? Of the above, qemu can target x86 and run on anything, Virtual PC ran originally on PowerPC, and FX!32 ran on the Alpha
    - monocasa 9y ago
      https://arstechnica.com/information-technology/2017/06/intel-fires-warning-shots-at-microsoft-claims-x86-emulation-is-a-patent-minefield/ https://arstechnica.com/information-technology/2017/06/intel...
      - mozumder 9y ago
        You know, the article refers to Windows ARM emulation, but I now see the Intel post as a warning to Apple, as patent licensing discussions may have started about that time last year. It usually takes 4-5 years to design a chip, and if Apple planned on releasing the Mac CPU in 2020, well...
      - andreiw 9y ago
        I see an article that is not congruent with actual tech facts. They can warn all they want, because the precedent says otherwise. Even if you wanted to make the argument that Intel is against emulation on non-x86 systems, that’s not true - see Virtual PC for Mac, FX!32, qemu, bochs and countless 8086 emulators to run VGA ROMs on various legacy free systems (either in OS or firmware). If you wanted to make the case that efficient emulation on non-x86 is is disallowed, well, FX!32, qemu and Virtual PC all used and use JIT. Intel’s position amounts to cage rattling and vague insinuations that efficient execution of x86 code is impossible without hardware support. That’s nice, and their ham-fisted approach already hurt customer choice in the past (Transmeta and nVidia Denver, before the later implemented the Arm ISA) but the Windows on Arm solution is completely software and not coupled to any specific Arm chip implementation.
DerekL 9y ago
> It wasn't until a little while later, presumably once most of the professional software companies had good working x86 versions of their software - that the high end mac pro systems started seeing Intel chips in them. The switch to Intel started six months [correction: seven months] from the announcement, and took seven months to complete. June 5th, 2005: Intel switch announced at WWDC January 10th, 2006: First Intel Macs (MacBook Pro and iMac) released August 7th, 2006: Mac Pro and Intel Xserve released, last PowerPC Macs discontinued https://everymac.com/systems/by_timeline/index-macs-by-timeline.html https://everymac.com/systems/by_timeline/index-macs-by-timel...

apetrovic 9y ago

There's two problems with that story: first, the famous switch from PowerPC to Intel happened when the desktop was the king, and app developers had strong incentive to recompile programs to the new platform. Right now? Not so much. Second, while Apple's mobile processors are currently the best in ARM world, for "pro" desktop computers Apple needs desktop-grade processor. Making something comparable to i7 isn't such an easy and cheap job.

collyw 9y ago
> the famous switch from PowerPC to Intel happened when the desktop was the king, I though one of the major reasons for the switch to Intel was the power consumption and performance on laptops.
- emodendroket 9y ago
  I'm thinking that in this case "desktop" is cast in opposition to tablets and cell phones and laptop computers can be considered "desktop" devices.
Lramseyer 9y ago
> Second, while Apple's mobile processors are currently the best in ARM world, for "pro" desktop computers Apple needs desktop-grade processor. Making something comparable to i7 isn't such an easy and cheap job. Qualcomm, NVIDIA, and Cavium have created very powerful ARM chips for non-mobile application. Apple definitely has the chip architects, the money, and the experience to do such a thing.
- lev99 9y ago
  I have serious doubts that the marketplace can support 3 desktop/laptop workstation CPU manufacturers. (Amd, Intel, and now Apple)
  - oneplane 9y ago
    I don't think it matters to Apple since they will be using their own chips. Intel will simply sell less, and AMD will not change at all. By the way, don't forget Broadcom, Marvell, Cavium, IBM etc. as they are doing big bucks with CPU's and SoC's as well.
    - lev99 9y ago
      It matters to Apple, CPUs are a very big expense for them, and the performance of their computers are very important to them. I'm specifically worried about the performance workstation/gaming cpu market, which Broadcom Marvell, Cavium, and IBM are not major players in. It's currently AMD and Intel. Apple would need to sell a lot of macs with the new CPUs to make the R&D cost worth it and the new CPUs would have to be very fast or they won't sell macs.
      - oneplane 9y ago
        I meant the impact on global CPU markets won't matter to Apple, they don't care if Intel makes less money.
tambourine_man 9y ago
Adobe for one is making a boatload of money from CC subscriptions, record numbers, so I wouldn't worry too much about that (I imagine MS is equally happy with Office 365 on the desktop). Besides, there's no Carbon to Cocoa or CodeWarrior to Xcode port, so aside from a very small percentage of optimized assembly code, it shouldn't be nearly as hard.
megy 9y ago
>Second, while Apple's mobile processors are currently the best in ARM world, for "pro" desktop computers Apple needs desktop-grade processor. Making something comparable to i7 isn't such an easy and cheap job. Good point, I guess Apple never though of that.
- freehunter 9y ago
  It's quite impressive how many people in the comments can so easily come up with big show-stoppers that Apple never thought about. Must be why Apple is struggling financially and can't get anyone to buy their products.
  - eklavya 9y ago
    While being sarcastic you might have missed the sarcasm of the parent :)
    - freehunter 9y ago
      I both understood and was adding to the sarcasm.
cerved 9y ago
Also worth noting is Windows and x86 desktop dominance at the time likely made porting Mac software to x86 easier

danellis 9y ago

I remember all the way back to the early 90s and the original x86 emulator running on ARM. Admittedly that was before optimizations like JIT, and a lot of research has gone into binary translation since then, but it was a painful experience at the time. I wonder how much performance would be lost these days.

tom_ 9y ago
If they’re making their own CPUs, they can presumably add whatever additional functionality and/or instructions are necessary to make emulation or JIT translation work effectively. But even if flat-out performance isn’t as good, what about if it proves cheap and easy to scale to more cores? Like, what if you get twice as many cores as the Intel equivalent has threads, albeit at 2/3 the speed (maybe worse or better for some workloads) but no hyperthreading-style contention? This wouldn’t be useless. And GPU-accelerated functionality will be mostly unaffected.

tootie 9y ago

Couldn't they achieve that today by running legacy apps in a hypervisor?

BillinghamJ 9y ago
The issue is implementing the features for the hypervisor’s emulation system without Intel-licensed hardware or software.
- andreiw 9y ago
  lol wut? VMware Workstation, Connectix/MS VirtualPC, Oracle VirtualBox, qemu, bochs...what Intel-licensed hardware or software? How about FX!32 or it’s latest AArch64 cousin seen on Windows 10? Of the above, qemu can target x86 and run on anything, Virtual PC ran originally on PowerPC, and FX!32 ran on the Alpha
  - monocasa 9y ago
    https://arstechnica.com/information-technology/2017/06/intel-fires-warning-shots-at-microsoft-claims-x86-emulation-is-a-patent-minefield/ https://arstechnica.com/information-technology/2017/06/intel...
    - mozumder 9y ago
      You know, the article refers to Windows ARM emulation, but I now see the Intel post as a warning to Apple, as patent licensing discussions may have started about that time last year. It usually takes 4-5 years to design a chip, and if Apple planned on releasing the Mac CPU in 2020, well...
    - andreiw 9y ago
      I see an article that is not congruent with actual tech facts. They can warn all they want, because the precedent says otherwise. Even if you wanted to make the argument that Intel is against emulation on non-x86 systems, that’s not true - see Virtual PC for Mac, FX!32, qemu, bochs and countless 8086 emulators to run VGA ROMs on various legacy free systems (either in OS or firmware). If you wanted to make the case that efficient emulation on non-x86 is is disallowed, well, FX!32, qemu and Virtual PC all used and use JIT. Intel’s position amounts to cage rattling and vague insinuations that efficient execution of x86 code is impossible without hardware support. That’s nice, and their ham-fisted approach already hurt customer choice in the past (Transmeta and nVidia Denver, before the later implemented the Arm ISA) but the Windows on Arm solution is completely software and not coupled to any specific Arm chip implementation.

DerekL 9y ago

> It wasn't until a little while later, presumably once most of the professional software companies had good working x86 versions of their software - that the high end mac pro systems started seeing Intel chips in them. The switch to Intel started six months [correction: seven months] from the announcement, and took seven months to complete. June 5th, 2005: Intel switch announced at WWDC January 10th, 2006: First Intel Macs (MacBook Pro and iMac) released August 7th, 2006: Mac Pro and Intel Xserve released, last PowerPC Macs discontinued https://everymac.com/systems/by_timeline/index-macs-by-timeline.html https://everymac.com/systems/by_timeline/index-macs-by-timel...