Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Why doesn’t Apple use their own chips?
- Thread starter Pangalactic
- Start date
- Sort by reaction score
It's 2018 and idiots are still doing the "X86 chips are at least 10x faster regardless" despite ALL benchmarks and applications indicating otherwise?
People have been saying that X86 chips are 100x faster since ARM chips took off, and now ARM chips are really over 100X faster while the X86 chips have only gotten maybe 2x as fast, yet people are still claiming that X86 is faster?
You know what Geekbench isn't biased for Apple or ARM at ll, if anything it only measures peak performance which doesn't take into account of sustained performance and thermals. So if you just take a look at Geekbench, yes the A12X is just as fast as a 8th-gen quad-core i7, with the GPU being about as fast as Intel's best integrated graphics as well.
In terms of sustained performance, obviously the iPad will not be as strong as the MacBook Pro because it's not got a fan in it. But yes in theory if you just stick a heatsink with a fan on the A12X it'll match the 13" MacBook Pro in almost every performance metric.
Sure there are some caveats to ARM performance, mostly with complex floating point and vector operations but that's completely irrelevant with any modern, consumer applications. Now whenever we need floating point performance, we use the GPU, so the fact that the i7 may be some 4x faster in some niche floating point application is just pointless, nobody cares. In fact it is precisely because x86's continued support of so many of these niche instruction sets, they are more difficult to optimize for common applications.
If you don't believe the benchmarks, go ahead and run the applications, I don't know about you but my experience with Lightroom CC on the iPad have always matched the benchmark results. The iPad Pro develops RAW images just as fast as mobile Intel i7 chips on laptops.
Why isn't Apple using their own chips on laptops? I'd predict that they will do just that with the next Macbook 12". They haven't done so because traditional "computer" applications are coded for X86, Microsoft has already released an ARM version of Windows that can emulate x86 software, and we have seen that emulation makes things 3x slower than normal, and the Snapdragon chips are already 3x slower than competing Apple high-performance chips so the performance is really abysmal with those ARM Windows devices. However, since the A12X is so fast, it is now completely feasible to do a Macbook 12 with this chip and also run lite traditional X86 applications in emulation with an acceptable performance while also having a very strong battery life.
People have been saying that X86 chips are 100x faster since ARM chips took off, and now ARM chips are really over 100X faster while the X86 chips have only gotten maybe 2x as fast, yet people are still claiming that X86 is faster?
You know what Geekbench isn't biased for Apple or ARM at ll, if anything it only measures peak performance which doesn't take into account of sustained performance and thermals. So if you just take a look at Geekbench, yes the A12X is just as fast as a 8th-gen quad-core i7, with the GPU being about as fast as Intel's best integrated graphics as well.
In terms of sustained performance, obviously the iPad will not be as strong as the MacBook Pro because it's not got a fan in it. But yes in theory if you just stick a heatsink with a fan on the A12X it'll match the 13" MacBook Pro in almost every performance metric.
Sure there are some caveats to ARM performance, mostly with complex floating point and vector operations but that's completely irrelevant with any modern, consumer applications. Now whenever we need floating point performance, we use the GPU, so the fact that the i7 may be some 4x faster in some niche floating point application is just pointless, nobody cares. In fact it is precisely because x86's continued support of so many of these niche instruction sets, they are more difficult to optimize for common applications.
If you don't believe the benchmarks, go ahead and run the applications, I don't know about you but my experience with Lightroom CC on the iPad have always matched the benchmark results. The iPad Pro develops RAW images just as fast as mobile Intel i7 chips on laptops.
Why isn't Apple using their own chips on laptops? I'd predict that they will do just that with the next Macbook 12". They haven't done so because traditional "computer" applications are coded for X86, Microsoft has already released an ARM version of Windows that can emulate x86 software, and we have seen that emulation makes things 3x slower than normal, and the Snapdragon chips are already 3x slower than competing Apple high-performance chips so the performance is really abysmal with those ARM Windows devices. However, since the A12X is so fast, it is now completely feasible to do a Macbook 12 with this chip and also run lite traditional X86 applications in emulation with an acceptable performance while also having a very strong battery life.
It would make a lot of sense for Apple to have ARM chips in some of their Mac lineup. A Mac mini ARM and MacBook ARM should be perfectly possible already. For developers, to get them to start porting apps. Performance will get there, no problem. But there are a couple of issues, and they're not really technical in nature:
There would be a delay for Mac apps to get moved to ARM. Some apps wouldn't get moved at all because developers went out of business, or moved on to other apps. Some users wouldn't be upgrading even if apps were moved. There is a massive logistics project to make the switch. They've done it before, and sure it can be done. But the world is different now. It's no certainty that they can draw too much on how things went the last time. They could potentially end up with significantly lower sales figures, and I think they have to have a very strong plan to keep users happy this time.
For myself, if I couldn't dual boot into Windows and run existing X86 apps, there's no way I could justify the extra cost of a MBP. If there exists a Windows 10 for ARM is completely irrelevant, because it doesn't have the apps. Until they can make the switch and have emulated x86 apps be as fast as on native Intel processors, it just won't be a win for me. And if I have nothing to gain from the switch, then why would I want it?
Getting the hardware, the performance, and the OS on ARM, that's the easy part. I have no doubt that they have that already. Getting the apps on it too, that's the challenge.
There would be a delay for Mac apps to get moved to ARM. Some apps wouldn't get moved at all because developers went out of business, or moved on to other apps. Some users wouldn't be upgrading even if apps were moved. There is a massive logistics project to make the switch. They've done it before, and sure it can be done. But the world is different now. It's no certainty that they can draw too much on how things went the last time. They could potentially end up with significantly lower sales figures, and I think they have to have a very strong plan to keep users happy this time.
For myself, if I couldn't dual boot into Windows and run existing X86 apps, there's no way I could justify the extra cost of a MBP. If there exists a Windows 10 for ARM is completely irrelevant, because it doesn't have the apps. Until they can make the switch and have emulated x86 apps be as fast as on native Intel processors, it just won't be a win for me. And if I have nothing to gain from the switch, then why would I want it?
Getting the hardware, the performance, and the OS on ARM, that's the easy part. I have no doubt that they have that already. Getting the apps on it too, that's the challenge.
Except it does provide x86 emulation:
https://docs.microsoft.com/en-us/windows/uwp/porting/apps-on-arm-x86-emulation
It just isn't particularly good.
Many iPad Pro 2018 reviews are mentioning that the hardware is currently being held back by the software. This powerful hardware may be hinting at Apple eventually doing more with the iPad Pro. Maybe we will see major changes in iOS 13. Maybe Apple is testing how the A12X chips are performing “in the wild” with customer applications before placing such chips in computers.
Wouldn't that only be hardware x86 emulation? Software x86 emulation surely must be possible without Intel interfering?
It does, but Intel has already started firing warning shots at it and other attempts to emulate X86. Obviously, some parts of X86 are patent free, but sorting that out is likely a non-trivial task.
https://arstechnica.com/information...t-claims-x86-emulation-is-a-patent-minefield/
Software emulation is just the same as hardware emulation from a legal perspective.
Ok, that went under my radar. That will make it quite interesting to see what Apple does with its ARM transition. I guess the remaining option is for either Intel or AMD to make hybrid x86/ARM processors. Or, I guess it's still possibly to have an ARM based machine and put a tiny x86 chip in there somewhere for legacy apps.
Either way, it is refreshing that ARM is doing so well, and RISC-V is being actively pursued. From a tech perspective, the x86 platform was inferior even when it launched, and it's nothing less than a miracle that it has lasted as long as it has.
What I never liked about Geekbench is that its this weird collection of tasks, often relying on third frameworks that can be implemented differently between architectures. I get that they wanted to have an idea of "real-world" performance, but this approach makes for a poor benchmark of CPU performance.
I'd like to see some microbenchmarks on the A12X, but I haven't found an in-depth analysis yet.
Well, 64-bit ARM's Neon is every bit as good as SSE, and I do like the intrusion set more. It doesn't have 256-bit SIMD registers like AVX, but that's less relevant I think. I don't know how many execution ports A12X has for vector ops, but I don't think it's that inferior to modern Intel offerings.
Couldn’t there just be some sort of a licensing agreement, like the one AMD and Intel have going on?
AMD and Intel Patent Cross License Agreement:
https://www.sec.gov/Archives/edgar/data/2488/000119312509236705/dex102.htm
AMD and Intel Patent Cross License Agreement:
https://www.sec.gov/Archives/edgar/data/2488/000119312509236705/dex102.htm
Oh I forgot, IBM has a license too, don't they? Or at least used to. Anyway, AMD and Intel have each other grabbed by the b*lls, that's why it works there. Any other player would probably end up having to pay so much that it's no longer worth it.
[doublepost=1542263223][/doublepost]
I've found Geekbench to give different results on the exact same hardware when booting different operating systems. Haven't tried on MBP yet but possibly I should. I don't think Geekbench results are very reliable cross platform, even though that seems to have been a design goal.
I've been resisting the urge to write solid and simple benchmarks that test various hardware aspects for cross platform devices in a clear and simple way. One of these days I'm sure my willpower will fail. Sadly it's a pretty big task probably.
It can be insanely complicated, especially if you want to make sure that you are benchmarking what you think you are. Geekbench uses the "easy way" by leveraging various real-world libraries and tools, but as you noted, implementation of those libraries can differ from platform to platform.
[doublepost=1542266356][/doublepost]
And that would accomplish what exactly? I don't think that Apple would be interested in developing an Intel-compatible CPU. Frankly, I wouldn't mind of x86 instruction set would just die. It became extremely messy and unwieldy. Ideally, I'd love the industry to move to Agner's Fog https://forwardcom.info but that probably won't happen
Well, a *lot* of people rely on that compatibility.. And the major software companies are so stuck in that world, that they'd also like to retain it, from a porting point of view, I guess.
Is that really so unlikely? Apple has always been upspoken about how flexibel the kernel and whatnot is, when it comes to other platforms. And they've in recent years shown that they actually prefer to go their own way, instead of using already existing things (A#, Metal, APFS, and likely other major things I can't remember). If they want to leapfrog their chips for high-end machines, they might as well go all the way..Frankly, I wouldn't mind of x86 instruction set would just die. It became extremely messy and unwieldy. Ideally, I'd love the industry to move to Agner's Fog https://forwardcom.info but that probably won't happen
Ah, but there are simpler ways to achieve compatibility that wouldn't need any licensing or hardware support. They could simply recompile x86 code to ARM on demand. Machine code is just another programming language after all and LLVM is perfectly suited for a task like this.
Wouldn't that violate the same patents that prevent hardware and software emulation?
Oh, does the patent also prevents software emulation? At any rate, what even counts as emulation?
that legal stuff is way above my paygrade...In the end, I think it depends on the exact wording of the patent. I would be surprised if it forbid other parties to do symbolic manipulation/transformation on x86 bytecide. Because then almost every compiler and disassembler would be illegal...
Last edited:
https://forums.macrumors.com/thread...-their-own-chips.2153468/page-2#post-26806050Oh, does the patent also prevents software emulation? At any rate, what even counts as emulation?
In the end, I think it depends on the exact wording of the patent. I would be surprised if it forbid other parties to do symbolic manipulation/transformation on x86 bytecide. Because then almost every compiler and disassembler would be illegal...
As far as I understand, you don't really know what's ok and not until it's been tried in court. It's not really something that interests me. I mostly find it interesting on a human scale that we're allowing the greed of a tiny minority to stand in the way of innovation and progress.
Everyone here waiting on ARM and all I want is a no Touch Bar option, a better keyboard, maybe FaceTime authentication/FHD camera . That would make the laptop more than perfect, I doubt anyone will be like "Wish I had some ARM performance, at the cost of bugs/emulation issues/compatibility issues" - these machines are pretty darn fast as they are. If I had to bet, I reckon the majority of users don't even use half the performance.
. That would make the laptop more than perfect, I doubt anyone will be like "Wish I had some ARM performance, at the cost of bugs/emulation issues/compatibility issues" - these machines are pretty darn fast as they are. If I had to bet, I reckon the majority of users don't even use half the performance.You can be pretty sure that for 99% of all personal computers in the world, it's mostly the computer waiting for the human, rather than the human waiting for the computer.Everyone here waiting on ARM and all I want is a no Touch Bar option, a better keyboard, maybe FaceTime authentication/FHD camera
TLDR. I don't know how far Apple have really advanced with it but if it was good enough to be in a MacBook. It would be.
Count on it.
"no matter if you believe that benchmarks across architectures are comparable or not", WHHAATTT??? If comparing power, you need to do an Apples to Apples comparison. Otherwise it's loses meaning.