I did some research into using ARM in Macs around Christmas time when I was ill and couldn't sleep. The TDP would shoot way high to be able to keep up with big time software that does a lot of CPU crunching in the background.
So it's true that the A series does have a high TDP at load.
The problem Intel has is that the A series TDP at load is still better than Intel's mobile TDP. If you stuck a fan on an A series chip it would outperform Intel right now at load. Easily. And the Apple's graphics would run circles around Intel's integrated graphics.
As Intel improves their TDP maybe this changes, but they need to catch up and keep pace, and they just aren't right now. They're still shipping the same CPU they've been shipping for what... the last five years?
The last time Apple had a CPU vendor (IBM) that fell behind on TDP and couldn't keep up they dropped them. This is history repeating itself.
Then there's the question of software companies having to keep two separate codebases. Recompiling is one thing, making sure it works and is as bug free as the traditional counterpart is another.
That's.... not a thing. You build ARM from the same code base. It's not two different code bases. Typically it's the exact same code with no changes. For almost all apps it's literally just a recompile on the existing code. I spend all day porting between x86 and ARM. Usually I'm just messing with build settings and not the code.
This whole thing is literally my job and I think you guys are all making a way bigger deal about it than it actually is.
That said, in a Mac Pro sphere, I can't imagine Adobe redeveloping software to work on high powered ARM chips when their time and effort is and has been Intel x86-64.
So it's worth repeating that I don't think the Mac Pro would be moving over to ARM in the near future, BUT...
- It's not a big job to move to ARM.
- Metal code moves between x86 and ARM with zero changes.
- Adobe just ported nearly all their performance code to Metal.
Tada.
Like I said, everyone here is assuming the transition hasn't started yet. Apple queued it up years ago. We're already in the transition. Moving to Metal is part of the transition. Look at everything Apple has done in the last few years and the next few years on macOS with an eye towards ARM.
If you look into it, even the Apple die-hards will admit an ARM chip that's as powerful as a 9900K, for example, on everything and not synthetic benchmarks, will consume just as much power. This is pure speculation as I simply don't see the costs for all involved parties being viable in the short term or even long term.
Wait let's back up.
Apple doesn't make an ARM chip right now that competes with the 9900K. I don't think they will any time soon.
But Intel is in a bad place right now, and if Apple chose to compete with the 9900K, I think they could build a much better ARM chip. It's not that ARM is implicitly better. It's that Intel is building such bad x86 chips right now.
(I'm leaving AMD off to the side for most these comparisons, but AMD is an option as well.)
I won't deny the A12 and A12X are amazing ARM processors, but I'd wonder how a higher TDP (say, 150 watts) version, would perform head to head against a Xeon or high end i9 mainstream or extreme processor in a variety of use-case tasks.
Extrapolating, it's would likely do very well. Apple has a process advantage so that alone would be a huge problem for Intel today. Apple would probably have a 30% advantage on that alone.
But such a chip does not exist and Apple can't simply pull one out of thin air.
AMD vs an Apple A series would be a much more interesting comparison in since they're both in the same place technologically.
Breaking down synthetics means looking at each test. If the ARM processors by Apple are amazing at say script rendering, then they're finely tuned for that alongside the native browsers that come with the device.
The latest ARM specification that Apple is using does have some optimizations meant to help JavaScript along, but generally this is not true. There is nothing in the A series chips that is specifically optimized for a specific use. Maybe really early versions, but the current chips are general use.
There are some server ARM chips floating around out there that are optimized for specific things, but Apple has increasingly been designing the A series as a general chip. Apple used to skip things like branch prediction, but these days the A series is a legit desktop grade chip. Just... smaller. Desktop grade but not built to scale for actual desktops.
[doublepost=1548743720][/doublepost]
The challenge isn't a $50 5-10 watt A-series ARM CPU that is faster than Intel at the same price/power point... That's a given, and a MacBook Air-size machine that uses it isn't hard to imagine. It would mean some sort of fat binaries to deal with "some Macs are Intel, while others are ARM". It could even be restricted to the Mac App Store as a method of app distribution - as long as ARM Macs are only a low price/low power option. Losing Boot Camp and Parallels isn't a big deal in that end of the market - that machine won't be running AutoCAD and ArcGIS.
The real challenge is a $350 45-watt laptop chip for the big MBP (with a $500 option for BTO, and with options down to $100 and 20 watts to fill in the 13" line), a series of $100-$450, 35-90 watt desktop chips for iMacs, and $500-$2000 workstation chips for iMac Pros and Mac Pros. Especially at the higher end of that range, Mac App Store only and/or losing Boot Camp and Parallels isn't an option.
This. Mostly.
The problem isn't ARM. It's that Apple doesn't have an ARM chip ready for the bigger Macs. If they had one, it would outperform Intel. But they don't. At least nothing we've seen publicly.
Where I disagree is the bit about the app store. I think ARM Macs will be unlocked from the app store, and Apple is already quietly prepping the fat binary support if you look at recent macOS changes. We'll probably see the first public rollout of ARM/x86 fat binaries with Marzipan. Even if they aren't downloaded as a fat binary, they'll be submitted as a fat binary most likely so Apple already has both architectures ready to go.
I don't think Apple will care that much about Boot Camp, but again, it doesn't have to go. Windows on ARM is a thing. It runs existing x86 software. You can still have your AutoCAD and ArcGIS on an ARM Mac. In theory. We'll see if Apple bothers with DirectX drivers for their A series GPUs.
