I thought it was mostly painless. Vast majority of applications provided an Intel binary right away without many issues. Rosetta wasn't too terrible for simple tasks for the rest of them.
Rosetta was mostly painless because the raw performance of the Intel chips were multiples of the G4 chips in the macs (especially considering most systems doubled the available core count). Even so Creative Suite was slower as high performance plugins used altivec which was not emulated. This wasn't an issue for compatibility as the G3's never had it, so almost every app had non altivec code paths.
The overhead was mostly acceptable because the performance boost from the chip change hid a lot of it. Not to mention the CISC vs RISC thing worked in Apple's Favour. There are few PPC instructions that were required to be supported that couldn't be covered with a small number of x86 instructions.
There is not a publicly available ARM chip that has single thread performance anywhere near the i5 / i7 chips being used. Add to this the complexity of converting SSE to equivalent instructions and the overheads dealing with edge cases, I am not sure there is much chance of equivalent performance.
The one thing going for Apple with an intel -> arm switch is that almost all major apps a built using Xcode. When PPC -> x86 happened, the apps that lagged the most were the ones stuck on CodeWarrior and other 3rd party IDEs that apple never gave sufficient notice to port in advance of the general announcement (however I will concede that codewarrior was owned by Motorola so it may not have made the jump anyway).
I think if Apple does release ARM based macs, it will be with apps specifically recompiled, not using an emulation layer.
Performance per watt is pretty good on ARM. Most performance these days is about IO. How fast is the memory subsystem?
There are intel chips with 400 gigs per second memory subsystems. But the cores run at around 1.4ghz. Even a single thread with that much memory bandwidth can be much faster. Likewise with cache. If you have 32MB of cache, you can do quite a lot more processing in cache, and be lots faster for many things.
Considering that macbookpro laptops haven't really gotten better single thread performance in years - the things that really matter are IO, and performance within the power/heat budget.
ARM is competitive now within the macbook power requirements. The A10 chip is 75% as fast as the i7-6600U in some single core benchmarks. At a much lower power budget.
Within the chromebook space, where ARM and Intel laptops are currently competing - Intel often produces faster laptops. However, they are also often more expensive. Lots of people in many-many reviews say their ARM based chomebooks are too slow.
Having been through the raspberrypi 1,2,3 performance leaps, I can say that the latest one is quite usable as a desktop machine for word processing, web browsing and media playing.
The Samsung Chromebook plus with ARM is shipping, and the one with Intel isn't yet. The ARM one is cheaper, and the intel one is apparently more performant. Both have a similar battery life. The ARM one has better support of android apps. GUI app development is now 10x more active on ARM compared to intel, and the apps are optimized more for ARM.
GPU performance is improving faster on mobile chipsets than Intel is improving. Because of the demand from mobile VR applications. Some are predicting performance will be faster than laptops in a mobile power budget this year (see ARM Mali-G71). It's even faster than some nvidia discrete mobile chipsets. Since many apps are performance limited by the GPU, and not the CPU, I'd say ARM chipsets will win for many users low-mid users this year.
ARM is already taking over the low end laptop market. I won't be at all surprised if they take the mid range market by the end of the year. Especially as the market for chromebooks, and the low end laptops picks up. Because then more laptop class ARM based chips will go into production.
I think the demand will be there, as the android using market also want to use laptops that sync better with their apps.
[ED: add note about high performance ARM chips coming]
High performance ARM chips are coming to super computers. The Post-K super computer by Fujitsu is expected to be one of the fastest computers in the entire world. If not the fastest. Will ARM take the super computer performance crown? I think so.
I have no doubt ARM will be in mid to high end laptops. I doubt though that the transition will be made emulating x86 apps for most users like it was with rosetta. 75% of Single core benchmarks is likely to become 40% even with a highly efficient emulating layer.
It makes sense to simply make vendors rebuild the app for the new platform then to use an emulating layer.
An emulator may be used for legacy apps, where performance doesn't matter, but apple has never worried too much about legacy in the past, so the mac community won't expect it. All the laptop needs to have is some other killer feature and the community will put enough pressure on the vendors to produce an OSX/arm build of their products.
Yes, good point. It will be interesting to see how popular they are this year. I also wonder if Office performance will be good enough for most people. I'm pretty sure MS know what they are doing with that, and the answer will be yes.
I'll buy one if they are cheap and include a sim slot. It would basically be a smartphone with Windows desktop software (mostly Office in my case) running on an 11.6-inch screen...
Speculating about what Apple's going to do is a great way to be wrong but I don't really see it. In both of their CPU migrations the switch was to substantially more powerful CPUs. I can see them making more 'pro' iOS devices, Macs with ARMs as coprocessors, etc. Ditching Intel altogether? I can't think of ways the advantages would outweigh the disadvantages.
My guess is that the CPU "leap" will be in power efficiency rather than raw performance. My guess is that the returns for each jump in CPU power these days isn't helping the vast majority of users, but they certainly care about battery life.
Licensing x86/AMD64 requires licenses from both Intel and AMD, which probably would be quite expensive, even if they can threaten to go ARM otherwise, and engineers with the low-level experience probably are way harder to find than for ARM (where quite a few companies hold various levels of licensing, and Apple likely already has quite a lot of talent inhouse)
> engineers with the low-level experience probably are way harder to find than for ARM
Eh, the instruction set doesn't matter that much for modern CPU designers (within reason). Apple's in house expertise came (in part) from their acquisition of P.A. Semi and their PowerPC engineers; Apple it seems immediately redirected them to working on ARM designs. And the Transmeta guys proved that a third party can put out an interesting x86 chip if you can get access to the licenses.
No he means that running x86 legacy mac programs on an arm mac with an emulator as a compatibility shim would be slow. Linux is totally different, because you can clone an entire DISTRO, cross compile it, and have a native arm environment with no compatibility shims slowing things down.