All these rumors are nice and all, but I'm a simple man: I only wish that Apple would add a heat pipe to MacBook Air. I've been looking to replace my ancient MacBook Air for ages. Come on Apple! I just want a simple, competent, and dependable laptop with easily replaceable batteries. You are almost there with the new MacBook Air, almost! It wasn't this hard to spend money on Apple products when Steve was alive.
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.
Top Stories: Redesigned iMac at WWDC?, Mac's Transition to Arm Chips, AAPL Hits $1.5 Trillion
- Thread starter MacRumors
- Start date
- Sort by reaction score
Not that simple a question. With what type of workload? Things like video encoding/transcoding may be MUCH faster on apple’s chip, while floating point math may be much slower, and single-threaded integer workloads may be a dead heat.
I figure that, across the board, if Apple releases a mac that is the new version of a prior mac, the new, ARM mac will be around 20% faster than the prior mac on most workloads.
So, for example, the 2021 Arm 16” MBP will beat the *base* 2020 Intel 16” MBP by 20% on most workloads, while the Arm iMac will beat the *base* Intel iMac by about 20%.
Query whether they offer multiple arm choices - a base model and an up-clocked one, for example. Doubt it, at least to start.
They could do it because PowerPC had fallen so far behind x86 by the time Apple transitioned that the x86 could easily take the performance hit and still run PowerPC code as fast as a native chip. In very low power consumption devices like ultrabooks, Arm may well be able to rival Intel/AMD chips now, but they certainly don't have the excess power needed to do seamless emulation.
Also, windows/x86 had and has 95% of the market. There was a huge amount of x86 native code that could never run on a powerPC but was easily ported to run natively on x86 macs. By the same token, there will be a huge amount of software that's not worth the developer's trouble to support two architectures when one of those is 5% of the market. You make think it's as easy as a compiler switch, but a lot of non-trivial software uses architecture specific coding techniques.
And don't forget that we've been talking about ultrabook class cpus. Unless Apple wants to support two architectures from here on out, they're giving up the entire high end and pro market. Arm has nothing to compete against higher power desktop CPUs much less server class CPUs. So no more Mac Pro or higher end iMacs at all. Any gains Apple would get from an Arm transition would be more than lost by having to architectures that are code-incompatible. Remember the utter debacle that was Windows RT. I don't see Apple going down that road.
Most of the time, said "architecture specific coding techniques" are nothing but latent bugs.
I can't speak for your coding experience. Back in the day, my code was pretty close to hardware level and I used a fair bit of inline assembly to get the needed performance.
Optimizing compilers still have not gotten as good as a skilled programmer, and hardware is still not fast enough that it never matters.
Even if you're right though, the developer is not going to fix the bug just to make the code they designed for x86 or x64 run on Arm. The market is just not worth the cost for all but the most mainstream apps.
I used to have to use in-line assembler to get around some compiler bugs or to improve performance of critical paths. Can’t remember the last time that happened. Probably 1995?
I might get a bit of hate for this, but I'll miss the bezels and chin. I have a 21-inch iMac that I use daily, and I feel like the bezels and chin five it a bit of personality that's uniquely iMac, compared to every other device on the market. It feels like people are too obsessed with edge-to-edge screens and getting rid of bezels and don't stop to think whether it would look good, and I personally think an iMac with super-thin bezels and iPad Pro design language wouldn't look very good at all. It just doesn't click for me, and I'll be gladly sticking with my 'fat' bezels.
I think you're underestimating the capabilities of Apple's chips here. The "ultrabook" class A12X was already beating many desktop/sever class Intel chips back in 2018. Apple surely haven't stood still since then, even if Intel has. The A12X also does this with only 4 high-performance cores.
Imagine what the 2020/2021 (A14X? A15XL?) versions are going to do. A desktop-class variant with 6 or 8 next-generation cores? Or a massively parallel Mac Pro with dozens of cores. At the rate Intel have been going recently, they will struggle just to keep up.
The stand is $1. The special screws to connect it to the monitor are $999.
First, the gen 8 i7 from 2 years ago was roughly twice as fast as the A12X in benchmarks. Remember we're talking about the whole Mac line eventually so it has to beat desktop CPUs as well as ultrabook CPUs, because it doesn't make sense to have 2 incompatible architectures, MS tried with win WinRT and it was a fiasco.
Second, I was also talking about using emulation to run legacy code like Apple did with the Intel transition. For that, you don't just have to match performance, you have to beat it by a good margin. IMO, that's Apple's biggest hurdle. There's just too much x86/x64 code that companies will not bother porting because Mac's market share is too small to bother with the expense. Even Microsoft couldn't convince devs to port their code to Arm for WinRT.
Third, the Arms cannot sustain the performance for more than a few minutes, then they have to throttle due to overheating. The i7 can maintain top speed 24 hours a day if need be. It's not a problem for most use cases in a tablet and even many use cases on a light-duty notebook. I'd hate to be a software dev or graphics designer using an A12X for compiling or rendering.
The arms you refer to have ipad/iphone cooling solutions. Power dissipation solutions involve dissipating W/cm^2. So you need more surface area and heat spreading. If you stick an a12x in a MBP chassis it would *never* throttle.
Begun the second RISC-CISC war has, and RISC-y for Apple it will be. An excellent compiler and pipeline scheme Timmy must provide. Grave doubts I have: overpromised on pipeline Timmy always has.
Last edited:
Darth Tulhu
macrumors 68020
It won't. The new iMac will just cost $1k more.
But seriously, as I've mentioned in another thread I'm hoping for an iMac that is a Surface Studio equivalent. A 27" iPad with Pencil support.
There were some old patent designs that pointed to this... now I can totally see it happening.
I couldn’t agree more with your last sentence. Buying Macs these days feels like a compromise. It isn’t an exciting prospect.
Why does a 1,5 T company only have 3 phone models on the market? And NONE of them with a headphone jack??
All the important "pro" apps will get updated pretty quickly. There's no doubt that Apple understands the importance of that and will be devoting significant resources to make sure it happens.
Where emulation is needed is for those random little apps you found in 2014 that are very useful but the developer has since disappeared and isn't providing updates any more. Generally speaking, these aren't really CPU-dependent, so even if emulation is a little slower, you aren't going to notice.
Likewise, a lot of older games will never be updated. But the performance-critical part of most games is their GPU code and shaders etc. Provided those still work well, you probably won't notice if the CPU is emulated a little slower.
Last time around, the "Rosetta" emulator only had a brief lifespan (10.4.4 - 10.5.x) before being made an optional install in 10.6 and removed entirely in 10.7. It just wasn't needed anymore because all the apps anyone cared about had been updated or replaced.
Finally, I agree with you that it's unlikely they'll be splitting the Mac line with different architectures in different models. Apple wouldn't be embarking on this whole endeavour if they didn't have a lot of confidence in their chip roadmap for years to come. They will be very confident that they have chips in the pipeline that can beat their competitors - across all segments.
In your whole post you are totally ignoring the power envelope. So which i7 SKU are you referring to? The 15W TDP, 25W TDP, 45W TDP or 95W TDP i7?
And which benchmark are you referring to?
Thats an excellent point. The best option to run x86 Windows apps on a future ARM Mac would be to virtualize ARM64 Windows and let the Windows x86 emulator do its job. As you mentioned, only the application layer SW is emulated but not the kernel and the drivers.
And then Windows has Linux virtualization built in via WSL2. For this Microsoft maintains a custom Linux kernel - but you can provide your own compiled kernel if you want to.