I can see this being the case for lighter-weight programs that don't significantly challenge the CPU. But aren't a lot of heavier-weight programs (by which I mean ones that need CPU optimization) hardware-optimized? E.g., suppose you have code that has been highly-optimized to be cache-friendly for current Intel x86 chips; would it need to be written differently to be cache-friendly for Apple Silicon?
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.
What makes ARM superior?
- Thread starter Jmausmuc
- Start date
- Sort by reaction score
DISCLAIMER: I'm only a hobbyist programmer, not one by trade. Nor am I intimately familiar with a lot of the internal works of large software.
I would guess that's the exact case for older programs. But from my understanding older pieces of code (so long as someone has access to the source) could be recompiled on an Apple (or ARM depending) and then work as intended (minus any bugs that may or may not crop up).
The problems arise when one doesn't have access to the source code, there's lots of programs that (I assume) people use that haven't been updated or maintained for awhile that would not work at all or work poorly under emulation(translation?). Which is a sticking point for some that I would not discredit.
It could be an issue, but I have doubts that it'll be a big issue.
What makes ARM superior?
It has more registers, reduced design complexity, a proper 3-operand instruction set, and more transistors allocated towards performance.
But, there are a few caveats....
It's not as widespread as x86-based designs in terms of traditional computing. Apple will be among the first to offer an ARM-based notebook and desktop solution. There are some ARM machines out there but most are power-hungry servers.
It has more registers, reduced design complexity, a proper 3-operand instruction set, and more transistors allocated towards performance.
But, there are a few caveats....
It's not as widespread as x86-based designs in terms of traditional computing. Apple will be among the first to offer an ARM-based notebook and desktop solution. There are some ARM machines out there but most are power-hungry servers.
ARM is generally viewed as being best for low power usage but its performance compared to x84 is unproven.
The older programs can be recompiled but I believe that if they are coded with important hardware-specific optimizations, those need to be rewritten for AS. And note that many of these older programs are important (like Office and Adobe CC). Indeed, it's precisely because it's *not* a trivial exercise to move these programs to AS that there has been (because there needed to be) significant effort, planning, and coordination betwen Apple and MS, and Apple and Adobe, to ready these respective software suites for AS.
On the contrary, this all assumes that the source code is available (which it should always be, since these programs come from their respective developers). Rosetta 2 doesn't translate x86 binaries to AS, it (imperfectly) translates source code written for x86 to AS.
Last edited:
ARM isn't based on a design that was brain-damaged even back in the 1970s (look up cpu architecture/programming in assembly for say, Motorola 68k vs. x86, which were designed around the same time), and then kludged together with 40-50 years of band-aids and sticky tape.
It's amazing what intel have done with x86 don't get me wrong - but when you start out with something that was plain bad, and then maintain compatibility with that for 4-5 decades, the end result isn't pretty. or secure. or easy to fix and maintain.
x86 succeeded due to IBM using it in the original PC, that architecture being cloned and mass produced. and later, intel's illegal/shady business practices to drive competitors out of business.
it's nothing to do with it being inherently "good".
ARM, even when it came out originally back in the archimedes (in the 80s) blew x86 out of the water performance wise. But intel were a much bigger company, everybody wanted pc compatibility, and the Acorn machines never took off outside of UK/Australian education. They were too expensive for the home market and business already settled on PCs.
It's amazing what intel have done with x86 don't get me wrong - but when you start out with something that was plain bad, and then maintain compatibility with that for 4-5 decades, the end result isn't pretty. or secure. or easy to fix and maintain.
x86 succeeded due to IBM using it in the original PC, that architecture being cloned and mass produced. and later, intel's illegal/shady business practices to drive competitors out of business.
it's nothing to do with it being inherently "good".
ARM, even when it came out originally back in the archimedes (in the 80s) blew x86 out of the water performance wise. But intel were a much bigger company, everybody wanted pc compatibility, and the Acorn machines never took off outside of UK/Australian education. They were too expensive for the home market and business already settled on PCs.
Last edited:
I believe you have that exactly backwards.
You by a new AS Mac, you migrate your software and data from your Intel Mac to your new AS Mac. The old Intel programs run on your new AS Mac thanks to Rosetta 2. You only have the binary files. You do not have the source code.
Rosetta 2 will allow AS Macs to run the software you already own by converting the Intel instructions to AS instructions. It does not translate source code.
This is correct. An "A" Chip for every task. And why not stack them or quad the A chips together. It is done now to get more processing power. Same process but soon to be in the Neg Nano Configs. All chips identical just programed to do different tasks...
ARM has been well tested against x86 servers in the data center. ARM compares favorably with Intel and AMD high-end CPUs for many tasks. ARM is not great for everything, but in many workloads it runs faster than more expensive and power hungry x86 CPUs. The big advantage is that is is often much cheaper to run ARM for a similar sized machine (about 50% of the cost) and this is hugely important for businesses.
Let's not confuse ARM with Apple Silicon though. The latter uses the ARM instruction set, but the SoC designs are Apple's and optimized for Apple hardware.
This sounds a little biased to me. In mobile devices, most of your graphics chips are very close to your cpu. On your PC, these are not physically close together, so any task mapped to it has to be large enough to amortize the dispatch cost. You can also achieve very high throughput on x86 by leveraging AVX and saturating max instructions per clock cycle. The issues are that Intel has too many opcodes for too many variations of things, and compiler technology has been playing a long game of catch up on auto vectorization. It's only in the last couple versions of gcc and clang that it started being useful for anything that isn't incredibly simple.
I tend to think the tendency to throw stuff to the gpu is just because it's compact enough. A few years ago this was still a bad idea, even on mobile devices..
Apple chips will use the ARM instruction set with some minor mods. There’s nothing distinct here other than marketing at play. BTW, performance isn’t just about running benchmarks. ARM chipset Apple will leverage is totally unproven. Let’s see how well that can drive high IO bandwidth.
I get what he's saying regarding ARM and Apple Silicon, but Apple Silicon is such an awkward thing to say. (especially the way he says it) There should be some shorthand term.
Apple uses the ARM instruction set but they don’t use the ARM-designed CPUs. That’s the only relevant distinction.
Not sure you quite understand the difference between chip design and instruction set , if i increase my L2 cache , design a better branch predictor , pay area for more ALU`s and a dozens of other Mircro arch decisions/improvements , you think it wont matter because i comply to the same instruction set ?
If you take 2 years ago Y series Intel x86 chip , which has the same instruction set as Ryzen 3950X , can you compare them in any metric ? you cannot , not core 2 core and not Chip vs Chip , different interconnects , different frequency , different pipelines depth , different HW accelerators , boost schemes , different OOO scheme , different ALU design , different Memory controller implementation , different QoS scheme , different Cache sizes , different Cache inclusion rules , different layout of the chip , different memory latency due to different Fabric topology, different power management design , and many many many more differences ,
but you know what ? they both have the X86-64 instruction set implemented!! so they are basically the same!! Its surprising they are able to dupe the users to buy those Ryzen`s when Intel can match them that easily with a fraction of a price! /s
Regarding "can drive high IO bandwidth" , can you clarify what you mean by that ? which IO specifically you think Intel can drive "well".
no the CPU is only one part of the chip, and even then, ARM has nothing to do with it other than allowing an instruction set to be used. Chips are enormously complex and unique across designs.
You are arguing semantics. What does “ARM chip” mean? Is it a chit that implements ARM ISA? Is it a chip designed by ARM? Is it a chip certified by ARM?
We should stop throwing around ambiguous labels and talk more context, then the discussion is more productive.
As many as they need. Unlike Intel, Apple isn't going to constrain lanes for marketing reasons. Apple will design their SoCs that meet the needs of the platform that it was designed for.
"Apple silicon" is no doubt a place holder term as Apple doesn't want to reveal which chips will be used. Each machine will no doubt have an A-number designation, like A12 or A14, once announced.
Possibly. But the key here is we need to see what this architecture looks like to make a decision about its suitability. Intel defines what these markets are and Apple's ARM chips (Apple Silicon if you must) has to play by those rules in the marketplace.
Why? Apple isn't going to market their SoCs. They are going to sell ASi Macs as complete units just like they do for everything they sell today. You have to dig to find out what Intel CPU they are using for different SKUs. When you buy a MacBook Pro, Apple doesn't tell you how many PCIe lanes it has, just that it supports 2-4 Thunderbolt 3 ports. ASi Macs will be the same.