Waiting on some details for the new Microsoft/Qualcomm SQ2 processor. Should be an interesting comparison to Apple Silicon when they are both available but I'm interested in performance numbers on the SQ2 vs. A12Z/A14 right now.well, big news:
Microsoft lends Windows on Arm a hand with emulation layer to finally run 64-bit x86 apps at last
News from the Department of Wait, It Didn't Already Do That?www.theregister.com
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.
ARM MacBook Pro I don’t want an Overpriced Netbook.
- Thread starter TheRealAlex
- Start date
- Sort by reaction score
Here you go with marketing again. I already said I don't care about Apple (or intel, or AMD - to be clear) marketing.
GO download any software for either of those systems. You'll have options for 32 bit (x86) or 64 bit (x64) - the x86 and x64 are in the damn filenames. go look at your C: drive on a windows box. You'll see references to AMD64 and x86 in program files, etc.
People call them x86/x64 processors. Just like people did, can and will continue to call the Apple Silicon machines "Arm". You don't have to like it, and I'm done arguing semantics with you. You seem more focused on Apple Marketing than the real world.
Irrelevant again.
The problem isn't that you are wrong when you make statements like this .. the real problem is you are hell bent on moving goal posts..without admitting your mistakes.
Explain what that statement has to do with the ISA of the processor. I don't care if Apple Silicon runs ARMv8, x86, SPARC, MIPS for some random new ISA.. What the hell does "they did not build Apple Silicon from scratch" mean?
Wrong.. people call them Intel or AMD processors. People call what run on them x86 binaries.. like you pointed out.
Case in point your own signature reads:
Macbook Air 2020 - i7, 16 GB, 1TB
Desktop: AMD R7-2700X, RX Vega 64, 3.7 TB of SSD, 64 GB RAM
Why didn't you write?
Macbook Air 2020 - x86_64, 16 GB, 1TB
Desktop: x86_64, AMD GPU, 3.7 TB of SSD, 64 GB RAM
Does it matter that it is an AMD Ryzen 7 or i7 (Intel marketing term)? After all they runs x86 binaries and are therefore just x86 processors, right? Why would you pander to AMD's or Intel's marketing department and call them by the marketing name? Double standards much?
I am not the one arguing semantics. You made blatantly false statements that Apple licenses ARM cores and extends them.. when called out you started arguing semantics on ISA. Stop projecting.
Last edited:
But hey, maybe you've got more credibility than anandtech.
The fact that you think this somehow helps your argument is sad.
But hey, maybe you've got more credibility than anandtech.
But hey, maybe you've got more credibility than anandtech.
Where do those Articles you posted say " did NOT build Apple Silicon from scratch", "Not really, they're ARM compatible, based on an ARM licensed core" and "ARM don't actually make processors, everybody licenses a core and extends it, Apple are no different."?
Copy paste exact sections in those articles that back up the points you made (pasted for your convenience).
Wow. You two, seriously.
Your weird argument comes down to two completely separate things that are being mixed by both of you, Hardware and Software.
HARDWARE
Apple Silicon and the A-series SoC’s since A5 have had absolutely nothing to do with ARM Cortex CPU core reference hardware designs.
SOFTWARE
Apple licenses the ARM Instruction Set Architecture and adds its own custom extensions to this base layer of software so that is can take full advatage of the custom hardware it designed and built (i.e. the SoC’s)
Notice how when Apple talks about its hardware, i.e. Apple Silicon as a circuit board with transistors, it never uses ARM terminology. The only time Apple refers to ARM is when they are talking about interfacing with the hardware, i.e. software.
It is that simple. Stop now. Please.
Your weird argument comes down to two completely separate things that are being mixed by both of you, Hardware and Software.
HARDWARE
Apple Silicon and the A-series SoC’s since A5 have had absolutely nothing to do with ARM Cortex CPU core reference hardware designs.
SOFTWARE
Apple licenses the ARM Instruction Set Architecture and adds its own custom extensions to this base layer of software so that is can take full advatage of the custom hardware it designed and built (i.e. the SoC’s)
Notice how when Apple talks about its hardware, i.e. Apple Silicon as a circuit board with transistors, it never uses ARM terminology. The only time Apple refers to ARM is when they are talking about interfacing with the hardware, i.e. software.
It is that simple. Stop now. Please.
There is no mix up from my side. I haven't said anything differently.
That's partially correct. Apple's custom core design started with the A6. A5 was a ARM Cortex 9 derivative. A6 on the other hand was a completely different design than ARM's reference design.
Sorry but this is just plain wrong. ARM ISA is not software it is an instruction spec consisting of opcodes and instruction definitions, regs spec etc.
No, its not but I agree. The horse is dead!
But hey, maybe you've got more credibility than anandtech.
Anandtech is simply, wrong. It is as simple as that. This is an incorrect statement, they have made then before, they are making one now, and will make some in the future.
The statement that you quoted, "....Apple is taking their future into their own hands with the Arm architecture....." is incorrect, and has been since the A6 or A7. At that point, Apple changed their license from an architecture license (which means that they take the then current ARM architecture (a Cortex of whatever version) and make copies of it, either in its entirety or with some additions) to an instruction set license (in which the only thing they use is the instruction set) in which not only do they NOT use the ARM architecture, but one of their own, in-house architecture designs), I don't believe that they have the right to even use the ARM design architecture. Apple has then started designing, from the ground up, their own cores, and had added whatever else they deemed was useful. This is what allows them the freedom to add things like ML cores, thier own type of GPU cores (usually licensed from Power VR, but they may have moved away from that; I could definitely be wrong), and the accelerator cores. This is what is also giving Apple the freedom to improve performance as they see fit, and why the A series has what is generally thought to be at least a 2 year speed advantage over the other ARM cores of a similar power consumption level.
By the way, Samsung does the same with the Exynos processors, and Qualcom used to do it with the older Snapdragons, but more recently, Qualvom have just started using the ARM architecture out of the box.
It's not like you helped in placating anyone. Someone is clearly wrong (on the internet) in this weird debate so there's no need to pretend it's a 'talking over each other' issue or whatever. Also, wasn't the A6 was the first fully custom one?
You are very confused about what CPU designers in general, ARM and Apple in particular, and that AnandTech article mean when they use the word "architecture". This is one of those cases where a common English word has acquired a somewhat different than standard meaning in the context of a specific technical field.
The ARM instruction set architecture (ISA), or just "architecture" for short, is not a physical design. It's a specification, and one of its most important jobs is to constrain conforming implementations such that software is highly portable between different implementations of the ARM ISA. At the same time, it tries to leave designers as much freedom as possible.
If you license an intellectual property (IP) core from ARM, such as the Cortex-A9 CPU complex found in Apple's A5 SoC, you get the right to use that core as-is in your chip design, and you have to pay ARM per-unit royalties. (ARM IP core licensing is more complicated than that in practice, but that's the essence of it.)
If you license the ARM architecture - the famous architectural license - you get the right to design your own CPU IP cores which conform to the ARM architectural spec. You do not have to pay royalties, because the resulting cores are your IP, not ARM's. However, to compensate ARM for the loss of royalties, you do have to pay a very high price up front for the architectural license. You are allowed to add extensions, but they aren't completely arbitrary (can't interfere with the baseline ARM ISA spec).
It's not clear to me whether the architectural license allows holders to expose their custom extensions to other parties. In practice, even if Apple can, they aren't - they've been shipping ARM CPUs with custom ML acceleration instructions for some time now, but there is zero public documentation or toolchain support. The only interface they provide to app developers is a set of library calls for accelerating common math intensive tasks. If the library has the function you need, great, but otherwise you're out of luck, there is no way to write your own code using those instructions.
So, AnandTech's article is correct about this, and you are not. Apple did not have to switch away from ARM's architectural license to design their first custom core ("Swift", shipped in the A6 SoC), they had to switch to it.
Again, the AS SoCs are NOT "......such as the Cortex-A9 CPU complex found in Apple's A5 SoC, you get the right to use that core as-is in your chip design..." as you describe. The AS cores, like has been true from the A6 generation on, are Apple internal designs. They are not based on ANY ARM physical design. Only the instruction set is common. Apple left the Cortex physical design in the rear view mirror over 5 years ago.
They licensed the Instruction Set, NOT the physical design (just to leave out the "architecture" and its multitude of meanings aside).Since the A6, the physical design has been Apple internal, down to the transistors. They did not use the core, or change the core design, they discarded it. Cortex is the physical design, the various ARM v8, v8a, v9, etc. are the Instruction Set versions. The Cortex is the reference physical design. Qualcom makes Cortex CPUs; Apple does not (even though TSMC is the foundry)
They licensed the Instruction Set, NOT the physical design (just to leave out the "architecture" and its multitude of meanings aside).Since the A6, the physical design has been Apple internal, down to the transistors. They did not use the core, or change the core design, they discarded it. Cortex is the physical design, the various ARM v8, v8a, v9, etc. are the Instruction Set versions. The Cortex is the reference physical design. Qualcom makes Cortex CPUs; Apple does not (even though TSMC is the foundry)
Last edited:
I think this is where you and many others are confused. Allow me to put on a computer scientist's hat for a second:
Instruction Set Architecture – Computer Architecture
An instruction set is the abstract architecture of a computer. But that's not quite what it is. What it is is a set of "specifications" that a physical computer system organization as well as the compiler of any high-level programming language have to adhere to. So in that sense, that makes it a set of "requirements" that your hardware and compiler have to meet.
"Software" in the strictest sense of the word, an organization of logic to achieve a certain purpose, is not necessarily adherent to the idea of an "instruction set". For instance, the logic I put together to control a cash register is "software". The translation layer (the compiler) is not necessarily a part of this "software". It's a part of the computing pipeline, or just another layer to get my "software" to work with a certain computer with a certain architecture.
Similarly, I think you are understanding "hardware" in this sense as the "organization of physical parts to perform computing tasks as defined by the software". So yes, in that case, I do agree with you that your computer or "hardware" is not strictly "the" abstract architecture. Your processor can meet the standards as defined in the instruction set, but it obviously can do so much more.
But... in software, there's a similar concept of an "interface". If your organization of logic, a "class" per se in some languages, is implementing an interface, then the convention is: it is what that "interface" is defining. So in that sense, if my software was written with the intentions to be adherent to a set of requirements as defined by an instruction set, then... it might well be what that instruction set is defining. This is is the current industry standard. And it applies to both software and hardware.
Let's bring that back to the current discussion: the ARM ISA is basically like the interface. It enforces certain constraints (requirements or specifications) that your processor has to meet. So your processor, by industry standard, is an ARM processor. And yes, that also means that, by industry standards, if your processor can meet both ARM and x86 specifications, then it's both ARM and x86.
So let's just say you are not wrong, but the industry enforces its way. Just like how ARM enforces constraints on Apple Silicon.
P.S.: that's also why I have to go back and recompile my software for them to work on ARM64. And again... I do understand what you are trying to say and I do agree with you. But that's not how the industry works. My software was previously x86, and now it needs adjustments to be "adherent" to the ARM compiler... and now it's both an x86 and ARM64 software. But hell, I didn't write even a single line of x86 or ARM64-specific instruction.
What a whole load of sermonizing bollocks!
Please don't try to explain to a comupter engineer of 20 years that has worked on kernel, device drivers, virtualization and found complier bugs, written handcoded assembly in very restrictive (high trap level, hyper priviledged with 8 registers context) environments, implemented instructions in emulators for the kernel, what any of this means!
I know I am not wrong. I clearly said Apple silicon is an ARM chip but it is not an "ARM" chip.
None of that you said has anything to do with the discussion in progress.
Do you think Apple's processors uses ARM Holding IP and merely extends them? That's the crux of the discussion.
And despite 20 years in the industry, you don't get what the "standards" are?
I do not "think". I know for a fact that Apple's processors have to adhere to specifications and standards set forth by ARM holding's IPs (specifically the instruction set). And by adhering to those standards, they are, by what the industry defines, ARM processors.
Whatever you or I think about the matter does not change the fact that this is what the industry has chosen to call them.
There is no discussion here. You're simply choosing to ignore facts that do not suit your narrative.
Clearly context is lost on you! What does "standards" have to do with my question?
No you are more confused than ever between Instructions Set and Silicon IP in typical fashion.
It is utterly irrelevant to this discussion that Apple implements ARMv8-A ISA in their processors. We are talking about Silicon IP.
From ARM's site you even see it mention IP clearly in the Link .
Products – Arm®
ARM is the world’s leading semiconductor IP company. We develop and license technology that is at the heart of many of the digital electronic devices sold each year, from sensors to smartphones to servers.
www.arm.com
The clearly call it Processor IP (Of which Cortex is a part).
Microprocessor Cores and Processor Technology
Arm offers top processor IP for AI, ML, and all device types, from IoT to supercomputers, & addresses performance, power, and cost with a broad core range.
www.arm.com
Physical Implementation for POP IP
Physical implementation for POP IP optimizes power, performance, area, and cost. Arm POP IP leverages advanced knowledge to deliver efficient, high-performance processors and speeds up core hardening and market entry.
www.arm.com
I'll ask again do you believe that Apple's A6+ processors dervive any of their Core Silicon IP from ARM Holdings the ones listed on their site?
If you believe they do then I have bad news for you about Santa Claus!
Last edited:
So you are saying that the ARM instruction set is actually not a part of ARM Holdings' IP portfolio?
Okay, I think we're done.
Again, you are ignoring facts that do not fit your narrative. Your narrative is this in a nutshell:
Because Apple's A6+SoCs are not using any of ARM's Cortex-A series cores, we can't technically call Apple Silicon an "ARM processor".
And I have pointed out that you are intentionally ignoring the fact that we call something an "ARM processor" not because of the physical silicon core it's implementing, but because of the instruction set architecture that it's implementing.
Just like Intel's Core i7 is an "x86 processor", just like AMD Ryzen is also an "x86 processor". Apple A14 is an "ARM processor".
But again, I know you'll just ignore the facts and try to push your narrative once again. I have tried to see it your way, but I guess your viewpoint is the only one that matters because you are the only computer engineer here with 20 years of experience.
That has never been my narrative.. go read the thread history. I have only gone after the claim the Apple extends ARM Cores Like everyone else. Like I said context is lost on you!
Yes, unlike computer scientists that don’t understand context, one of the basic concepts in the field!
No, context is not lost on me.
You meant this, right?
That statement does not read anything out of the ordinary to me.
Apple has to obtain a license from ARM before they can develop an ARM-compatible processor core. So Apple's processor core is "licensed by ARM".
I'm guessing you are reading that differently. I don't see where he claimed that the ARM licensed core, the bolded part, was also ARM-designed.
Let's just agree to disagree.
No. Apple is a founding member of ARM Holdings and has a architecture license for ARM.
There is no such thing as a core licensed by ARM that is also not designed by ARM. You either get the core IP or not. Please don’t make stuff up. Familiarize your self with the different licenses ARM offers, first.
ARM architecture family - Wikipedia
A long look at how ARM licenses chips
People routinely ask SemiAccurate about ARM CPUs which is an odd question because ARM makes absolutely no silicon.
semiaccurate.com
“Architectural licensees get a set of specs and a testing suite that they have to pass, the rest is up to them. If they want to make a processor that is faster, slower, more efficient, smaller, or anything else than the one ARM supplies, this is what they have to do.”
ARM licensed core means ARM designed there is no other way to interpret that.
Apple hasn’t used an ARM designed or licensed core sine the A5.
Last edited:
Thank you! Somebody gets it. Apple has a perpetual arm license and all of their cores are derivatives of that license as they implement the ARM ISA.
How Apple have implemented that, and whether they redesigned it or whatever is irrelevant. It's like saying American English isn't "English".
Now, Apple silicon does however cover their integrated GPU which IS 100% in house design. But this GPUs inclusion is implicit when running "Apple Silicon" or ARM based Mac or iOS processors.
So, they're ARM macs. Also called Apple silicon. They're both. But to claim they are not also ARM (and ARM is much easier to type, have been discussed as ARM based for 5+ years at this point, etc.) is simply not correct.
If someone wants to call them Apple Silicon i have no issue with that. However if you're going to take people to task for calling them ARM and argue they are not - well you're simply not correct. They're both.
from what I can see, that’s the poster who’s been disagreeing with you up until now...