Kuo: Apple to Launch Several Macs With Arm-Based Processors in 2021, USB4 Support Coming to Macs in 2022

[cmaier]

No, lot of peoples do care and I think Apple too, especially if you work for a company with a Windows only internal software

Why would apple care if you work for a company with windows only internal software? Those companies, almost without exception, buy windows machines. Apple makes very little money selling macs compared to everything else it sells. Only a small percentage of those run windows. Why would apple let its product strategy be hamstrung by such a small contingent? Switching to arm allows them to make machines that have unique form factors, are much more attractive to their huge iOS install base, and have features nobody else has. It’s the future.

If you run windows on Mac you aren’t using apple’s services on Mac, you are less likely to use the rest of Apple’s ecosystem, etc. You aren’t making enough money for them to move their decision needle.

 

[Stephen.R]

future.

If you run windows on Mac you aren’t using apple’s services on Mac

I mean, you know that iCloud on windows is a thing right? iTunes (and the iTunes Store) on windows is a thing.

 

[4487549]

Isn't it possible that this rumor is only half true? Couldn't Apple be working on their own x86 CPU?

 

[ingambe]

Why would apple care if you work for a company with windows only internal software? Those companies, almost without exception, buy windows machines. Apple makes very little money selling macs compared to everything else it sells. Only a small percentage of those run windows. Why would apple let its product strategy be hamstrung by such a small contingent? Switching to arm allows them to make machines that have unique form factors, are much more attractive to their huge iOS install base, and have features nobody else has. It’s the future.

If you run windows on Mac you aren’t using apple’s services on Mac, you are less likely to use the rest of Apple’s ecosystem, etc. You aren’t making enough money for them to move their decision needle.

it’s not a small pourcentage, for example my previous company had a Windows only bug ticket report software, BUT lot of peoples like me asked for a Mac, when we need a ticket, I was booting on Mac, download the HTML report, boot on Mac and do the job. I was way more productive with MacOs
i explain my usage, but lot of people do have these kind of usage, lot of company have 20 years old software running and these software doesn’t work with MacOs.

 

[Unregistered 4U]

1) people who really want windows buy a windows machine. Apple can’t base its strategy on windows. And windows-on-mac is a small portion of mac users.

2) there will initially be MUCH MORE software compatible with ARM mac, because it will be able to run most iPadOS and iOS software natively. There are orders of magnitude more of such software than is currently available for the mac.

And 3) I think everything in the AppStore (where the vast majority of users download their Mac software), will likely be compatible on day one because Apple will require it and delist the app if it can’t be downloaded in an ARM/Intel version.

 

[ingambe]

Well, it is superior. In terms of engineering, ARM is better than x86.

CISC are way more complex than ARM RISC architecture

 

[chucker23n1]

I mean, you know that iCloud on windows is a thing right? iTunes (and the iTunes Store) on windows is a thing.

Sure, but Apple doesn't lose that if they move the Mac to ARM. You can still run iCloud for Windows and iTunes for Windows. Heck, you can run beta.music.apple.com on your Chromebook if you want.

 

[Stephen.R]

And 3) I think everything in the AppStore (where the vast majority of users download their Mac software), will likely be compatible on day one because Apple will require it and delist the app if it can’t be downloaded in an ARM/Intel version.

There’s stuff in the App Store that won’t even run on Catalina, how are you supposing Apple is going to recompile all those existing apps for a different architecture?

 

[fairuz]

Geekbench, Anandtech 's article. Along with AWS Numbers. Apple A13 Single Thread Performance. All of these are Data. You pointed to KeyDB as if Single Thread Performance doesn't matter in KeyDB and other DB while ignoring all others. There are number of company running their software on M6g at this moment.

Geekbench isn't even the same test on phone chips. You could run the same benchmark in theory, but nobody has ever shown me an ARM CPU with the same single-core performance as an i5. When I say there's no data, I mean the speculation that Apple could beef up an ARM CPU to be on similar terms as the Intel ones has no evidence.

 

[chucker23n1]

Isn't it possible that this rumor is only half true? Couldn't Apple be working on their own x86 CPU?

One, it's not clear that they could (Intel and AMD have cross-licensing agreements, and Apple would likely have to negotiate with both AMD and Intel just to get x86-64 going).

And two, it's not clear that there would be a big benefit to doing so. They'd lose out all the experience they have for ARM-specific stuff, and for what? Just to be slightly better than Intel, if even?

 

[Stephen.R]

Sure, but Apple doesn't lose that if they move the Mac to ARM. You can still run iCloud for Windows and iTunes for Windows. Heck, you can run beta.music.apple.com on your Chromebook if you want.

No, not saying they do.

but the argument was that using windows on a Mac means you aren’t/can’t use Apples services.

 

[chucker23n1]

Geekbench isn't even the same test on phone chips. You could run the same benchmark in theory, and so far there's been no ARM chip with single-core performance comparable to an i5. So there's no proof that it can be just as fast.

On the contrary, A13 beats both a Coffee Lake Refresh-S i9 and a Cascade Lake Xeon-W 3200.

(edit)

Just in case I get attacked for this later: lots of caveats to this. Only in single-core, only for short bursts, and so on.

 

[Jury]

Given that there isn't a single ARM-based Mac for sale, I would love to know what makes you come to this conclusion.



So Apple can sell more stuff if they reduce their margin? Sure. But sounds like a pyrrhic victory.
[automerge]1585407861[/automerge]
?

The cost of chip is lower with ARM-Based Processor and that will help reduce the product price without significantly affecting the margin.

 

[fairuz]

There is nothing different about work load on an N1 ARM than a x86 Desktop CPU other than the AWS CPU has many more core.

That by itself means a totally different workload. I think the AWS test was also multi-machine, but it's not clear.
[automerge]1585419508[/automerge]

On the contrary, A13 beats both a Coffee Lake Refresh-S i9 and a Cascade Lake Xeon-W 3200.

Oh, those *are* the same test on Geekbench? Then it does look comparable. Multi-core is worse but not by much if you compare to a new-ish i5.

 

[cmaier]

it’s not a small pourcentage, for example my previous company had a Windows only bug ticket report software, BUT lot of peoples like me asked for a Mac, when we need a ticket, I was booting on Mac, download the HTML report, boot on Mac and do the job. I was way more productive with MacOs
i explain my usage, but lot of people do have these kind of usage, lot of company have 20 years old software running and these software doesn’t work with MacOs.

Yes, it is a small percentage. Anecdotes are not statistics.
[automerge]1585419994[/automerge]

CISC are way more complex than ARM RISC architecture

If true, that’s even more evidence that ARM is superior.

of course, the difference between a typical x86-64 processor and an A13 is probably not all that great (we don’t know all the details of what Apple is doing, unfortunately, but, for example, I designed both the AMD K6+ and the PowerPC x704, and they were pretty similar in complexity. I also worked on UltraSparc V and designed Athlon 64, and those were pretty similar in complexity.

 

[ksec]

Geekbench isn't even the same test on phone chips. You could run the same benchmark in theory, but nobody has ever shown me an ARM CPU with the same single-core performance as an i5. When I say there's no data, I mean the speculation that Apple could beef up an ARM CPU to be on similar terms as the Intel ones has no evidence.

Which part of the Geekbench is not the same on the phone chips?

Or do you not want any synthetic benchmarks, and simply want to run Handbrake or other computational intensive Software as a reference point for it to be considered as Data or Fact?

 

[chucker23n1]

Oh, those *are* the same test on Geekbench? Then it does look comparable. Multi-core is worse but not by much if you compare to a new-ish i5.

Yeah, I was referring to single-core. Sorry, should've been clearer on that.

We know for a fact that Apple's chips don't scale as well to multiple cores as Intel's do. For example, the 2018 iPad Pro has almost identical single- and multi-core scores as the i7-7700K in the 2017 iMac. https://browser.geekbench.com/ios_devices/58, https://browser.geekbench.com/macs/407

Which is impressive if you consider that the iMac requires many times the TDP (although I'm sure the iMac can sustain this performance for longer). OTOH, the iPad Pro requires eight cores to accomplish this, rather than the iMac's four. This is in part because four of the A13's cores have lower performance, but even so, the four high-performance cores don't do as well as the four Intel cores do.

As for being the same test: Geekbench's knowledge base never seems to explicitly say that scores are comparable across architectures. But the workloads explanation specifically mentions all five platforms (Android, iOS, Linux, macOS, Windows), strongly implying that scores are meant to be platform-agnostic. So yes, I think that's the idea.

 

[fairuz]

Which part of the Geekbench is not the same on the phone chips?

Or do you not want any synthetic benchmarks, and simply want to run Handbrake or other computational intensive Software as a reference point for it to be considered as Data or Fact?

Chucker corrected me, they're the same. Geekbench's site doesn't allow any comparison between iPhones and Macs unless you scroll to page 950 of the top CPU benchmarks, so I thought they were different.
Single-core Handbrake? Sure, that's a decent test.

 

[Nütztjanix]

Isn't it possible that this rumor is only half true? Couldn't Apple be working on their own x86 CPU?

Very unlikely, since they‘d need licensing agreements with Intel and AMD to produce such a CPU.

CISC are way more complex than ARM RISC architecture

Since when is „more complex“ equivalent to „superior“?

 

[ksec]

So does Amazon N1.

Um. Sigh. No it doesn't.

It is becoming completely clear that you didn't read any of the Data provided. What is each individual test in GeekBench specifically testing ( https://www.geekbench.com/doc/geekbench5-cpu-workloads.pdf , and that is the *proper* usage of the term work load ) , and not just bunch of numbers. What did Anandtech specifically tested, which even includes single thread benchmark. And the actual AWS Data on multiple work loads. And you should properly read on https://en.wikipedia.org/wiki/Amdahl's_law

There is no point in discussing if anyone is simply refusing to read. Where there are actual Data listed in the Anandtech article.

I pointed those out in the interest for the community to learn something. But I guess that was me in good faith everyone wants to learn. May be it wasn't a wise idea to discuss anything in technical on MacRumours.

Anyway, I am out.

 

[Nütztjanix]

I fear, there are loads of people who refuse to learn. Instead, they repeat the same debunked myths over and over again just because it fits their point of view - regardless whether it’s true or not.

 

[fairuz]

Yeah, I was referring to single-core. Sorry, should've been clearer on that.

We know for a fact that Apple's chips don't scale as well to multiple cores as Intel's do. For example, the 2018 iPad Pro has almost identical single- and multi-core scores as the i7-7700K in the 2017 iMac. https://browser.geekbench.com/ios_devices/58, https://browser.geekbench.com/macs/407

Which is impressive if you consider that the iMac requires many times the TDP (although I'm sure the iMac can sustain this performance for longer). OTOH, the iPad Pro requires eight cores to accomplish this, rather than the iMac's four. This is in part because four of the A13's cores have lower performance, but even so, the four high-performance cores don't do as well as the four Intel cores do.

As for being the same test: Geekbench's knowledge base never seems to explicitly say that scores are comparable across architectures. But the workloads explanation specifically mentions all five platforms (Android, iOS, Linux, macOS, Windows), strongly implying that scores are meant to be platform-agnostic. So yes, I think that's the idea.

It's interesting that they have the same single and multi-core scores but a different number of cores. Usually single-core score times number of cores roughly equals multi-core score, but not so on the iPad. Maybe the iPad thermal-throttled even during the test.

Same performance with lower TDP sounds like a winner. Only thing is the i7 is older by a year, but whatever.

 

[Nütztjanix]

It's interesting that they have the same single and multi-core scores but a different number of cores. Usually single-core score times number of cores roughly equals multi-core score, but not so on the iPad. Maybe the iPad thermal-throttled even during the test.

Same performance with lower TDP sounds like a winner. Only thing is the i7 is older by a year, but whatever.

Apple’s A-Series CPUs don‘t have identical cores, in contrast to Intels CPUs.
Some are high performance, some high efficiency. So you can‘t just multiply single core score by number of cores.

 

[fbr$]

Please correct if I'm wrong:

If under the hood the current x86 chips are RISC + CISC-to-RISC-translater, then they are actually inferior than a pure RISC chip in terms of:

- Performance: because a pure RISC chip doesn't have to waste time making the CISC-to-RISC translation. And maybe more important, the internal space used/wasted by the CISC-to-RISC-translater could be used to make a pure RISC chip bigger, with more cores;

- Energy: because the CISC-to-RISC-translater consumes energy, it makes heat.

CISC has the advantage of producing smaller, more compact, Assembly code. But this was important when storage was expensive. Now that we have very cheap storage...

 

[ingambe]

Very unlikely, since they‘d need licensing agreements with Intel and AMD to produce such a CPU.


Since when is „more complex“ equivalent to „superior“?

read again
my reply concerned the comment about the supposed « engineering arm superiority », in term of engineering it’s a metric !