What Today's 27" Intel iMac Refresh Tells Us About the Apple Silicon Mac Release Order

[sublunar]

Well the newer W2200 series Xeons, an RDNA 2 GPU from AMD when they’re ready later this year and miniLED display with the nanotexture screen would be on my wish list for a 1Q2021 iPad Pro. I still believe lol.

Let's point out that Intel are discontinuing the Skylake Xeon CPUs in the current iMac Pro with orders being stopped in January 2021 before the final products are shipped in July 2021. Vega56/64 is also on the chopping block and might be unavailable sooner - AMD must be doing Apple a favour if they still keep making them until deep into next year.

And if you look at the 2013 Mac Pro as a template, they bumped the base spec model to match the upper SKU before discontinuing it shortly before the 2019 Mac Pro came out. I reckon the iMac Pro will go away when the Apple Silicon Mac Pro comes out.

 

[funkahdafi]

I think your information might be a bit outdated. Fact is: current Apple mobile CPUs already offer better per-core performance than Xeon/EPYC.

You say that's a fact as if you knew more than the rest of us. Care to share your sources? And don't give me those synthetic benchmarks that say nothing about real world performance, please.

Also, "current Apple mobile CPU" - which one?
"better per-core performance than Xeon/Epyc" - which one?

An Apple A13 running at 2.5Ghz is more or less equivalent to a Intel Skylake running at 3-3.3 Ghz.

Right. So you are comparing a 7nm 2019 Apple CPU to a 14 nm Intel 2015 architecture, one that was already outdated when it was released. Sounds like a fair comparison.

What you should be comparing against (since we are talking about the Mac Pro here) is AMD's Zen 4 architecture and its Epyc variant, which will hit the streets shortly before Apple allegedly brings their magic ARM Xeon killer.

Now, there are obvious a lot to "buts" and "ifs" here.

Another big question mark is whether Apple has the interconnect technology to efficiently wire a bunch of cores.

We don't know what Apple has in store this year.

This, this and that. Nobody here knows anything and yet every iPad fanboy suddenly seems to be an expert in CPU tech. if you know anything about the tech, you'll realize just how unlikely it is Apple will release an ARM CPU that will outperform top end 2022 HEDT CPUs. It's not going to happen. Not by 2022.

In regards to consumer computing, they are a step ahead of AMD and two steps ahead Intel right now.

Again, you make it sound like you knew more than everybody else. Can you back up your claim that Apple is ahead of AMD? In what area exactly? And by what margin? Sources?

Also, we were talking about the Mac Pro, not about "consumer computing".

Based on their progress so far and the properties of their architecture, I do see reason for cautious optimism.

What are the properties of their architecture? Can you point them out for me, please? A link to the architecture documentation will suffice. Thanks!

Look, if Apple manages to pull this off, I am as happy as the next guy, but you guys all need to make a reality check and take off those rose-colored glasses. Apple produces great tech, but they are not magicians. This stuff is way more complicated than slapping a few more CPU cores in the die. And we haven't even talked software yet.

 

[jlocker]

This is why I say that is the best time to buy a higher end Mac. Get the AppleCare, use it for three years then when AppleCare has run out. Take it to a Apple store trade it in for credit, use your Apple Card and get a new model with 12 months to pay of the new computer.

 

[Realityck]

The Tim Cook quote says very clearly that within 2 years, “Apple Silicon will bring amazing technologies, industry leading performance, and a common architecture across all of our products.”

A common architecture across all of our products isn’t ambiguous.

Yes but it could also be interpreted as across all our product lines, not that every computer that is offered would use Apple Silicon.

Its a joke to say that within 2 years Apple will completely abandon high end workstations using their own unproven expanded ARM SoC's to replace multiple CPU's and high end GPU's along with some of the outboard interfaces they use for professional video editing, photography, and like. Apple couldn't make a new Mac Pro for years, why is it so simple now?

 

[leman]

You say that's a fact as if you knew more than the rest of us. Care to share your sources? And don't give me those synthetic benchmarks that say nothing about real world performance, please.

Yes, synthetic benchmarks. What does qualify as ”real world performance“ to you? We have a lot of independent comparison points, as imperfect as they might be, SPEC benchmarks, Geekbench, browser tests, various algorithms people were running on iOS vs x86 etc. All them point towards the same conclusion: current Apple mobile CPU are comparable with what desktop-class x86 can do in terms of core-to-core performance.

Also, "current Apple mobile CPU" - which one?
"better per-core performance than Xeon/Epyc" - which one?

A13, Cascade Lake, Rome

Right. So you are comparing a 7nm 2019 Apple CPU to a 14 nm Intel 2015 architecture, one that was already outdated when it was released. Sounds like a fair comparison.

I’m sorry, what should I compare it to? I though we are talking about Xeons? Latest Xeons are still built as an iteration of Skylake SP, which in turn is Skylake with an AVX-512 patched in. It’s funny that you call Skylake outdated in 2015 while praising a Intel at the same time. Look into their roadmap. They’ve been milking this architecture for 5 years and they will still milk it for another two.

If you want to compare it to 10nm sunny cove, sure, it doesn’t change much. The two year old A12Z outperforms most if not all Ice Lake SKUs.

What you should be comparing against (since we are talking about the Mac Pro here) is AMD's Zen 4 architecture and its Epyc variant, which will hit the streets shortly before Apple allegedly brings their magic ARM Xeon killer.

Kind of hard to do that not knowing how Zen 4 will turn out, is it? I am comparing based on data we have about current architectures. Maybe AMD will manage some sort of dramatic breakthrough. Who knows.

Also, who is even talking about a Xeon killer? Apple is not interested in server platform, their Mac Pro is a hybrid HPC workstation. Thats why they use the W variants and not a server chip such as EPYC.

This, this and that. Nobody here knows anything and yet every iPad fanboy suddenly seems to be an expert in CPU tech. if you know anything about the tech, you'll realize just how unlikely it is Apple will release an ARM CPU that will outperform top end 2022 HDTE CPUs. It's not going to happen. Not by 2022.

I do hope to know a bit about tech, and I don’t see why this should be impossible. They have the microarchitecture, the have the cache, they have the power management. The only question is whether they have the interconnect.

Also, we were talking about the Mac Pro, not about "consumer computing".

And what does that mean exactly? Mac Pro as in build server? Mac Pro as in render farm? Mac Pro as in video editing? Mac Pro as in scientific HPC platform? No, current Apple CPUs won’t even come close to AVX-512 optimized number cruncher that runs on an Intel Xeon (neither will AMD btw). For other tasks such as software development? Sure.

What are the properties of their architecture? Can you point them out for me, please? A link to the architecture documentation will suffice. Thanks!

Anandtech has some in-depth analysis. If you want official documentation you should ask Tim.

 

[JMacHack]

Apple couldn't make a new Mac Pro for years, why is it so simple now?

Nobody says it's simple. We're saying it's possible.

 

[Realityck]

Nobody says it's simple. We're saying it's possible.

This reminds me of the old project management triangle from the fifties.
Back then it was known by these rules

1. The quality of work is constrained by the project's budget, deadlines and scope (features).
2. The project manager can trade between constraints.
3. Changes in one constraint necessitate changes in others to compensate or quality will suffer.

Most people are familiar with these signs.

While this should be looked at as a bit of humor, technology hurdles can be indeed challenging. No one can contest Apple's attention to detail on putting out great products however, this is a very high hurdle to achieve to make a Apple Silicon based Mac Pro that compares to todays models.

 

[jdb8167]

Yes but it could also be interpreted as across all our product lines, not that every computer that is offered would use Apple Silicon.

Its a joke to say that within 2 years Apple will completely abandon high end workstations using their own unproven expanded ARM SoC's to replace multiple CPU's and high end GPU's along with some of the outboard interfaces they use for professional video editing, photography, and like. Apple couldn't make a new Mac Pro for years, why is it so simple now?

One thing about Apple and keynotes, they are carefully scripted and every word is chosen for clarity. If Apple meant product lines, they would have said product lines. What they said is they will have a common architecture across all their products.

I see no reason to doubt what Tim Cook and Apple have said and I take them literally.

 

[Realityck]

One thing about Apple and keynotes, they are carefully scripted and every word is chosen for clarity. If Apple meant product lines, they would have said product lines. What they said is they will have a common architecture across all their products.

I see no reason to doubt what Tim Cook and Apple have said and I take them literally.

You have too much faith. Technology progression doesn't always go smoothly on very complex products. But we'll see.

 

[jdb8167]

You have too much faith. Technology progression doesn't always go smoothly on very complex products. But we'll see.

I guarantee Apple already has prototypes in their labs for every Mac model that they intend to replace (vs. retire). Two years is a pretty short cycle for new products. It took took more than 2 years from the original discussion of the Mac Pro with journalists until its release in late 2019. Apple already knows what performance they will get from the Mac Pro replacement at least to a reasonable degree.

 

[smoking monkey]

Good breakdown.
The one thing I would question is the stipulation that the new ARM chip has to be significantly faster than the previous Intel chip. I don't believe this to be true for laptops. I think if Apple can get an ARM chip out in a laptop that is equal to or just a little faster, but moderate to significantly cooler than Intel, that is all that matters.

IN saying that, I wouldn't be surprised if the 16 Arm isn't release until July-Oct next year. I hope not, but I'm prepping for that.

 

[anshuvorty]

Good breakdown.
The one thing I would question is the stipulation that the new ARM chip has to be significantly faster than the previous Intel chip. I don't believe this to be true for laptops. I think if Apple can get an ARM chip out in a laptop that is equal to or just a little faster, but moderate to significantly cooler than Intel, that is all that matters.

IN saying that, I wouldn't be surprised if the 16 Arm isn't release until July-Oct next year. I hope not, but I'm prepping for that.

Good point. I am sure most laptop users will gladly give up some performance points for significantly increased battery life if they had the choice.

 

[johngwheeler]

There is no way in hell that ARM based CPUs will match the performance of high end x86 processors like Xeon within just two years. That would "break" Moore's law by such a wide margin that it is next to inconceivable. Even Apple's marketing is limited by physics, not to mention their engineers. And don't forget, high end x86 CPUs will make substantial progress in these two years as well. Not just Intel's Xeon, but also (and probably more importantly) AMD's Epyc.

Given there are already 80 and 128-core ARM CPUs in production from Ampere, Marvell, AWS which roundly beat the best AMD Epyc & Xeon CPUs in the data center on compute/watt and are close or better in absolute performance, I don't think it is at all inconceivable that an ARM-based desktop CPU can beat the best workstation Xeon or even AMD Threadripper within a couple of years.

ARM is still not best in all workloads, and needs more development, but it is already competing and in some cases winning against AMD & Intel.

Go and do some research.

 

[johngwheeler]

I remember that article. It was based in synthetic benchmarks and was highly biased. ARM performance would literally need to make at least a 10-20x performance jump to get anywhere near the levels of Xeon/Epyc performance. Something that takes 10 years under normal conditions. Even if Apple managed to get there faster, we are most definitely not talking 2 years. Unless you are happy with the performance levels of some low end Xeons or 8th gen Intel consumer CPUs. We're talkin Mac Pro here, not iMac or Macbook Pro (which both use mobile CPUs, to top it all off).

Again, think about what Apple said: Intel Macs are here to stay for *many* years to come.

Check out some real-world tests:

Stacking Up Arm Server Chips Against X86

It is pretty clear at this point that there is going to be a global recession thanks to the coronavirus outbreak. Maybe it will be a V-shaped recession

www.nextplatform.com

Benchmarking the AWS Graviton2 with KeyDB – M6g up to 65% faster | KeyDB - The Faster Redis Alternative

We’ve always been excited about Arm so when Amazon offered us early access to their new Arm based instances we jumped at the chance to see what they could do. We are of course referring to the Amazon EC2 M6g instances powered by AWS Graviton2 processors. The performance claims made and the hype...

docs.keydb.dev

There are plenty of other benchmark results out there. I'm not saying that they are all completely objective (some are from manufacturers, not independent testers, and aim to highlight the best possible performance).

ARM does poorly in some workloads, but does well in terms of price/performance, which is much more of a concern in the data center than it is for consumer machines.

My point is that ARM is much closer to AMD & Intel in raw performance that you seem to think. I expect that Apple will be optimizing their SoC for their hardware to address the weaker areas of ARM architecture. In any case, Apple is just using the ARM ISA, so could be producing something that is totally different to ARM chips from Ampere, Marvell, AWS etc.

What seems clear is that Apple will not release computers that are slower than the current Intel models, and they are confident that they can do this within 2 years.

Time will tell if they succeed, but I would bet that they will.

 

[leman]

Given there are already 80 and 128-core ARM CPUs in production from Ampere, Marvell, AWS which roundly beat the best AMD Epyc & Xeon CPUs in the data center on compute/watt and are close or better in absolute performance, I don't think it is at all inconceivable that an ARM-based desktop CPU can beat the best workstation Xeon or even AMD Threadripper within a couple of years.

ARM is still not best in all workloads, and needs more development, but it is already competing and in some cases winning against AMD & Intel.

As it has been pointed out over and over again, instruction set doesn’t matter much. Microarchitecture does. I wouldn’t draw too many conclusions from ARM datacenter chips, simply because they are not made by Apple. To put differently, projecting From Ampere or A64FX to Apple Silicon is as useful as projecting from AMD EPYC to Apple Silicon.

Another thing is that Apple has no need to compete with data center CPUs. Server CPUs trade single threaded performance for the ability to execute many jobs at the same time - either asymmetrically (web services, virtualization - there is a reason why AMD is pursuing four way SMT) or symmetrically (supercomputing for which A64FX is a prime example). Apple however needs both single and multithreaded performance in the Mac Pro, is why they use Xeon-W with its wide range of turbo boost. The Mac Pro CPU needs to be able to adapt to hybrid workflow. Not nessesarily what data center class CPUs are good at.

 

[funkahdafi]

Given there are already 80 and 128-core ARM CPUs in production from Ampere, Marvell, AWS which roundly beat the best AMD Epyc & Xeon CPUs in the data center on compute/watt and are close or better in absolute performance, I don't think it is at all inconceivable that an ARM-based desktop CPU can beat the best workstation Xeon or even AMD Threadripper within a couple of years.

ARM is still not best in all workloads, and needs more development, but it is already competing and in some cases winning against AMD & Intel.

Go and do some research.

I am not sure your last comment was really necessary. I have commented on these monstrous 128 core ARM CPUs in this thread before, but I will gladly bring you up to speed. These CPUs you are mentioning have several issues:

1.) Cost. They are prohibitly expensive and not viable for consumer or prosumer devices. Put one of these in a Mac Pro and its pricing becomes even more ridiculous than it already is. By a very hefty margin. You will spend as much as you spend for a shiny new BMW. And once you have that, you will be massively disappointed, because:

2.) Amdahl's Law. Google it. It's a scientific thesis describing the scaling of CPU cores. In a nutshell, at a certain point, more CPU cores does not equal more performance. In other words, a 128 core CPU is not necessarily faster than a 64 core CPU. At least not in consumer computing. Your Photoshop won't perform better just because someone slapped another 100 cores onto a CPU die. That's not how it works. Again, since you pointed out I should do my research, I suggest you get familiar with Amdahl's Law.

And just to mention it again: I am not saying Apple can't reach competitive performance levels. All I am saying is that it won't happen within just two years.

EDIT: Also, refer to what leman said just one post above this one.

 

[funkahdafi]

Do yourself a favor and watch this:

 

[jinnyman]

One thing about Apple and keynotes, they are carefully scripted and every word is chosen for clarity. If Apple meant product lines, they would have said product lines. What they said is they will have a common architecture across all their products.

I see no reason to doubt what Tim Cook and Apple have said and I take them literally.

Yeah right.
What if MP 7,1 is Apple's last MP?
You don't know, and I also don't know. It's all Apple's word quiz.

 

[Waragainstsleep]

Delete me

 

[Waragainstsleep]

And just to mention it again: I am not saying Apple can't reach competitive performance levels. All I am saying is that it won't happen within just two years.

I suspect they have already been at it for a number of years. Its not like Apple didn't know they'd need a workstation class CPU. And we know they are more than capable of playing the long game. They ran OS X on x86 for at least 5 years before anyone even suspected it was a possibility they would switch to Intel. Back then they had far less in the way of resources, if anything their long term planning will be even more magnificent now.

Yeah right.
What if MP 7,1 is Apple's last MP?
You don't know, and I also don't know. It's all Apple's word quiz.

New CPU, new model names (for some).

iMac = Internet Mac
gMac = Gaming version of the iMac
UltraMac = Low end Mac Pro replacement
HyperMac = Mac Pro replacement

 

[Shivetya]

if they do release a 24 this fall I will buy it to go alongside my 2019 i9 iMac as both a comparison and possibly a future system to hand off to my parents to see how well they adapt to using a Mac

 

[JMacHack]

While this should be looked at as a bit of humor, technology hurdles can be indeed challenging. No one can contest Apple's attention to detail on putting out great products however, this is a very high hurdle to achieve to make a Apple Silicon based Mac Pro that compares to todays models.

Believe me, I'm well familiar with the "good, fast, cheap" dilemma. And I don't dispute that this will be a challenge, and while I do believe this is something that Apple can accomplish, I won't rule out the possibility that they might fail.

Yes, it's a high hurdle, that's probably why the Mac Pro will be the last to transition to Apple Silicon. But I don't believe for a minute that Apple would take the gamble of moving their entire lineup to their own processors if they didn't think they could beat Intel. Apple's betting the entire Mac lineup on this.

Yeah right.
What if MP 7,1 is Apple's last MP?
You don't know, and I also don't know. It's all Apple's word quiz.

You're right, we don't know for 100% certain. However, after the huge "mea culpa" in 2017, and forming a "team" to determine what the pro Mac market wants I find it difficult to think that they'd just say "nah we're done here" after 3 years. If they were gonna drop the Mac Pro, dropping the 6,1 and sticking with the iMac Pro is what they would've done and they wouldn't have spent millions in R&D to do it.

Conservatively, I'd put it at an 85% chance they plan on keeping the Mac Pro, even after the ASi move.

 

[jinnyman]

You're right, we don't know for 100% certain. However, after the huge "mea culpa" in 2017, and forming a "team" to determine what the pro Mac market wants I find it difficult to think that they'd just say "nah we're done here" after 3 years. If they were gonna drop the Mac Pro, dropping the 6,1 and sticking with the iMac Pro is what they would've done and they wouldn't have spent millions in R&D to do it.

Conservatively, I'd put it at an 85% chance they plan on keeping the Mac Pro, even after the ASi move.

Well i don't disagree on that much, and I personally hope Apple continue offering MP. But then, given how loooooooooooooong Apple takes to build a desktop of MP caliber, I doubt Apple's got ability to do it in timely fashion

 

[Realityck]

I suspect they have already been at it for a number of years. Its not like Apple didn't know they'd need a workstation class CPU. And we know they are more than capable of playing the long game. They ran OS X on x86 for at least 5 years before anyone even suspected it was a possibility they would switch to Intel. Back then they had far less in the way of resources, if anything their long term planning will be even more magnificent now

You do know that NextStep what OS X is based on was running on both Motorola and Intel processors before Steve Jobs rejoined Apple? In fact they ran it on multiple processors. This was a normal process with gauging performace on other processor platforms in case you needed to change the processors you use.
“NeXTStep ran on PA-RISC, Intel x86, Motorola 68K and SPARC architectures since well before 2000.”
Apple was having a terrible time trying to develop a next generation OS, and if you look at the history of Apple, before they choose NextStep over Gassée BeOs you can see it’s wasn’t long term planning that they were good at.

 

[transpo1]

What about GPUs? Does anyone think, due to the increased power savings, that they will be able to include dedicated graphics in the Arm MBP 13"?