News Apple Will Move Macs to Custom Silicon, Details Transition From Intel

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

bit_user

Polypheme
Ambassador
I would phrase it a bit different.
No, I meant what I said. I think the data backs it up. ARM has been routinely delivering IPC gains in the 20-30% range for a while, now. That is much faster than x86 cores have been improving, for a long time.

But laptops are not the place where you don't care about single core performance as a lot of software is older and will be absolutely murdered by that change.
I guess you didn't see where I quoted that "we see that the A12 outperforms a moderately-clocked Skylake CPU in single-threaded performance".

That's from post #17. You replied to #20. Please try a little harder.

If they change AIR series, it should be fine, nobody sane uses it to work.
With ARM's power-efficiency advantage, it might in fact be the AIR series that has the most to gain.

I am not mentioning desktops, that would be a hilarious choice for first wave.
According to the article, developers can already get a Mac Mini with Apple Silicon. If interested, we should be able to find some reporting on their impressions, soon if not already. Apple also reportedly demonstrated a bunch of software running on it, at the event.

BTW this is most powerful supercomputer (is ARM BASED). by simple proportions it uses like ~13% more cores to get same performance. It's what you're loosing on single core speed for ability to scale, especially that at this scale x86 solutions already degrade, so I expect ~25% on single core speed in laptop scale is to be expected.
Fugaku7,299,072415.5 petaflop
Summit2,414,592148.6 petaflop
Wow. Where to start?

First, you're mixing up GPU and CPU performance. So, that data is quite simply irrelevant to anything we're discussing - Summit gets its main FLOPS from Nvidia V100 GPUs. Even the performance of Fugaku (which is impressively 100% CPUs - no GPUs or other accelerators at play) is irrelevant to this discussion, since it's using SVE which is good for HPC workloads but probably won't show up in mobile cores.

Second, it's using custom Fujitsu-designed ARM cores, so it's like using data about a system made with VIA x86 chips to say something about AMD CPUs.

Third, Summit uses POWER CPUs. So, it's not even relating to x86, as you claim.
 

bit_user

Polypheme
Ambassador
Bad news for Apple users, no more Windows compatibility. Lots of ppl dual boot Windows on their Macs.
MS supports Win10 on ARM. You can go out and buy a ARM-based Windows laptop, today. It can even emulate x86 code, though as @TerryLaze pointed out, I last heard that works only for 32-bit x86 apps.

There's also the whole emulation angle. The article cites VMs for hosting Linux and their Rosetta 2 emulator, although that doesn't sound like a recipe for very usable performance. So, let's hope Windows on ARM gets x86-64 support sorted out (if it's not already).
 
MS supports Win10 on ARM. You can go out and buy a ARM-based Windows laptop, today. It can even emulate x86 code, though as @TerryLaze pointed out, I last heard that works only for 32-bit x86 apps.

There's also the whole emulation angle. The article cites VMs for hosting Linux and their Rosetta Stone emulator, although that doesn't sound like a recipe for very usable performance. So, let's hope Windows on ARM gets x86-64 support sorted out (if it's not already).
I just looked at Surface ARM geekbench, 3800X about 2x more single thread, 4x multithread. I am not saying its impossible to catch the 3800x but by the time Apple releases arm, others will be on 5nm. Apple really need to work hard.
 

bit_user

Polypheme
Ambassador
Apple is just a customized ARM? Exynos is what?
Apple's CPUs are not customized cores, they're fully-custom cores. They're designed from the ground up. Again, it's like Intel vs. AMD. Both design chips which run the same software, but were designed in complete isolation from each other.

Exynos 9810 uses Samsung's own M3 core, which you can read about, here.


Their internal CPU design team has been disbanded, leaving Apple as the only independent designer of mobile ARM cores.

The 7300U is a very slow and old CPU,
It still uses the same Skylake micro-architecture found in all of Intel's current desktop and server CPUs. So, the comparison is very relevant, especially as it relates to IPC (Instructions Per Clock).

I still don't see how ARM even if Apple will compete with 8-64 cores from Ryzen 3000 and the upcoming Ryzen 4000. By the time Apple releases ARM, AMD will be already on Ryzen 5000, at 5nm, DDR5, and more than 64 stronger cores.
It seems clear they're targeting laptops first, where power-efficiency is king and ARM already dominates.

What remains to be seen is how well their cores can scale up their clock speed. That will be a major factor in their competitiveness on the desktop.

As for the rest, there's no reason Apple can't match (or even beat) AMD in adoption DDR5. And ARM cores generally being smaller than x86 means they can potentially deliver more cores at a lower price point (and thermal envelope).
 
Last edited:

bit_user

Polypheme
Ambassador
I just looked at Surface ARM geekbench,
Okay, I have no idea what that means, in terms of which ARM cores it has. Please include a link, so we can see the actual model of the CPU in it.

It'd be like someone talking in general about "the performance of Intel CPUs", based on data they're seeing from an old Core 2, made on 65 nm. Which core you're looking at matters a lot!

by the time Apple releases arm, others will be on 5nm.
Apple is typically the first customer on TSMC's new nodes. They're the ones leading!

Apple really need to work hard.
If you're not going to actually read my posts, I'm going to stop replying to yours. Seriously, the data I posted indicates that Apple is already where it needs to be.
 
Okay, I have no idea what that means, in terms of which ARM cores it has. Please include a link, so we can see the actual model of the CPU in it.

It'd be like someone talking in general about "the performance of Intel CPUs", based on data they're seeing from an old Core 2, made on 65 nm. Which core you're looking at matters a lot!


Apple is typically the first customer on TSMC's new nodes. They're the ones leading!


If you're not going to actually read my posts, I'm going to stop replying to yours. Seriously, the data I posted indicates that Apple is already where it needs to be.
Its the Surface X, there is only one arm cpu I believe:

View: https://amp.reddit.com/r/Surface/comments/dqiqm5/surface_pro_x_first_geekbench_scores/
 
Okay, I have no idea what that means, in terms of which ARM cores it has. Please include a link, so we can see the actual model of the CPU in it.

It'd be like someone talking in general about "the performance of Intel CPUs", based on data they're seeing from an old Core 2, made on 65 nm. Which core you're looking at matters a lot!


Apple is typically the first customer on TSMC's new nodes. They're the ones leading!


If you're not going to actually read my posts, I'm going to stop replying to yours. Seriously, the data I posted indicates that Apple is already where it needs to be.
We both can't predict the future.
We need Apple ARM Geekbench scores.
 

bit_user

Polypheme
Ambassador
We both can't predict the future.
I'm just talking about the present. The only points of speculation are about manufacturing nodes and DDR5. We can't know about DDR5, because Apple keeps those details under tight wraps, but I'm pretty sure I read that Apple has already bought up a lot of TSMC's 5 nm capacity. If you're interested, you can probably find more on that without much trouble.

We need Apple ARM Geekbench scores.
I already posted some, in post #24!

You can click the articles I linked to get more.
 

bit_user

Polypheme
Ambassador
Thanks. That seems to have Qualcomm's 8cx, which is based on a variant of the A76 cores discussed in post #24. As you can see from that post & links, they're still a ways behind the last 3 generations of Apple's cores.

Also, why would you compare a tablet to a 105W desktop CPU (Ryzen 7 3800X)??? What's that supposed to prove?

If you want a more reasonable comparison, I think MS recently released a Surface based on a Ryzen APU.
 
I guess you didn't see where I quoted that "we see that the A12 outperforms a moderately-clocked Skylake CPU in single-threaded performance".
I don't believe it moderately-clocked Skylake CPU means anything but the 5W intel cpu.
Its deliberate that they don't mention what is it AND any marketing would not agree for moderately-clocked comment unless they really cannot squeeze more from the cpu.
And that's why I mentioned the AIR series that is using 5W but not the 28W+ from pro series.
According to the article, developers can already get a Mac Mini with Apple Silicon. If interested, we should be able to find some reporting on their impressions, soon if not already. Apple also reportedly demonstrated a bunch of software running on it, at the event.
Mac mini is a performance joke even with intel, I have one for whatever that have to run though their OS.
It will be even worse with ~3W 12Z chip.

you can put whatever you want in whatever format, but performance will be similar like all other formats with same chip.
I expect much more from desktop, than from handheld phone/tablet.

Rdslw said:
If they change AIR series, it should be fine, nobody sane uses it to work.
With ARM's power-efficiency advantage, it might in fact be the AIR series that has the most to gain.

that's why I mentioned it might be ok with AIR for first wave, but desktop is very missmatched choice.

unless ofc, they really plan to go from 4/6 core design to 24/32 instantly.
 
Last edited:

bit_user

Polypheme
Ambassador
I don't believe it moderately-clocked Skylake CPU means anything but the 5W intel cpu. Its deliberate that they don't mention what is it
OMG. What will it take for you to actually read a link?? The data is in there, but I guess you'd rather argue than learn anything.

Of course, there's a small IQ test here, because the actual data he's referring to is via a second link. It's right there in my original quote, but since you missed that, I expect you'll never find the link.

Mac mini is a performance joke even with intel, I have one for whatever that have to run though their OS.
It will be even worse with ~3W 12Z chip.
You missed the point, which was that because the mini's are now getting released, people can actually try it and you can read firsthand accounts of how it performs. But, that was before I realized that you'd probably rather not have actual facts and simply go on believing whatever best fits your worldview.

unless ofc, they really plan to go from 4/6 core design to 24/32 instantly.
With AMD's chiplet approach, it wouldn't be hard to scale up.
 

Chung Leong

Reputable
Dec 6, 2019
493
193
4,860
Bad news for Apple users, no more Windows compatibility. Lots of ppl dual boot Windows on their Macs.

Dual-booting is completely impractical for everyday use. Even with games, I think most people would choose to pay the performance penalty of virtualization.

Inability to host x86 VM is gonna be a problem even for those who only use Mac software. Imagine you're a web developer. You use Docker to create your dev environment. You upgrade to a new Macbook and to your dismay, you find your work setup is now totally broken due to missing container images.
 
  • Like
Reactions: Zizo007

Chung Leong

Reputable
Dec 6, 2019
493
193
4,860
MS supports Win10 on ARM. You can go out and buy a ARM-based Windows laptop, today. It can even emulate x86 code, though as @TerryLaze pointed out, I last heard that works only for 32-bit x86 apps.

I don't think the ARM version of IE supports ActiveX. Oftentimes that's the reason why Mac users need Windows: they need to access some ancient intranet site.
 
OMG. What will it take for you to actually read a link?? The data is in there, but I guess you'd rather argue than learn anything.

Of course, there's a small IQ test here, because the actual data he's referring to is via a second link. It's right there in my original quote, but since you missed that, I expect you'll never find the link.
Only data I see is the mobile codec that have little sense in desktop. its not apple-to-apple comparison, so I did not want to use it as an reference.
Any performance benchmarks on a77 is compared to 4 year old 7300u AND that 5nm part is like 20% faster than 4 yeard old budget mobile part.
so lets use it as reference point and compare to ryzen 4800u
https://cpu.userbenchmark.com/Compare/Intel-Core-i5-7300U-vs-AMD-Ryzen-7-4800U/m223355vsm1005639
30-250% faster. so ARM 5nm part is loosing already.
and my point:
we did not even get to the desktop parts.
but if you want to:
https://cpu.userbenchmark.com/Compare/Intel-Core-i5-7300U-vs-Intel-Core-i9-10900K/m223355vs4071
70-800% slower.
but that's a random graph with random, maybe incomplete image based on FAVORABLE for arm data
(who would be marketing with whatever is not showing them in good light?)

and closest comparison that we did get with
Snapdragon 835
with A73 cores:
which from this graph should be 2.5x slower:
View: https://i.imgur.com/Q707MFkl.png

so
https://static.techspot.com/articles-info/1599/images/11.png
0.3 single core x 2.5 = ~0.8
Which is LOWER than 5Y old 5W intel i7.
https://ark.intel.com/content/www/p...3-6y30-processor-4m-cache-up-to-2-20-ghz.html
and multi core:
1.6 x 2.5 which is unreliable by core count but whatever:
4 ~~ 8550u which is 2 year old MOBILE 15W part.

so what we will get is ~5W performance on single chip
and ~15-25W on multi core.

another graph:
https://static.techspot.com/articles-info/1599/images/12.png
similar result
performance in between 5W (single) and 25W (multicore)
Which is horrible for DESKTOP, and bad for PROFESSIONAL MOBILE LAPTOP as I mentioned if you would read. AND THAT WAS MY POINT.

You missed the point, which was that because the mini's are now getting released, people can actually try it and you can read firsthand accounts of how it performs. But, that was before I realized that you'd probably rather not have actual facts and simply go on believing whatever best fits your worldview.
box is for devs to prepare their software to use correct drivers that does not translate though x86 to avoid loosing 30% from the start.
it's not meant for wider audience. Its for benchmarking for everyone who make software for apple to prepare themselves and avoid issues windows had on first ARM based box.
 

bit_user

Polypheme
Ambassador
Only data I see is the mobile codec that have little sense in desktop. its not apple-to-apple comparison, so I did not want to use it as an reference.
I don't know if you're arguing in bad faith or just have useless web skills, but either way this can't continue.

The data cited was SPEC CPU2006 numbers, and the CPU was a Xeon 8176. I don't expect you to take my word for it, so I guess we're done, here.

I think you'll be in for a surprise, when Apple's new products hit the market, presuming your apparent bias doesn't still prevent you from seeing what they've achieved. You'll probably just keep arguing about the performance of irrelevant cores that have nothing to do with what Apple is even using.

box is for devs to prepare their software
I know what it's for. My point was that it will be the first chance for the general public to run desktop software on their A12 (which isn't even the latest - A13 has already been out for a bit) and get real reactions from people.

I should add: because I think it's important, that I don't even like Apple! I've never owned an Apple product in my entire life, and I don't foresee that ever changing. But, I'm simply unwilling to let my dislike of them, their walled garden philosophy, and their extortion of developers and users alike blind me to the technical achievements they've made. Seeing their technical progress for what it is can teach us all something about the potential of the ARM ISA and reveal the approaching sunset on the dominance of x86.
 
Last edited:
  • Like
Reactions: Rdslw
I don't know if you're arguing in bad faith or just have useless web skills, but either way this can't continue.

The data cited was SPEC CPU2006 numbers, and the CPU was a Xeon 8176. I don't expect you to take my word for it, so I guess we're done, here.

I think you'll be in for a surprise, when Apple's new products hit the market, presuming your apparent bias doesn't still prevent you from seeing what they've achieved. You'll probably just keep arguing about the performance of irrelevant cores that have nothing to do with what Apple is even using.


I know what it's for. My point was that it will be the first chance for the general public to run desktop software on their A12 (which isn't even the latest - A13 has already been out for a bit) and get real reactions from people.

I should add: because I think it's important, that I don't even like Apple! I've never owned an Apple product in my entire life, and I don't foresee that ever changing. But, I'm simply unwilling to let my dislike of them, their walled garden philosophy, and their extortion of developers and users alike blind me to the technical achievements they've made. Seeing their technical progress for what it is can teach us all something about the potential of the ARM ISA and reveal the approaching sunset on the dominance of x86.

I am judging by 4 factors:
performance numbers (mid-low)
upgradability (non existing)
build quality (superb)
price (very high)
and I dont care what people say, once I see the benchmarks showing that it's not slower than previous gen,
I will take back my word, but based on numbers I have now and their 2.5x promise my expectations are very low, especially for desktop. In my opinion it will be glorified pricy paperweight.