Discussion Quo vadis Apple Macs - Intel, AMD and/or ARM CPUs? ARM it is!

Page 17 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
Due to popular demand I thought somebody should start a proper thread on this pervasive topic. So why not do it myself? ;)

For nearly a decade now Apple has treated their line of Mac laptops, AIOs and Pro workstations more of a stepchild. Their iOS line of products have surpassed it in market size and profit. Their dedicated Mac hardware group was dissolved. Hardware and software updates has been lackluster.

But for Intel Apple clearly is still a major customer, still offering custom chips not to be had outside of Apple products. Clearly Intel is eager to at all costs keep Apple as a major showcase customer.

On the high end of performance Apple's few efforts to create technological impressive products using Intel parts increasingly fall flat. The 3rd gen of MacPros going up to 28 cores could have wowed the audience in earlier years, but when launched in 2019 it already faced 32 core Threadripper/Epyc parts, with 64 core updates of them already on the horizon. A similar fate appears to be coming for the laptops as well, with Ryzen Mobile 4000 besting comparable Intel solutions across the board, with run of the mill OEMs bound to surpass Apple products in battery life. A switch to AMD shouldn't even be a big step considering Apple already has a close work relationship with them, sourcing custom GPUs from them like they do with CPUs from Intel.

On the low end Apple is pushing iPadOS into becoming a workable mutitasking system, with decent keyboard and, most recently, mouse support. Considering the much bigger audience familiar with the iOS mobile interface and App Store, it may make sense to eventually offer a laptop form factor using the already tweaked iPadOS.

By the look of all things Apple Mac products are due to continue stagnating. But just like for Intel, the status quo for Mac products feels increasingly untenable.
 
  • Like
Reactions: Vattila

lobz

Platinum Member
Feb 10, 2017
2,057
2,856
136
The performance will be interesting to see. Intel has been stagnant for years, and still stuck on 10nm for most chips. TSMC is building 5nm chips for Apple.

As for the transition, the PPC to Intel transition went very smoothly. I see no reason why this won't also go smoothly. Although its a bit comical that Apple has gone from CISC (68k), to RISC (PPC), to CISC (Intel), and now back to RISC (ARM).
I'm pretty sure you wanted to say 'stuck on 14nm for most chips'.
 
  • Like
Reactions: Stuka87

uzzi38

Platinum Member
Oct 16, 2019
2,565
5,575
146
First step is 18 core Xeons, a relatively low bar to set in the long run.

And they have something for that. But don't expect something to handle 64 core EPYC or Threadripper so soon, the first target is smushing those Xeons.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
The GPU also is in another world.

13mm2 GPU in the Tablet A12Z performing at least like Intel's Iris Plus G7 graphics or Vega 7 in AMD's APU. 3x perf/area. It's not just the CPU ARM vendors excel in!

The primary issue with x86 lies with the vendors in that both Intel/AMD completely falls flat on their faces every few years.
 
  • Like
Reactions: Etain05

blckgrffn

Diamond Member
May 1, 2003
9,111
3,029
136
www.teamjuchems.com
I think it's interesting they'll be fully competing with the Intel and AMD of ~18 months from now.

Especially on the Intel front, it seems somewhat likely they get their manufacturing woes straight (I mean, percentage chance has to be pretty decent?). Will we be seeing DDR5 based CPUs at that point from all vendors? I felt like we were going to have another CPU performance lul after pushing to some leading edge manufacturing nodes, but maybe not.

Then again, Mac computer sales were a total of ~ 18M units in 2019.


Compare that to 30.7 million Apple watches... which can't be sold competitively with a third party chip.

This seems like a pretty big investment long term to gain some margins, further diffuse their CPU investment and decrease risk of their partners really dropping the ball.

I am not sold that this transition brings as much value to their personal computing products as it does to their mobile and wearable divisions.

(For further perspective, worldwide PC shipments in 2019 were ~261M units)
 

Eug

Lifer
Mar 11, 2000
23,583
996
126
I think it's interesting they'll be fully competing with the Intel and AMD of ~18 months from now.

Especially on the Intel front, it seems somewhat likely they get their manufacturing woes straight (I mean, percentage chance has to be pretty decent?). Will we be seeing DDR5 based CPUs at that point from all vendors? I felt like we were going to have another CPU performance lul after pushing to some leading edge manufacturing nodes, but maybe not.

Then again, Mac computer sales were a total of ~ 18M units in 2019.


Compare that to 30.7 million Apple watches... which can't be sold competitively with a third party chip.

This seems like a pretty big investment long term to gain some margins, further diffuse their CPU investment and decrease risk of their partners really dropping the ball.

I am not sold that this transition brings as much value to their personal computing products as it does to their mobile and wearable divisions.

(For further perspective, worldwide PC shipments in 2019 were ~261M units)
Well, I could see this meaning their prices drop, comparatively.
 

Richie Rich

Senior member
Jul 28, 2019
470
229
76
I think it's interesting they'll be fully competing with the Intel and AMD of ~18 months from now.
Complete picture is that they will have to compete also with ARM's Cortex X1 core introduced receintly (which has 140% PPC of Zen2 already, 30% IPC jump over A77 is huge) and Cortex X2 next year (based on Matterhorn ARMv9 2048-bit SVE2 vectors). Probably some new versions of laptop Snapdragon 8cx will come based on X-series to MS Surface and other laptops. AMD and Intel has problems with relatively old A76 in Graviton2 server CPU. It will be interesting to see how Intel and AMD will be competing to new Cortex X1 cores and Apple's A14 cores. That battle will be epic. Such a clash was not in IT for long time.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Well, I could see this meaning their prices drop, comparatively.

Well their margins have been dropping. They probably sell it at the same price but give discounts based on vendors bundling bunch of other products. It's a finance trick to make it look like the revenue is high yet be price competitive.

At the peak they were at 62-65%. Now they are at 56-59%. That's a quite a big reduction.

Their prices for server and notebook chips are very misleading. They were selling a laptop with 1035G4, a $320 RCP part for $399 last Black Friday. For the most part it was the 1035G1, which is $297 but still. Actual price must have been in the $50-70.

The motherboard that included the Pentium Silver J5005 sold for $120. The RCP is $161 for the J5005. The chip may be $30.

How much are they discounting their "$13000" server chips?
 

HurleyBird

Platinum Member
Apr 22, 2003
2,670
1,250
136
I expect Macs will gain marketshare now unless AMD and Intel can keep up, but at the same time I also don't think it's a given that, say, Zen 4 will be slower than A15 (or whatever is around during the same time period). My best guess, albeit with huge error bars, is that AMD will have the ultimate performance crown while Apple will have the best laptop chips.

Also, I bet Apple will try breaking into servers again, especially if the Mac Pro remains a thing.
 

Doug S

Platinum Member
Feb 8, 2020
2,203
3,405
136
My best guess, albeit with huge error bars, is that AMD will have the ultimate performance crown while Apple will have the best laptop chips.


Well it depends, are you talking about fastest single thread performance or fastest multithread performance? I wouldn't be at all surprised if Apple wins in the first. Especially because Apple is getting their new chips on TSMC's latest process faster than AMD, so while Apple will be shipping 5nm CPUs in the Macs sold six months from now AMD will still be on 7nm. When AMD is moving to 5nm Apple will be on 5nm+, when AMD is on 5nm+ Apple will be on 3nm.

Multithread is a different story. I don't see Apple even trying to compete with the number of cores AMD is doing on their high end. The Mac Pro is already a niche market for Apple, and matching AMD with 64 cores would be a niche market of a niche market that isn't likely to be pursued. It will be interesting to see how Apple does if they have for instance 32 cores (which I'd expect to be done using four 8 core chips, no way Apple designs a monolithic chip for the Mac Pro alone) versus AMD and Intel's latest with the same 32 core count.

There's zero chance it wins when the Mac is running macOS. It just hasn't had the work to support multiple cores that Windows and especially have Linux have, because Apple doesn't sell servers so it hasn't all the extra reason the server market provides to make it more efficient at high core counts. If both are running the same OS and same number of cores I think it would be an interesting battle.
 

blckgrffn

Diamond Member
May 1, 2003
9,111
3,029
136
www.teamjuchems.com
I expect Macs will gain marketshare now unless AMD and Intel can keep up, but at the same time I also don't think it's a given that, say, Zen 4 will be slower than A15 (or whatever is around during the same time period). My best guess, albeit with huge error bars, is that AMD will have the ultimate performance crown while Apple will have the best laptop chips.

Also, I bet Apple will try breaking into servers again, especially if the Mac Pro remains a thing.

I personally don't think it will move market share all that much. I think it will increase the value of Apple as a company, both tangibly and intangibly as it gives them further secret sauce and the ability to customize their products for performance AND margins. I always thought Apple was really smart with focusing on keeping their memory footprint low on their mobile devices, for example.

Keeping the amount of ram small by compressing memory and keeping memory overhead low was not really a performance advantage, but when you sell hundreds of millions of phones per year every $ in the BoM adds up dramatically. I suspect that most of their PC skus will similarly be "good enough" to be competitive and make them solid profits.

Lastly, I don't think most people care what chip is in the thing. If they want a Mac, they probably already bought a Mac. If they are in the vast majority that didn't care about MacOS or getting a "premium" Apple product for a PC, then I doubt this makes an iota of difference.
 

blckgrffn

Diamond Member
May 1, 2003
9,111
3,029
136
www.teamjuchems.com
Well, I could see this meaning their prices drop, comparatively.

Maybe, my guess is that they return value to the shareholders by keeping prices the same and increasing their margins as they are able.

As a shareholder, I'd be I want them to increase their mobile market share by 5% than spending any time at all increasing their PC market share. Apple reliably takes care of their shareholders.

(Forgot that I hold Apple stock through other means :p)
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136

3D performance under binary translation, at least for 3DMark has minimal loss compared to native.

Where did you get 3dmark native numbers from? The linked scores from notebookcheck are results under emulation.

But yeah, the Surface Pro X under some games handily beat a Surface Pro 7 - despite emulation.
 

Etain05

Junior Member
Oct 6, 2018
11
22
81
Oh, look at that, Apple will transition its entire Mac lineup to ARM, and the entire transition will take 2 years. Who could have possibly predicted such a shocking thing? Guess who predicted that down to the number of years ;P
Where is @Glo. when you need them to tell us how impossible that is and to denounce anyone who would suggest otherwise?
 

SarahKerrigan

Senior member
Oct 12, 2014
339
468
136
Complete picture is that they will have to compete also with ARM's Cortex X1 core introduced receintly (which has 140% PPC of Zen2 already, 30% IPC jump over A77 is huge) and Cortex X2 next year (based on Matterhorn ARMv9 2048-bit SVE2 vectors). Probably some new versions of laptop Snapdragon 8cx will come based on X-series to MS Surface and other laptops. AMD and Intel has problems with relatively old A76 in Graviton2 server CPU. It will be interesting to see how Intel and AMD will be competing to new Cortex X1 cores and Apple's A14 cores. That battle will be epic. Such a clash was not in IT for long time.

Why do you keep saying Matterhorn is going to be 2048-bit vectors? I have seen absolutely zero indication of this, and it would be an inefficient use of area for a consumer core. I'd be surprised if it goes past 256b units, and shocked if it goes beyond 512.
 

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
The performance will be interesting to see

They wouldn't switch if they aren't certain it's in total mostly better. It's like with consoles. They are vertically integrated. They can put as much dedicated hardware on the SOC for all kinds of things up to AI/ML and other matrix/vector processing. Since they control the OS and APIs, that makes it much easier. However I think these macs can't really be used anymore for running linux on them nativley (windows obviously is out anyway due to ARM) or lose a ton of the benefits of the SOC.

13mm2 GPU in the Tablet A12Z performing at least like Intel's Iris Plus G7 graphics or Vega 7 in AMD's APU. 3x perf/area. It's not just the CPU ARM vendors excel in!
These GPUs also don't have to support a ton of APIs. That works well with Apple that is vertically integrated. They need to run metal API and that is basically it. No 10 different versions of DX and OPenGL and vulcan and so forth. Plus the mentioned Apple GPU has video encode/decode in a seprate block where as in case of AMD or Intel they are part of the GPU and in these small gpus, a very significant part. You would have to count the size of the CUs only.
 
  • Like
Reactions: Tlh97

Tup3x

Senior member
Dec 31, 2016
945
925
136
This was pretty obvious outcome considering all various rumours. It was pretty obvious that if they want to make the switch they could do it rather painlessly.

I hope that ARM would gain traction on Windows too. It would increase the competition greatly and break the current duopoly.
 

Richie Rich

Senior member
Jul 28, 2019
470
229
76
Why do you keep saying Matterhorn is going to be 2048-bit vectors? I have seen absolutely zero indication of this, and it would be an inefficient use of area for a consumer core. I'd be surprised if it goes past 256b units, and shocked if it goes beyond 512.
Because there is difference between ISA extension and HW units implementation. The ISA extension SVE2 can process vectors from 128-bit up to 2048-bit which means that even the Little core like post-A55 will be able to handle 2048-bit SVE2 vectors (probably using 1x128-bit FPU/SIMD unit). You are talking about HW units which we don't know the configuration yet. Fujitsu A64FX has 2x512-bit SVE units what is overkill for normal use out of supercomputing world. ARM Cortex X1 introduced first ARM core with 4xFPU units (IIRC Apple core has 3xFPU) so IMHO Matterhorn Cortex X2 next year will have 4x128-bit or 4x256-bit HW units.

Apple A15 can have 3x256-bit FPU/SIMD HW units, theoreticaly doubling the throughput. Anyway, it's clear that ARMv9 ISA with it's 2048-bit wide vectors is leap frog (from 128-bit NEON) a will be much more advanced than x86-64 being stuck at AVX512. x86-64 can stay powerfull thanks to stronger HW SIMD units but this can change.

Who introduced world's first 6xALU core? ................ Apple A11 Monsoon in 2017 is ARM ISA.
Who introduced world's first 2048-bit vectors? ........ Fujitsu A64FX in 2019 is ARM ISA.

Clearly ARM ISA has bigger development speed and momentum right now. I wouldn't underestimate this. Especially when you take into account that CPU development takes 4 years which means all of those 2048-bit SVE2 chips like Apple's A15 and ARM's Matterhorn is about to finish development and tape out soon this year (to be available as silicon at the end of next year in mass production). Apple is currently working on A16 and A17 architecture....
 

DrMrLordX

Lifer
Apr 27, 2000
21,583
10,785
136
They must be cooking up a beast of a SoC to compete with Xeon and Epyc processors.

Or they're effectively abandoning that corner of the market. The Mac Pro of tomorrow may not be the same as the Mac Pro of today (or, really, yesterday). AMD is soldiering on into the future with 64c/128t workstation CPUs, while Apple may not go that route at all.

Because there is difference between ISA extension and HW units implementation.

Every time you bring this up, people jump down your throat for insisting that such-and-such future SVE2 implementation is 2048b SVE2. Just because SVE2 allows up to 2048b-wide vectors doesn't mean diddly squat. The only thing that really matters for any given CPU is the actual hardware implementation of SVE2. How wide are the pipes? That's all we really care about.

You can stop telling us about 2048b SVE2 because nobody's gonna have pipes that wide.
 

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
For the duration of the transition Intel Macs will be between rocks and a hard place, with new "Apple Silicon" Macs pushing from the one and AMD x86 systems on the other side. Will be interesting to see how big the market for Boot Camp really is now that that feature will disappear.
 

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
Will be interesting to see how big the market for Boot Camp really is now that that feature will disappear.

On the other hand many devs use mac to develop software. Software that will then run on x86-based linux servers or even windows. I'm not a fan of macs and never got their appeal (overpriced hardware really) but if one of your main uses is to develop software running on x86 system, it's probably time to switch away from mac if you can't really test the stuff you develop on your local machine. It's one thing to develops on Windows for linux or vice-versa (or cross-platform) which often already poses issues but developing cross-ISA??? Can't see how that would work. I think they will actually lose market share simply due to that, at least in US.
 

Richie Rich

Senior member
Jul 28, 2019
470
229
76
I think they will actually lose market share simply due to that, at least in US.
Or Apple can gain big market share in ARM developing workstations. Because really there is almost non right now. There is eMag but this has half IPC of old A76 used in Graviton2 so it's useless.
https://www.anandtech.com/show/15733/ampere-emag-system-a-32core-arm64-workstation

With big ARM push in cloud and servers it looks like Apple will have the only powerful ARM development rig for those people. Sure they will loose x86 development people but they can more than compensate that IMHO.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
Why not move all of their x86 "stuff" to the cloud? Instead of BootCamp or other solutions that run Windows on the local mac hardware, why not just further integrate running virtual Windows instances in the cloud and have that be their solution? It allows them to reduce the price, or at least the cost of goods sold, for their desktop products while opening up a possible revenue stream through cloud virtualization services. MAcs, being the premium products that they are, are more likely to be used by people that would be paying for cellular data services or at least be on high speed wifi almost all the time. Why not just make the assumption that they will always have access to the cloud? Apple certainly won't care about those few road warriors that want to do their dev work out in the sticks without having any data connection at all, and with Starlink taking flight, even those devs would have access to a portable Starlink base station to give them a wifi hotspot.

Times are changing, we don't HAVE to have desktop supercomputers any more. Those that need it, likely only want to pay for what they actually use. Also, those that need it, likely need the best that they can get. Why spend $5-10K on a computer that's going to be obsolete in three years when you can spend a few hundred a month on having the fastest thing running? And, if you're dealing with such massive amounts of data that it can't be moved on and off the cloud easily, you're likely doing video production work that's going to be able to be more effectively handled with a dGPU attached via Thunderbolt, which can be easily replaced and upgraded as needed?
 
  • Like
Reactions: Richie Rich

SarahKerrigan

Senior member
Oct 12, 2014
339
468
136
Because there is difference between ISA extension and HW units implementation. The ISA extension SVE2 can process vectors from 128-bit up to 2048-bit which means that even the Little core like post-A55 will be able to handle 2048-bit SVE2 vectors (probably using 1x128-bit FPU/SIMD unit). You are talking about HW units which we don't know the configuration yet. Fujitsu A64FX has 2x512-bit SVE units what is overkill for normal use out of supercomputing world. ARM Cortex X1 introduced first ARM core with 4xFPU units (IIRC Apple core has 3xFPU) so IMHO Matterhorn Cortex X2 next year will have 4x128-bit or 4x256-bit HW units.

Apple A15 can have 3x256-bit FPU/SIMD HW units, theoreticaly doubling the throughput. Anyway, it's clear that ARMv9 ISA with it's 2048-bit wide vectors is leap frog (from 128-bit NEON) a will be much more advanced than x86-64 being stuck at AVX512. x86-64 can stay powerfull thanks to stronger HW SIMD units but this can change.

Who introduced world's first 6xALU core? ................ Apple A11 Monsoon in 2017 is ARM ISA.
Who introduced world's first 2048-bit vectors? ........ Fujitsu A64FX in 2019 is ARM ISA.

Clearly ARM ISA has bigger development speed and momentum right now. I wouldn't underestimate this. Especially when you take into account that CPU development takes 4 years which means all of those 2048-bit SVE2 chips like Apple's A15 and ARM's Matterhorn is about to finish development and tape out soon this year (to be available as silicon at the end of next year in mass production). Apple is currently working on A16 and A17 architecture....

This is nonsense. Literally no structure in A64FX's architectural SIMD is 2048 bits wide. It's not like SX where the registers are 16384b regardless of the functional unit width - SVE just provides architectural features to intelligently act based on the vector width present in a given microarchitecture. An SVE core with 512b vectors is 512b, period - 512b vector registers, 512b execution width, 512b instructions. There is no such thing as "able to process 2048-bit SVE vectors" at all, and the fact you think there is tells me you haven't read the SVE manual; the whole point of SVE is that you write code that works with knowledge of available vector length, not that you just run a single 2048b op across 16 cycles or whatever it is you're trying to imply.

I love SVE as much as the next girl, but come on. "Matterhorn has 2048-bit SIMD because someone could theoretically implement a different SVE-capable core with 2048-bit SIMD" is possibly the weirdest argument I've heard this week.
 
  • Like
Reactions: blckgrffn

Richie Rich

Senior member
Jul 28, 2019
470
229
76
There is no such thing as "able to process 2048-bit SVE vectors" at all, and the fact you think there is tells me you haven't read the SVE manual; the whole point of SVE is that you write code that works with knowledge of available vector length, not that you just run a single 2048b op across 16 cycles or whatever it is you're trying to imply.
Ok. So code written for ARM Fujitsu A64FX with 512-bit SVE using 512-bit registers will not run at Matterhorn with 256-bit SVE2? Another one, 2048-bit SVE2 can be scaled with 128-bit increment so there are 16 versions of SVE2 length (128 * 16 = 2048)...... this looks to me very messy. This would also mean that you cannot pair Little cores with bigger cores with different SIMD widths.

I think this solves Vector Length Agnostic instructions (VLA):
  • Vectors cannot be initialised from compile-time constant in memory, so...•INDEX Zd.S,#1,#4: Zd= [ 1, 5, 9, 13, 17, 21, 25, 29 ]
  • Predicates also cannot be initialised from memory, so...•PTRUE Pd.S,MUL3: Pd= [ T, T, T , T, T, T , F, F ]
  • Vector loop increment and trip count are unknown at compile-time, so...•INCD Xi: increment scalar Xiby # of 64-bit dwordsin vector •WHILELT Pd.D,Xi,Xe: next iteration predicate Pd= [ while i++ < e ]
  • Vector register spill & fill must adjust to vector length, so...•ADDVL SP,SP,#-4: decrement stack pointer by (4*VL)•STR Z1,[SP,#3,MUL VL]: store vector Z1to address (SP+ 3*VL)
https://indico.math.cnrs.fr/event/4705/attachments/2362/2940/ARM_SVE_tutorial.pdf
page 8.


As far as I understand VLA it will able to chop whatever long vector into lengths suitable for local SIMD unit to able to process it regardless its HW width. So little cores with slow 128-bit SIMD FPU can be paired with 4x256-bit SIMD FPUs thanks to VLA. Or do you see VLA function in different way?
 

pete6032

Diamond Member
Dec 3, 2010
7,443
3,004
136
The 13" MacBook Pro receiving an ARM chip will be interesting. Apple must have done some demographic evaluation and figured out that most of the 13" MBP users aren't actual Pro users -- or Apple have rapidly expedited full versions of the pertinent Pro software for these ARM laptops.

I still have my reservations about heat / performance of ARM in these laptops doing long runs of multicore processing, considering my iPhone 11 gets hot just running Zoom calls over cellular.
It is my opinion that base MacBook pro 13 inch specs are very pedestrian (i5, 8GB, 128GB). The "pro" moniker for a system like that is kind of laughable. I would imagine that most base model MacBook pro buyers are not using that type of system for anything other than text editing, email, internet, maybe some light photoshop. It is much more reasonable to call the higher end ones "pros" when they have i7, i9 and 16GB RAM and 512/1TB SSD.