Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Different performance levels of P and E cores between Apple and Intel may be related to the fact that Apple design is prioritising mobile applications (iPhone specifically) and Intel design might be prioritising performance applications (desktops and, especially, servers).

Yes, that’s probably the historical background. Intel never had much incentive to pursue ultra-low power, while for Apple this was the central design goal from the start. And the practical consequence of this is that Intel is stuck in a thermal corner (which is also the main reason why it has to go horizontal) and Apple has much more scalable tech. Their P-cores offer performance close to that of Intel P-cores at much lower performance, and their E-cores consume almost no power at all.

And I think it’s entirely incorrect to say that Apples E-cores have no place on desktop. Quite on contrary. They can take care of low-priority background tasks, minimizing context switches and cache trashing on the main cores and freeing them up to do more performance-sensitive work.

It is entirely possible that Apple will eventually develop the E-cores more into throughput support processors (there were major performance improvements post A14 in this domain), but the focus in energy efficiency and background work execution is unlikely to change. This is also evident from Apples balanced core configurations which are optimized for real world usage and concerns rather than more artificial metrics.

Apple E cores would be absolutely useless on the servers.

No, they wouldn’t be useless (for the reasons I explain above). But Apple doesn’t need a specially designed core to maximize server throughput. Their P-cores are both faster and use less power than Intels E-cores.

And if Intels’s P-core scaled well in respect to power usage they wouldn’t need the E-core for throughput either. AMD doesn’t.
 
She announces a processor that rivals the M1 Pro (which is 15 months old at this point), and you say she should replace the guy who brought us that original chip?
Reports from The Information have shown that people in Apple’s chip division are leaving and the morale is low. Basically, all the brightest minds behind the M1 have left Apple. If we look at AMD, which has arguably the best leader in the chip business, that company is run like a well oiled machine and is plowing ahead generation after generation. It all comes from the top.

The A15 and A16 have shown that Apple is in a big slump. The blame falls on Johny Srouji. As a leader you need to rally the troops. Let’s see if he can strike lightning in a bottle twice.
 
  • Angry
Reactions: AlexMac89
The A15 and A16 have shown that Apple is in a big slump. The blame falls on Johny Srouji. As a leader you need to rally the troops. Let’s see if he can strike lightning in a bottle twice.

What a load of nonsense. A15 had massively redesigned E–cores, improved coprocessors and new GPU features. Admittedly, A16 was a letdown but this can be explained by delays in TSMC‘s roadmap. There is evidence that the big thing in the A16 was supposed to be a new GPU but these plans had to be scrapped as it really required 3nm node.

All in all, there are bigger architectural changes in one year from A14 to A15 than AMD had in two years from Zen3 to Zen4. And you make claims about better execution.
 
Apple E cores would be absolutely useless on the servers.
I would disagree there. Servers are not just meant for high performance computing. One of their main use is to push data around between I/O channels, e.g. from disk to network. The E-cores are ideal as they are fast enough to deal with the interrupts and DMA circuitry takes care of the rest.
 
Nobody would grant this many excuses for AMD or Intel if they have released two underwhelming chips back to back.

Nobody would claim that Intel or AMD were doomed if they released chips with 10% performance improvements each year either. With Intel, it took Intel over four years of single–digit yearly improvements for the customers to become worried. But of course, Apple not releasing a major P–core redesign after two years is seen as a sign of disaster.

P.S. Fun fact - AMD claims a 18% improvement in performance on their top laptop chip from the previous generation 6900HS (Zen3 to Zen4). That’s comparable to A16 to A14. But of course, AMDs execution is so much better. Note that 20% faster single–core than 6900HS is still slower than the passively cooled M2.
 
Last edited:
All in all, there are bigger architectural changes in one year from A14 to A15 than AMD had in two years from Zen3 to Zen4.
Let's make the comparison. What parameters did you use?

ab0d302d-57d3-462a-a738-0599b87a6985_4096x4096.jpg
553cf263-224b-475d-8949-f97d2d13a187_2000x1604.jpg



02820a31-70d1-4dd2-8490-faab93235e26_1736x1307.jpg

 
Last edited:
Let's make the comparison. What parameters did you use?

Right. So Zen3 to Zen4 tweaked the size of the internal structures and caches, introduced AVX512 (big one) and went from 7nm to 5nm.

Apple also tweaked the size of internal structures and caches in both A15 and A16. They didn’t introduce as major new ISA features (only small stuff like native bfloat16 support and possibly other small features). A16 is manufactured on 4nm, but folks say that’s it’s a refinement if 5nm rather than a completely new process. However, the E–core was made much bigger in A15 (more execution units etc), and there were also major changes to the matrix coprocessor (double the throughtput) and the NPU. From the link you posted the cores on A16 are a new design as well. Of course, you won’t find marketing slides about all these things. Just a few enthusiasts measuring these things in their spare time and publishing results either on Twitter on in their microblogs.

The funny thing is that what constitutes a major new architecture for x86 is traditionally a fairly minor revision for Apple. But folks don’t perceive it this way because a) Apples marketing does not release technical details and b) x86 architectural updates are rarer while Apple does these kinds of tweaks every year. And they still do although it is also true that recently their execution has slowed down.
 
To think that Dr. Su would take another opportunity that’s anything less than CEO of another company after successfully turning around AMD as CEO is just silly. Leading a division of another company is not the trajectory someone like that takes. It goes beyond simply offering a lot of money.
Two words: Angela Ahrendts. Sure, by my quick math, Burberry is about 1/4th the revenue of AMD (and I don't really feel like looking up market cap and other metrics), but she left a CEO job to lead a division at Apple. Made a bundle of money too...
 
Also, an observation: at a high level, AMD's resurgence is at least partly based on the same factors driving Apple silicon. They realized, either when they decided to spin off GlobalFoundries or when, presumably with AMD's input, GlobalFoundries decided to stop investing in new processes, that smartphones and their ecosystem, not the PC (broadly defined, including x64 servers), would be what would drive the semiconductor industry going forward in the second half of the 2010s and beyond.

And so you can let the smartphone industry take the lead on the investment required for new fabs, miniaturization R&D, etc and then benefit from that technology at a dramatically lower cost than trying to do it yourself. And that means that both Apple and AMD, which both have much lower volume on PC chips than Intel, are now able to be much, much more competitive. And Intel, which can't leverage the smartphone industry to drive its investment (selling the ARM business in 2006 will go down as a major blunder, as was not being interested in being a fab for Apple designs), is still operating at PC-market scale...

I'm not sure if that was Lisa Su's insight or her predecessors' - she only joined AMD in 2012, so certainly the GlobalFoundries spin-off preceded her - but the continuation of that strategy and the shift towards TSMC may have happened under her watch.
 
Seem like you have outdated information. The A16 and Snapdragon Gen 2 is way closer than people think.


When Apple was ahead, people ran with the narrative that the performance of Apple Silicon left everyone else in the dust and Apple bragged about it during presentations. Now that the others have caught up, we can’t go and say performance doesn’t matter to Apple. That’s hypocritical.

Lol. Caught up is:
Snapdragon 8 Gen 2 scored 1490 in the single-coretest and 5131 in the multi-core test. Whereas in our testing, Apple’s A16 Bionic chipset scored 1879 in the single-core test and 5307 in the multi-core test.

Doesn’t seem caught up to me! Oh and which devices are shipping with this specific spec on high and efficient performance cores? Pray tell that the GHz spec matches with production numbers?
 
Nobody would grant this many excuses to AMD and Intel if they have released two underwhelming chips back to back.

Seems that AMD is making their own excuses for everyone else.

Why is Zen4 intermixed with Zen3 within the same processor series?! Hmmm.

Also the TDP power measurement of their UltraThin laptop CPUs still is about 10-15W higher than M1/M1 Pro or M2 chips from Apple and this is AMDs latest currently released chips not Apple’s 6mths or older chips.

You still want to claim they’ve caught up?

Regarding “top chip designers” from Apple only 1 was mentioned as being quite instrumental at Apple. So please drop some full names & your references.
 

Attachments

  • 2389961D-2071-4FAD-A572-1C5F133322D5.jpeg
    2389961D-2071-4FAD-A572-1C5F133322D5.jpeg
    130.2 KB · Views: 85
  • C81103B4-9FB0-4911-A22F-BA35F6E4E2D1.jpeg
    C81103B4-9FB0-4911-A22F-BA35F6E4E2D1.jpeg
    110.5 KB · Views: 80
I didn't read the thread but joining Apple's chip team would be a huge downgrade for Lisa Su in terms of a career. The only upgrade for Lisa Su at Apple would be CEO or next in line for CEO.
 
  • Like
Reactions: Retskrad
Regarding “top chip designers” from Apple only 1 was mentioned as being quite instrumental at Apple. So please drop some full names & your references.
Manu Gulati (now Nuvia), John Bruno (now Nuvia), Gerard Williams III (now Nuvia), Mike Filippo (now Microsoft)

Why is Zen4 intermixed with Zen3 within the same processor series?! Hmmm.
The nomenclature of AMD mobile CPUs is extremely confusing to make older and cheaper IP cores more marketable. Only the 7040 series has new AMD IP cores.
 
Regarding “top chip designers” from Apple only 1 was mentioned as being quite instrumental at Apple. So please drop some full names & your references.
There has been bit of a brain drain in Apple's chip design department.
 
Reports from The Information have shown that people in Apple’s chip division are leaving and the morale is low. Basically, all the brightest minds behind the M1 have left Apple. If we look at AMD, which has arguably the best leader in the chip business, that company is run like a well oiled machine and is plowing ahead generation after generation. It all comes from the top.

The A15 and A16 have shown that Apple is in a big slump. The blame falls on Johny Srouji. As a leader you need to rally the troops. Let’s see if he can strike lightning in a bottle twice.

Well, those engineers are not leaving Apple for AMD, but for Qualcomm, Intel and Microsoft so your whole analysis is strange and faulty. By your logic Apple should hire chef engineers from those companies, not from AMD.
 
  • Like
Reactions: Jeven Stobs
Manu Gulati (now Nuvia), John Bruno (now Nuvia), Gerard Williams III (now Nuvia), Mike Filippo (now Microsoft)


The nomenclature of AMD mobile CPUs is extremely confusing to make older and cheaper IP cores more marketable. Only the 7040 series has new AMD IP cores.
I already posted the 7040 series and still it’s a mix-match of old and new. Don’t believe me have a look.

Oh and for Manu Gulati - he’s NOT at Nuvia and his experience with Apple is with older devices and chipsets. Nothing with Apple Silicon for laptops or desktops. The last chip was with A12x and it shows with Qualcomm.

John Bruno P also not at Nuvia but Qualcomm. Notice his entry regarding Apple is a one liner for 5yrs!???? Very odd and it’s his OWN LinkedIn account his control.


Gerard Williams III,
Now THIS guy has some serious chips! An Arm Fellow! Been around since Texas Instruments yes this guy leaving to create Nuvia (which may have been acquired) knows his stuff.

Mike Filippo possibly he’s featured i Bloomberg as key architect but what exactly did he work on at Apple for his 2yr stint. https://www.bloomberg.com/news/arti...ts-key-apple-engineer-to-work-on-custom-chips

???!
 

Attachments

  • 3D57BEE0-22D0-4EDD-9A52-3AF9A0860669.png
    3D57BEE0-22D0-4EDD-9A52-3AF9A0860669.png
    607.9 KB · Views: 61
  • DB41D9C6-E208-4AAA-BF26-1E02D7EA8226.png
    DB41D9C6-E208-4AAA-BF26-1E02D7EA8226.png
    368.9 KB · Views: 64
  • A86C5DAF-26D6-4609-AB57-9C3ECFB87696.png
    A86C5DAF-26D6-4609-AB57-9C3ECFB87696.png
    536.1 KB · Views: 71
  • D9C4C6A8-260B-4EFF-A194-D531F57853D3.png
    D9C4C6A8-260B-4EFF-A194-D531F57853D3.png
    554.3 KB · Views: 58
  • 72AF9AC9-E366-423B-8A13-E8318FBEFE03.png
    72AF9AC9-E366-423B-8A13-E8318FBEFE03.png
    582.7 KB · Views: 65
  • 92DD3D45-5495-4849-B7B3-F37137905189.png
    92DD3D45-5495-4849-B7B3-F37137905189.png
    353.7 KB · Views: 51
  • 8085AABF-4105-404E-AB1D-240161C31D1E.png
    8085AABF-4105-404E-AB1D-240161C31D1E.png
    359 KB · Views: 64
  • 861CB5D5-843B-4085-B016-F83695762BD9.png
    861CB5D5-843B-4085-B016-F83695762BD9.png
    659.2 KB · Views: 59
  • Like
Reactions: Xiao_Xi
Oh and for Manu Gulati - he’s NOT at Nuvia and his experience with Apple is with older devices and chipsets. Nothing with Apple Silicon for laptops or desktops. The last chip was with A12x and it shows with Qualcomm.

John Bruno P also not at Nuvia but Qualcomm. Notice his entry regarding Apple is a one liner for 5yrs!???? Very odd and it’s his OWN LinkedIn account his control.

https://www.linkedin.com/in/john-bruno-p-eng-88616a1
Gerard Williams III,
Now THIS guy has some serious chips! An Arm Fellow! Been around since Texas Instruments yes this guy leaving to create Nuvia (which may have been acquired) knows his stuff.

Mike Filippo possibly he’s featured i Bloomberg as key architect but what exactly did he work on at Apple for his 2yr stint. https://www.bloomberg.com/news/arti...ts-key-apple-engineer-to-work-on-custom-chips
Nuvia was founded by Manu Gulati, John Bruno and Gerard Williams III and bought by Qualcomm. Although Nuvia no longer exists, they continue to work together, making them a "subdivision" of Qualcomm.

Although Mike Filippo's contribution may not be as relevant as others, it may be a sign that Apple has a problem. Apple hired him because he had a very important role at ARM and, after two years, he moved to Microsoft.

Those people are important enough to make headlines. It seems more important to know if the employee turnover rate in that division is higher than in any other.
 
Nuvia was founded by Manu Gulati, John Bruno and Gerard Williams III and bought by Qualcomm. Although Nuvia no longer exists, they continue to work together, making them a "subdivision" of Qualcomm.

Although Mike Filippo's contribution may not be as relevant as others, it may be a sign that Apple has a problem. Apple hired him because he had a very important role at ARM and, after two years, he moved to Microsoft.

Those people are important enough to make headlines. It seems more important to know if the employee turnover rate in that division is higher than in any other.

By this same logic AMD is doomed since it lost Jim Keller and Raja Koduri. Good engineers like new challenges and switching companies is good for their career. The fact that some good engineers at Apple decided to roll their own startup is neither unusual nor alarming.

What is relevant is execution. If Apple fails to produce relevant advances in 2023, yes, that could be a sign of trouble. Right now it’s too early to tell.
 
  • Like
Reactions: Homy
If Apple fails to produce relevant advances in 2023,
People are already complaining the sky is falling in regards to the M2, and I just don't see that. I don't think any chip designer is going to offer huge performance increases every chip generation. I think its perfectly acceptable see modest improvements in the M2, and then expect a bigger jump with the M3. This sort of mimics Intel's tick-tock strategy, but obviously they failed in the execution of that.
 
  • Like
Reactions: DeepIn2U and leman
Nuvia was founded by Manu Gulati, John Bruno and Gerard Williams III and bought by Qualcomm. Although Nuvia no longer exists, they continue to work together, making them a "subdivision" of Qualcomm.

Although Mike Filippo's contribution may not be as relevant as others, it may be a sign that Apple has a problem. Apple hired him because he had a very important role at ARM and, after two years, he moved to Microsoft.

Those people are important enough to make headlines. It seems more important to know if the employee turnover rate in that division is higher than in any other.
These people ARE important to make headlines - simply because of keywords: Apple + Silicon + Engineers

The KEY part to focus on, again besides Gerard Williams III (whom I suspect felt restricted in vision based on his stellar expertise and experience under Srouji), is WHAT they worked on and WHEN at Apple:

A12x/z is worlds behind Apple silicon M series!
This could be why they left - believing to use learning of A series chips back with A12 or prior was not suitable for desktop macOS processing. maybe restrictions of Displays or using embedded graphics with no eGPU or external card support - We'll NEVER know! But it's a key part to focus on which 'Headlines' does NOT reflect nor even mention which is HUGE factor.

Looking at claims that Qualcomm has caught up is very premature and shows the limitations in expertise or reverse engineering.

think of it like this:
A) this person studied for the exam - passed.
B) this person copied off person A or paid for the exam cheat sheet and is good ONLY at passing tests.
C) is person A but not up to date on knowledge and new applications of working with said new knowledge.

results when asking both person A & B to do the actual work and check their work validity; WHOM do you think will shine or be able to do the real work? I wonder if John Bruno P falls more into category B here or all fall into person C.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.