Apple loss its chief CPU architect is a great loss for Apple, see A16.

[exoticSpice]

Gerard Williams III designed the A series from A7 to A12 and laid the foundations for M1.

After he left the A15 and A16 that Apple released had barely any IPC increases?
I think Apple's CPU progress will slow down.

Because of him Apple was the first to ARMv8 AND released the first 64bit ARM cpu. Nowadays, ARMv9 A/M series is no where to be seen.

 

[chromite]

Gerard Williams III designed the A series from A7 to A12 and laid the foundations for M1.

After he left the A15 and A16 that Apple released had barely any IPC increases?
I think Apple's CPU progress will slow down.

Because of him Apple was the first to ARMv8 AND released the first 64bit ARM cpu. Nowadays, ARMv9 A/M series is no where to be seen.

I don’t know what any of this means

 

[exoticSpice]

I don’t know what any of this means

So the designer of A7 to M1 chips left Apple in 2019 and because of the will we see any slow downs in CPU advancement from Apple?

 

[T'hain Esh Kelch]

TIL Apple only has one chip designer.

 

[Xiao_Xi]

The iPhone got a big performance boost every time Apple used a better node. And the same will be true for Mx SoCs.

iPhone 7's Performance Is A Massive 16% Better Than iPhone 6s

In my last post, I was pretty much disappointed with the design of iPhone 7 since there is negligible difference between it and iPhone 6s according to our sources. But a new benchmark has surfaced on the interwebs which details the performance difference between the new and former flagship and I...

www.thecrydsdaily.com

 

[exoticSpice]

TIL Apple only has one chip designer.

I did say "Chief" designer.

 

[exoticSpice]

The iPhone got a big performance boost every time Apple used a better node. And the same will be true for Mx SoCs.

View attachment 2112313

iPhone 7's Performance Is A Massive 16% Better Than iPhone 6s

In my last post, I was pretty much disappointed with the design of iPhone 7 since there is negligible difference between it and iPhone 6s according to our sources. But a new benchmark has surfaced on the interwebs which details the performance difference between the new and former flagship and I...

www.thecrydsdaily.com

Wow, that chart basically confrims what I mean. Look at that jump from A6 to A7 and then a huge jump from A8 to A9. A9 to A10 was a huge jump as well.

Gerald first designed the A7 and boy is it ever clear.

The A14 got a new node family but the jump was not huge due very low IPC increases.

 

[Xiao_Xi]

Wow, that chart basically confrims what I mean. Look at that jump from A6 to A7 and then a huge jump from A8 to A9. A9 to A10 was a huge jump as well.

We need more data to draw solid conclusions. We should compare how the improvements of the new Samsung/TSMC nodes are reflected in performance and power consumption across all iPhones.

 

[leman]

Contrary to the popular opinion, Apple doesn't actually release a fully new CPU every year. Different subsystems are redesigned on different cadences, each of them usually taking multiple years.

For example, A14 has introduced a new big core architecture. A15 and A16 were minor tweaks bringing in some new features but mostly tweaking the cache subsystem and focusing on power efficiency (and A16 is particularly impressive here). At the same time, A15 had an upgraded GPU and other components, and both A15 and A16 had major changes to the efficiency cores (the E-cores of A16 are a significant upgrade over E-cores of A14).

So far there is no evidence to support the conclusion that Apple's chip design efforts are slowing down. Maybe changing focus a bit, e.g. they go for power efficiency instead of absolute performance on their A-series, but that makes perfect business sense. The big test however will be the next iteration of Apple prosumer chips. If they will still use minor tweaks to A14 architecture, Apple might have a problem.

 

[exoticSpice]

So far there is no evidence to support the conclusion that Apple's chip design efforts are slowing down. Maybe changing focus a bit, e.g. they go for power efficiency instead of absolute performance on their A-series, but that makes perfect business sense. The big test however will be the next iteration of Apple prosumer chips. If they will still use minor tweaks to A14 architecture, Apple might have a problem.

oh yes its slowing/slowed down. Look at the A7, A9, A10, A11, A12X. These chips are vastly different and brought about huge increases in pref and IPC. These 5 chips have different core layouts and are NEW core designs which Apple brought every year.

The last chip that Gerald designed was the A12/A12X. The A13, A14, A15 and A16 got pref gains due to node changes and frequency increases and but VERY minor IPC increases.

The Bionic design/core layout that Gerald designed since A11 is still being used in the A16. 2p+4e.

Losing this man as the chief designer was a misake that Apple made. If the next A17 chip does not bring a huge IPC increase or a new core design then Apple is slowed dow for sure.

hey go for power efficiency instead of absolute performance on their A-series

But that's the thing the A series cores are used in their laptops and desktops. They have to find a balance.

 

[leman]

oh yes its slowing/slowed down. Look at the A7, A9, A10, A11, A12X. These chips are vastly different and brought about huge increases in pref and IPC. These 5 chips have different core layouts and are NEW core designs which Apple brought every year.

The last chip that Gerald designed was the A12/A12X. The A13, A14, A15 and A16 got pref gains due to node changes and frequency increases and but VERY minor IPC increases.

On the IPC side I don't see the improvements slowing down at all. Sure, A16 and A15 have exact same P-core IPC as the A14 (because there were not many changes), but E-core has improved massively. Similarly, A14 was a big jump in IPC over A12. In the past Apple brought a redesigned P-core u-arch every third year or so. A16 is obviously a tad dissapoiniign in this regard, but it also looks like a conservative placeholder chip designed to bridge the gap caused by COVID slowdowns and 3nm delays. Again, I wouldn't interpret too much into it until this trend continues next year as well — then it will be time to get worried.

The Bionic design/core layout that Gerald designed since A11 is still being used in the A16. 2p+4e.

I don't understand what's that supposed to show? 2P+4E sounds like the most useful configuration for mobile phones if you have good P-core performance. What are they supposed to do instead? Android needs to come up with other configurations because their core performance sucks compared to Apple.

Losing this man as the chief designer was a misake that Apple made. If the next A17 chip does not bring a huge IPC increase or a new core design then Apple is slowed dow for sure.

Agree with the second sentence, not so much with the first. If the gut wanted to leave and pursue other projects, well, that's his right, no? A company is not a slave owner. If his interests and the company's interests have diverged, why was it a "mistake"?

But that's the thing the A series cores are used in their laptops and desktops. They have to find a balance.

Until now it worked out fine. The next prosumer laptops will show where they are going. Wouldn't be surprised if they're come up with a new high-perf Mac-only core.

 

[Gudi]

I think Apple's CPU progress will slow down.

I don't care. It's the Dutch ASML machines and the Taiwanese TSMC factories, which will enable the next process node shrink.

 

[fatTribble]

I’m possibly missing the point, but this thread sounds like another “I have evidence that Apple is doomed” threads. Apple lost SJ and survived. I’m sure they’re able to cope with other departures.

 

[T'hain Esh Kelch]

I did say "Chief" designer.

Yeah, and you also wrote that he designed all the chips.

 

[Love-hate 🍏 relationship]

Gerard Williams III designed the A series from A7 to A12 and laid the foundations for M1.

After he left the A15 and A16 that Apple released had barely any IPC increases?
I think Apple's CPU progress will slow down.

Because of him Apple was the first to ARMv8 AND released the first 64bit ARM cpu. Nowadays, ARMv9 A/M series is no where to be seen.

The gains in perf and efficiency are indeed kinds disappointing compared to previous gens

It is as I feared

 

[mcnallym]

Is there any CPU family that as matured the gains didn‘t slow down.

 

[teh_hunterer]

oh yes its slowing/slowed down. Look at the A7, A9, A10, A11, A12X. These chips are vastly different and brought about huge increases in pref and IPC. These 5 chips have different core layouts and are NEW core designs which Apple brought every year.

The last chip that Gerald designed was the A12/A12X. The A13, A14, A15 and A16 got pref gains due to node changes and frequency increases and but VERY minor IPC increases.

The Bionic design/core layout that Gerald designed since A11 is still being used in the A16. 2p+4e.

Losing this man as the chief designer was a misake that Apple made. If the next A17 chip does not bring a huge IPC increase or a new core design then Apple is slowed dow for sure.

But that's the thing the A series cores are used in their laptops and desktops. They have to find a balance.

You don't know anything about chip development, so you're grasping for something that's easy to understand (x person left y company) in order to explain things.

 

[Sydde]

Is there any CPU family that as matured the gains didn‘t slow down.

POWER has been around for more than 3 decades, and their performance increase has been pretty steady. The POWER10 is server-huge but still a significant gain over POWER9. There probably will be a point where μarches cannot really be further tweaked for more than trivial gains, but that point is still a ways off.

 

[exoticSpice]

You don't know anything about chip development, so you're grasping for something that's easy to understand (x person left y company) in order to explain things.

Not really. I do know enough that Apple had a brain drain in most of its departments

 

[Lounge vibes 05]

Gerard Williams III designed the A series from A7 to A12 and laid the foundations for M1.

So the designer of A7 to M1 chips left Apple in 2019

Gerald first designed the A7 and boy is it ever clear.

The last chip that Gerald designed was the A12/A12X.
The Bionic design/core layout that Gerald designed

The fact that you think one man has this much influence over the speed difference between processors is very odd.
He was far, far from working alone.
Theres the team he worked with, which was a part of the even bigger hardware team lead by Johny Srouji, which is only a small part of Apple.
The operations and marketing departments get their hands in all of this as well, deciding what’s a big enough upgrade to be able to market while keeping costs low enough for Apple’s massive margins.
On top of all this, there’s TSMC, Samsung, QUALCOMM, and all of the other companies who help with apples product chips which all have their own delays and issues and costs.
And then finally add on top of all that that for the last two years (almost 3 years) we’ve been in a pandemic/recession with supply shortages, delays and everything that goes with it.
There’s also just the fact that as products mature, the differences between models become smaller and smaller.
Sure, going from the iPhone 6 to the iPhone 6S might’ve felt more substantial than going from the 13Pro to the 14Pro, but that had only been seven years since the original iPhone, it’s been 15 years now.
Times change, Apple is not doomed, certainly their processors are not doomed, and this has been a pointless exercise.
PS: Apple isn’t working one processor at a time, if he left in 2019 then he definitely worked on the A14, M1, and more than likely he was there for the very early development of the A15 and M2 as well.