Report suggests Apple's A15 Bionic lacks significant CPU upgrades due to chip team brain d...

AppleInsider · September 15, 2021 9:42PM

Apple failed to detail the relative performance gains achieved by its new A15 Bionic SoC during Tuesday's iPhone 13 reveal, a rarity for a major hardware launch. One report believes the omission was not an oversight as Apple has little to tout.

While Apple mentioned certain aspects of its new system-on-chip design, specifically improvements related to graphics capabilities, executives were mum on raw CPU performance. A morsel was handed out during the presentation of iPad mini, with Apple saying A15 packs a 40% faster CPU compared to the tablet's A12-powered predecessor, but the comparisons ended there.

Rumblings on social media suggested that CPU numbers were intentionally left out because it is not as favorable a compare as previous A-series introductions.

In a report published shortly after the unveiling of iPhone 13 and iPad mini, SemiAnalysis suggests Apple's chip team is beset with internal troubles after a supposed brain drain sapped key executives and personnel from projects like A15. After conducting a preliminary die analysis of the A15 based on total transistors, estimated die size and other metrics, the publication found no major CPU improvements and concluded that Apple's next-generation core was delayed into 2022.

While there is little evidence that an emigration of talent is impacting Apple, the company has lost a number of key players in recent years. Apple's lead A-series chip designer from A7 through A12X, Gerard Williams III, left the company in 2019 to found Nuvia, a firm that was later snapped up by Qualcomm for $1.4 billion. Apple sued Williams on claims that he started Nuvia while working at the tech giant and recruited from its ranks, allegations the chip architect denies.

Today's report also mentions Rivos, a RISC-V startup that that includes a number of former senior Apple engineers.

"We believe Apple had to delay the next generation CPU core due to all the personnel turnover Apple has been experiencing," the report says. "Instead of a new CPU core, they are using a modified version of last year's core."

Alternatively, Apple could have sunk its transistor budget into the improvement of GPU cores and silicon designs that enhance graphics- and camera-related features. SemiAnalysis agrees that GPU gains are "quite impressive," but maintains the theory that a lack of CPU upgrades points to internal issues.

For its part, Apple in marketing material for iPhone 13 Pro lauds the advancements made with A15, saying that it is capable of making improvements to A-series silicon because it creates long-term product road maps "in ways no other company can." Many of the graphics-intensive tasks like ProMotion variable display refresh rates.

"That's how we deliver features like ProMotion, which have to be planned years in advance," Apple says. "Our chip team fully understood the needs of the display hardware, display software, and operating system teams and took them into account for A15 Bionic."

More detailed analysis of the A15 should come to light after the iPhone 13 and iPad mini are released next week.

Read on AppleInsider

longpath · September 15, 2021 9:46PM

It seems to me that Apple focused on a more power efficient cpu for this time around, as increased battery life was stressed.

tjwolf · September 15, 2021 9:55PM

longpath said:

It seems to me that Apple focused on a more power efficient cpu for this time around, as increased battery life was stressed.

Could be - or the increased battery life could simply be due to the bigger battery LOL.

cpsro · September 15, 2021 9:59PM

We should know in a year (2 years since the iPhone 12) if this speculation is true. Apple develops products over the course of years. What we see today didn't necessarily need all hands on deck during the past 2 years to bring to market.

mark fearing · September 15, 2021 10:04PM

It's impossible to know without a contact inside, and even then, that contact may be biased or ill informed. As mentioned, these things are in the pipeline for years. And with chip shortages across the entire supply chain, there may be manufacturing reasons to push off changes. I don't think Apple depends on 'one really smart person' or two or three. No company does at this point. Time will tell, and there must be some limits to what we can expect given certain sizes.

rinosaur · September 15, 2021 10:07PM

I noticed this comparison to the “competition” vs. the previous chip in the presentation and feared the worst…

edited September 2021

justsomeguy1 · September 15, 2021 10:22PM

(Posted this on another article, but it belongs here.)

[...] it seems likely that CPU performance is not significantly improved - lacking more facts, my money is on Andrei's analysis (in AnandTech), maybe 5-6%. But there are two wildly divergent ways to look at this.

It is possible that this is simply the result of a brain drain. That's a popular take in the press, right now. It's not clear how that analysis lines up with other known facts, like the massively improved NPU.

There is another possibility though. Apple is now designing a pair of cores for use not just in the phone, but also in the Mac. What are the needs of those two devices?

- For the phone, the biggest need is NOT more CPU performance. It's lower power use, which leads to greater sustained performance or longer battery life.

- For the Mac, it *is* more performance. But Macs are very different from phones, even the laptops. They can afford to burn more power on increased clock speed, unlike phones... IF the chip has the ability to run at higher clocks. It seems likely that the A14/M1 does NOT have that ability, simply based on the MBPs not clocking past 3.2GHz even when on wall current. (This is normal - every chip design has a maximum beyond which it can't go, no matter how much power you throw at it.)

The A-series chips have sped up from ~2.3GHz to ~3GHz over the last five years, since the iPhone 7, but most of the performance has come from widening the cores. But this leaves a ton of performance on the table- they should be able to get at least 4GHz, and possibly close to 5GHz, out of the process node they're using now, with a newer design. (Power requirements prevent that in the phone, of course.)

Now... what would such a redesign look like? Really, you'd want to try to preserve the IPC of your existing design while allowing for higher clocks. And you'd probably also want to increase your caches to compensate for the fact that every cache miss is going to cost more cycles (as each cycle is quicker). Once that design is done, if you don't need the max performance out of that chip in one situation, you'd run it slower and pocket the power savings.

This looks like it might be what Apple has done. They're claiming better battery life, despite a high-refresh-rate screen, a brighter screen, and a doubled system cache. And oh yeah, that doubled cache seems telling.

So, I think we can't really know what's going on at Apple until the new Macs ship. And maybe not until a new desktop (27" imac and/or Pro, not so much the mini) ship. If my guess is right, what we're seeing is Apple being very smart about maximizing the RoI on a single pair of core designs (high-perf & high-efficiency). They get better power efficiency in the A15, which is their primary design goal this time around, while being able to drive the cores much faster (4-4.5GHz, maybe?) in the M2 Macs. That would give the cores +25%-+40% performance PER CORE from clockspeed. You'd lose some performance due to longer pipelines, cache misses, etc, probably made up for by the larger cache (which might be where the +6% is coming from in the A14).

Next month will be *fascinating*.

verne arase · September 15, 2021 11:12PM

While brain drain is certainly possible, another possibility is that the CPU has gone as wide as possible, and further increasing the number of decoders would yield little benefit. As it is, the silicon design team probably raised a toast when they got eight instructions to run in parallel.

Once you get past a certain point, you're counting on higher clocks, eliminating bottlenecks, or better manufacturing techniques (like mask shrink) to achieve faster speed, and as your scale gets smaller you have quantum effects disrupting the works.

At that point, it would make sense to look for other areas to improve your workflow like optimizing ASICs (IP blocks) and other more specialized processors.

Apple doesn't design their computers to improve their SPECMARK numbers - they've always been about implementing new features in their products.

inspiredcode · September 15, 2021 11:14PM

What would be more useful to the end user? Upgrading the CPU, GPU, bionic cores? The CPU might not be the most important thing at this point. It already far surpasses everything else on the market.

chadbag · September 15, 2021 11:38PM

Dear editor: it seems that a piece of a sentence is
missing. This sentence fragment occurs at the end of a paragraph, but is not a sentence itself so makes no sense:

Many of the graphics-intensive tasks like ProMotion variable display refresh rates.

blastdoor · September 15, 2021 11:51PM

A brain drain is very much a possibility.

Lots of other good points too.

COVID could have also slowed design process.

Also, one nice feature of this new phone (at least in the US), is no price increase. Maybe apple decided better to limit feature improvements than raise prices, given supply constraints.

These are crazy times — hard to know what’s going on

swineone · September 16, 2021 12:56AM

Maybe the chip team was too busy complaining about returning to the office.

Or about Apple not condemning Israel against Palestine.

Or about the hazardous chemicals in Apple Park.

Or about pay inequality.

Or about #AppleToo.

swineone · September 16, 2021 1:01AM

RudolfGottfried said:

appleinsider said:
Apple failed to detail the relative performance gains achieved by its new A15 Bionic SoC during Tuesday's
iPhone 13 reveal, a rarity for a major hardware launch.

Not true. Apple made exactly the same kind of performance claims as last year. No comparisons with the previous generation were made there either. So why does the media continue to serve this clearly false narrative? On the one hand, of course, to generate clicks, and on the other hand, to promote moronic elaborations like this one from Daylan Patel. He already spread stupid stuff about the A14 last year, and this year you could also set the clock until he would once again spout absurd nonsense about Apple's chip development on his ridiculous rumor platform. This time the hook is: Apple is suffering terribly from the loss of engineers and therefore had to postpone its core development. Yes, of course. In 2019 about half a dozen Apple engineers (and not more than 100 as good Daylan claims) went and founded Nuvia, a startup that until today has produced nothing more than hot air and bizarre graphics sheets, but created so much wind that they were bought by Qualcomm. And since then, nothing more has been heard from them. Two more engineers went to Rivos this year and that's it. Daylan, who likes to speculate about how unbelievably startups will change the chip scene and who also claims that Google's self-developed chip (in reality a modified Exynos) will replace 10 million Intel chips, simply can't stand the fact that Apple's chip development generates such enormous performance increases year after year. So, he tries to fight it year after year in the short time frame between keynote and launch to get at least a bit of media attention. Okay, at the end of next week the spook is over. Until next year.

DED, is that you?

mikethemartian · September 16, 2021 1:43AM

Yeah, I’m sure no other design team can plan like Apple’s.

magman1979 · September 16, 2021 2:26AM

While there is little evidence that an emigration of talent is impacting Apple

That, right there, is all you need to know... This is all speculation and clickbait hyperbole from a bunch if anal-ysts with a quota of factless BS to pump out.

Has anyone seen benchmarks on the A15 yet? NO
Has anyone like iFixit done an SoC x-ray to examine changes to the die? NO
Did Apple not state in the presentation the A15 has "two all-new high-performance cores"? YES

Full stop

magman1979 · September 16, 2021 2:31AM

swineone said:

Maybe the chip team was too busy complaining about returning to the office.

Or about Apple not condemning Israel against Palestine.

Or about the hazardous chemicals in Apple Park.

Or about pay inequality.

Or about #AppleToo.

Or maybe NONE of the FUD you just wrote is true and you just need to put down the tin foil hat? *facepalm*

entropys · September 16, 2021 4:15AM

Well this article didn’t age too well. Up to 55% jump in GPU, over 15% CPU is pretty good IMHO.

it’s almost like Apple was hoping praying for such an article to generate angst ridden press, then repeat another bunch of press once the benchmarks start to dribble out.

in Australian parlance it is called feeding the chooks (the chickens being the media).

bestkeptsecret · September 16, 2021 5:42AM

rinosaur said:

I noticed this comparison to the “competition” vs. the previous chip in the presentation and feared the worst…

In the real world, there really isn't anything to fear. The iPhone 12 is a beast even now and the iPhone 13 is better.

bestkeptsecret · September 16, 2021 5:44AM

swineone said:

RudolfGottfried said:

appleinsider said:
Apple failed to detail the relative performance gains achieved by its new A15 Bionic SoC during Tuesday's
iPhone 13 reveal, a rarity for a major hardware launch.

Not true. Apple made exactly the same kind of performance claims as last year. No comparisons with the previous generation were made there either. So why does the media continue to serve this clearly false narrative? On the one hand, of course, to generate clicks, and on the other hand, to promote moronic elaborations like this one from Daylan Patel. He already spread stupid stuff about the A14 last year, and this year you could also set the clock until he would once again spout absurd nonsense about Apple's chip development on his ridiculous rumor platform. This time the hook is: Apple is suffering terribly from the loss of engineers and therefore had to postpone its core development. Yes, of course. In 2019 about half a dozen Apple engineers (and not more than 100 as good Daylan claims) went and founded Nuvia, a startup that until today has produced nothing more than hot air and bizarre graphics sheets, but created so much wind that they were bought by Qualcomm. And since then, nothing more has been heard from them. Two more engineers went to Rivos this year and that's it. Daylan, who likes to speculate about how unbelievably startups will change the chip scene and who also claims that Google's self-developed chip (in reality a modified Exynos) will replace 10 million Intel chips, simply can't stand the fact that Apple's chip development generates such enormous performance increases year after year. So, he tries to fight it year after year in the short time frame between keynote and launch to get at least a bit of media attention. Okay, at the end of next week the spook is over. Until next year.

DED, is that you?

You swine!!

lowededwookie · September 16, 2021 7:25AM

swineone said:

RudolfGottfried said:

appleinsider said:
Apple failed to detail the relative performance gains achieved by its new A15 Bionic SoC during Tuesday's
iPhone 13 reveal, a rarity for a major hardware launch.

Not true. Apple made exactly the same kind of performance claims as last year. No comparisons with the previous generation were made there either. So why does the media continue to serve this clearly false narrative? On the one hand, of course, to generate clicks, and on the other hand, to promote moronic elaborations like this one from Daylan Patel. He already spread stupid stuff about the A14 last year, and this year you could also set the clock until he would once again spout absurd nonsense about Apple's chip development on his ridiculous rumor platform. This time the hook is: Apple is suffering terribly from the loss of engineers and therefore had to postpone its core development. Yes, of course. In 2019 about half a dozen Apple engineers (and not more than 100 as good Daylan claims) went and founded Nuvia, a startup that until today has produced nothing more than hot air and bizarre graphics sheets, but created so much wind that they were bought by Qualcomm. And since then, nothing more has been heard from them. Two more engineers went to Rivos this year and that's it. Daylan, who likes to speculate about how unbelievably startups will change the chip scene and who also claims that Google's self-developed chip (in reality a modified Exynos) will replace 10 million Intel chips, simply can't stand the fact that Apple's chip development generates such enormous performance increases year after year. So, he tries to fight it year after year in the short time frame between keynote and launch to get at least a bit of media attention. Okay, at the end of next week the spook is over. Until next year.

DED, is that you?

Whatever happened to DED? I haven't seen anything from him for ages.

lowededwookie · September 16, 2021 7:27AM

Personally I'm not bothered by a lack of CPU upgrades. Back in the day Commodore offloaded all graphics stuff to the GPU and all audio stuff to the Audio processor. This meant the CPU didn't have to be as powerful because it wasn't doing as much work. Given that GPUs these days do a lot more heavy lifting than they previously did the supposed lack of a CPU upgrade isn't that much of an issue.

Report suggests Apple's A15 Bionic lacks significant CPU upgrades due to chip team brain d...

Comments