Key M1 Mac Engineer Departs Apple for Intel
macrumors.com
macrumors.com
This seems like a meaningful (re-)hire for Intel. They're obviously weak in SoCs. The Atom was an expensive flop, AFAICT.
You've magnificently missed the point of my comment.
And then, of course, for most of the relationship Apple was quite unhappy with what Intel provided, ultimately deciding to take the matter into their own hands.
Good for him.
When it's one person they probably rely on the fact that the team and structure in place can take a loss. When it's a mass departure they will give bonuses as they have done recently with the possible departures to Meta.
I don't get this. Saying you can't raise someone's salary above a certain % is pure arbitrary BS. There's no rule they can't do it unless they create the rule. This is what keeps salaries depressed if you stay at a company long enough. They'll have to pay market for your replacement when you leave... I mean it's just stupid.
From management perspective it’s unwise to pay more than you have to. And if management is competent, then the departure of a few should not be a hardship. That’s a big “if”.
Now, the great resignation may change this calculus, that remains to be seen.
He might:
- already be rich enough to not care at all about making more
- disagree with his emplyer's ethos
- have had an asshole for a boss.
If Intel is offering a competitive salary and promising interesting work, this is basically a zero-risk move for him.
Apple's star on the other hand is rising. The M1 is simply amazing. The article states this guy headed up the transition from Intel to M1. If so, he did an amazing job. Prior to the initial M1 release, I thought there'd be an annoying if not painful transition with lots of issues. To the contrary, it was smooth and a monumental success by any measure.
So Apple would be paying through the nose to keep this guy. Intel, being in shambles, would have to pay bank to lure him away. I'd honestly expect his compensation at Intel to easily be $10m+ per year. Easily. Good for him. I really hope Intel recovers. Competition is good.
> Intel CEO Pat Gelsinger in October said that he hopes to win back Apple's business in the future by creating "a better chip" than Apple can make.
I don't know what to make of this statement. Apple wants to control every aspect of the product. They don't want to be dependent on Intel for chips. They initially moved to Intel because Intel at that stage was the clear leader and PowerPC just couldn't compete. Now Apple has its own silicon it's never giving that up so it's so strange for him to say this.
EDIT: Corrected POWER to PowerPC.
Yup. There is zero chance they're going back to Intel. If a competitor could actually deliver better silicon than them, they'd be going to AMD, because AMD actually builds custom solutions and can be coaxed into at least getting decent security (see: Xbox), but that still leaves the power management gap on the table, and would force another ISA transition. It would be a monumental failure on Apple's part if their silicon engineering fell back to the point where it makes sense to give up all the control they have and the benefits that come with that to switch again to third-party chips. It's not going to happen unless Apple somehow ends up so far downhill that they are on the brink of failure and have no other choice.
OTOH, if Intel starts offering its fabs to third parties and somehow ends up ahead of TSMC again at a competitive price, sure, Apple will totally go for that. But they will definitely keep doing their own SoC designs.
AMD has an ARM architectural license, they've released ARM processors (e.g., the Opteron A-series, the AMD Platform Security Processor, etc.), and were at one point experimenting with a common ARM/x86 platform (Project Skybridge).
If Apple wanted a third party to design an ARM-compatible processor for them, AMD would be capable of doing so. Indeed, when there was speculation last year about AMD developing an ARM processor, I suggested that Apple would be the only customer where that would make sense.[1] However, given the success of M1 and Apple's scaling strategy for it, it's unlikely that AMD has anything to offer right now that Apple would want.
AMD do have an architectural license, but they have never used it. All the ARM cores in their products have been licensed from ARM, not in-house designs, so far.
There is also some Wireless / WiFI IP Apple could leverage, since Broadcom / Apple relationship has been unstable for years.
Intel is pushing benchmarks on a 45 watt processor and trying to compare it to M1s using half the power or less. As a Mac user with an M1 Pro, I don't ever want to go back to a laptop that requires the fans to spin up all the time like my 2020 Intel MacBook Pro. Apple has shown that it can continue to churn out processors that are better than Qualcomm's (or anyone else) on the mobile side. It seems likely that they'll be able to maintain their pace with the M-series processors and it gives them the chance to build the best part for their needs.
I don't think it's strange for Intel to say that. They want to put the impression in your head that their future chips are going to be so great that the richest company in the world will come back and say, "we're sorry for leaving, could we get some of your amazing chips?" If they say, "we've written off Apple," they're saying that they won't ever make chips that are compellingly better than Apple's. The best Intel can hope for is to make chips competitive with Apple's. If that's the best Intel can do, what's to stop Microsoft, Lenovo, or HP from designing an ARM processor that's competitive with Intel and leaving Intel just as Apple did?
Intel needs to project the image that their future chips aren't just going to be competitive, but so good that they'll win Apple's business back. If they aren't that good, it leaves the door open to more companies ditching Intel.
Even if Intel knows they're never going to win Apple's business back, they want companies like Microsoft, Lenovo, and HP to see investing in ARM processors as a waste of money. If Intel processors are going to be so great in the next couple years that Apple is going to regret their decision, then any money I spend trying to create "Lenovo Silicon" ARM chips is just wasted money. By the time my chips are ready, I won't want to bring them to market since Intel will be so far ahead.
Likewise, it reassures shareholders and your board of directors that you believe that Intel's future is so bright that you'll leapfrog Apple. Your comment on how far Intel has fallen and how it's a disaster is a perception held by many. Apple leaving Intel was a very public rebuke of the company's ability to deliver good chips. Saying that you hope to win Apple back puts the impression in shareholders and directors minds that there's no reason to question Intel's new trajectory (new as of 2021). Sure, they haven't been able to deliver on that yet, but their new roadmap looks fancy. Intel 7 and Alder Lake sounds like a return to the Intel of old where they're improving their process - 7 is a nice shrink from 10, right? And Intel 4 is coming real soon! That should keep them competitive against 3nm from TSMC, right? To be fair, I think Intel's trajectory is improved, but not enough to surpass Apple. But management will want shareholders/directors to stick with them long enough for their plan to play out.
It is initially strange to say something like that, but it serves an important purpose for Intel.
I assume this is CEO-speak to Intel shareholders to keep them from getting skittish.
What Intel should be worried about is whether Apple is hiring some top tier server systems designers. If Apple is able to leverage their M1 work into server systems that are fast and power efficient, they can hollow out Intel's profitable server business.
Thinking not only that desktops still matter, but that the current proportion of desktops matter where the future is concerned, and not looking at growth rates? That's a grievous error.
https://www.tomshardware.com/news/arm-6-7-billion-chips-per-...
Graviton 2 has an absolutely huge number of cores (64), but lacks the memory bandwidth to keep them all fed at once.
>the L3 cache of the Graviton2 was shared amongst all its cores, and we also discovered how only 8-16 cores were able to saturate the memory controllers of the system.
Scaling linearly across cores might be easy for some workloads, but for anything that even remotely has some kind of memory pressure should see greater slowdowns given that all the threads are competing for the shared L3 and DRAM resources.
https://www.anandtech.com/show/15578/cloud-clash-amazon-grav...
There's more going on than "ARM is ARM".
I think it's pretty clear I shot that all to hell by pointing out that M1 is ARM, and ARM is going everywhere, now and in the future.
Now, it's quite possible that the architectural details you point out may be design foibles for most workloads in just the Graviton2 series, but the writing is on the wall, the other 3 walls, the floor, and the ceiling that the near future is not x64 but some combination of ARM and RISC-V
People should actually learn about CPU architectures and hardware design.
AArch64 is the ARM architecture API supported by every Linux variant that has an ARM build that I've ever seen to date. The very reason https://asahilinux.org/ was able to get up to speed so quickly is due to this fact (although they're still working on driver support for various subsystems).
If Intel has already licensed ARM, then what the hell is taking them so long? The M1 runs Intel code at similar speed but 1/6 the power cost. The fact that Apple pulled this off, and not Intel themselves, is hilarious. This would be like Intel coming out with an x64 chip that ran PowerPC code slightly faster than PowerPC itself but at 1/6 the power. x64 is a dead-man-walking CPU API, and it's about damn time.
And M1 is a process node ahead of Intel's offerings.
Intel fell laughably far behind on the manufacturing side, but people touting M1 as some design achievement and not simply a more cutting edge manufacturing process don't know what they're talking about.
The big gains in CPU performance in the modern era have always come from process node jumps. The big/little concept is a great architectural change for mobile as well. But you're not going to get substantially improved performance without improving the manufacturing side.
Both AMD and Intel's 5nm equivalent offerings will perform similarly to M1 on both power and perf.
For example, desktop processors will use up to 3-4x power as mobile processors, yet not nearly 3-4x more powerful. More like a dozen or two percent at same core count.
M1 wouldn't be nearly twice as performant if given twice the power. Just as any other processor.
The early indications of next gen Intel/AMD mobile processors is that they're looking to be more powerful than m1. Right now they are a technology generation behind.
It's like comparing performance of PS5 to Xbox 360 and saying Sony is a better hardware creator.
I'm not knocking m1 at all, it's a good achievement. But people are worshiping it under false assumptions and incorrect comparisons. The primary advantage apple had was to buy out all the 5nm production at TSMC before AMD could
Of course it isn't, but it is in the same direction, i.e. A class of products of power efficiency that are not seen in X86 chips.
In terms of data centers/cloud, this is about the most important metric (cost) there is.
> Intel can just license ARM again as they already did once.
Of course they can, but for the larger players why would they want the middle man? The barrier of entry is already gone.
Whether Intel's SoC chips will end up being good is orthogonal.
Competition is not about a party making better stuff than another, it's about been given the chance to work to make it (and not have BS inpediments, like implicit or explicit aggrements to never cross-hire between two companies).
Am i the only one who lacks the fantasy that this could become true?
That's more a message to Intel and their customers than to Apple.
Intel also has little shot of competing with Apple on a talent perspective. They won here, but I suspect Apple has a plan B for who they are going to poach to replace Wilcox, and that person is probably better.
Apple had an entire line of really good wifi routers before shutting it down. A line of servers and storage they also abandoned.
You can never really guess what Apple's plans are because they don't work like traditional business. They have little loyalty to their own products and will absolutely go with an outside supplier if they determine they can do it better.
The integration is delivering dividends. Apple has shown itself to be nimble. If Intel brings a better chip to the table, I wouldn't be surprised if they changed tack once again.
Right now that's a liability for the vertically integrated company because TSMC is ahead of Intel. Maybe it'll stay that way, maybe it won't. If Intel should retake the lead, Apple's choices will be to have a slower / less efficient processor or switch back to Intel.
Even then they might not do it. Just being competitive isn't enough when they'd have to go back to paying Intel's margins. But if Intel pulls it off, it doesn't leave Apple with a lot of great choices.
Power efficiency is not something that Intel has ever led the industry in, regardless of process node.
Even selling mobile SOCs at a loss didn't allow them to break into the handheld mobile device market.
>Intel’s Mobile and Communications Group lost $1.1 billion in just the past quarter and just over $2 billion so far this year.
https://www.extremetech.com/extreme/186367-intels-mobile-div...
Have people forgotten why Apple switched from PowerPC to Intel to begin with?
> Even selling mobile SOCs at a loss didn't allow them to break into the handheld mobile device market.
They're not selling at a loss. They're not selling at all. They spent money to develop something nobody wanted. The bulk of the "handheld mobile device market" is low performance low power low cost processors.
Intel showed up with a medium performance medium power medium cost processor, which nobody asked for. It wasn't cheap enough for the low end and it wasn't fast enough for the high end.
They might have actually made money by charging less for it. That's what the low end of the market cares about, and if they were competitive with ARM on price and better on performance, that could have succeeded. But Intel didn't want the ARM vendors' margins, so what they got instead is nothing.
Where the M1 sits is a different market. "Core i7, but with a process advantage instead of a disadvantage, so more power efficient" would win there pretty easily. See also Ryzen.
We aren't talking about being more power efficient than IBM's desktop workstation architecture of the 90's. We're talking about having better power efficiency than ARM.
Intel has never been anywhere near that, regardless of process node.
> They're not selling at a loss.
Now that they have abandoned their attempt to enter the handheld mobile device market? Sure.
However, to attempt to break into that market, Intel intentionally sold their chips at a very large loss for quite some time.
>Intel is subsidizing its chips — practically giving them away to partners — and recording the transactions as "contra revenue." Intel is shipping its mobile processors and offering refunds or rebates to the vendor for using them. If devices with the processors don't sell, Intel gives its partner, say Lenovo or Dell, a refund.
https://www.zdnet.com/article/intel-to-hit-40-million-mobile...
What we're talking about is the thing that Apple puts in laptops. And there is a reason they switched to Intel and not ARM at the time when they did.
> We're talking about having better power efficiency than ARM.
There are two classes of ARM processors you could be talking about.
One is the kind they put in low end Android phones. These have excellent performance per watt because they prioritize efficiency over performance. They are thereby slower.
The other is processors with performance competitive to Intel Core processors. The only ARM processor in this class is Apple's, and it has a process advantage.
> However, to attempt to break into that market, Intel intentionally sold their chips at a very large loss for quite some time.
This is confusing two things.
One is how much it costs to manufacture a chip vs. how much you sell it for. The other is the net profit or loss of the business unit, including the R&D cost of developing the microarchitecture.
Intel's problem with Atom isn't the performance per watt, it's the price/performance. It's too slow for how expensive it is. It's not that much faster than much cheaper ARM processors, so nobody wants it for that price. What they were asking was so excessive that even after heavily discounting it, it was still too expensive. The only place it really got used was netbooks running Windows where being x86 was the thing that made it worth the premium. So they never reached the volumes needed to recover the R&D.
To compete with low performance ARM processors in the low end of the market, you have to compete on price. Intel doesn't want to do that, but nothing else will work there, because that market is highly price sensitive.
In the high end mobile market, i.e. workstation-class laptops, Intel had the best performance per watt up until the point that they lost their process advantage. Before then no ARM processors in that market had better efficiency because no ARM processors in that market existed.
A lot of the swing of the pendulum in tech to de-vertical is still was collateral damage from the core competencies excuse in the 90s MBA era to basically chop companies up and make the financials artificially better for a time.
This matters a lot when the vendor lacks competition and therefore has high margins. In a commodity market it's a liability.
If five independent vendors each pay money to develop a processor, you get to pick the best one to use. If four independent vendors and you each pay money to develop a processor and yours isn't the best one, oops.
> A lot of the swing of the pendulum in tech to de-vertical is still was collateral damage from the core competencies excuse in the 90s MBA era to basically chop companies up and make the financials artificially better for a time.
It's quite the opposite. Companies that do one thing and do it well result in more stability. If one of them screws up and falls behind, they fail, but that's as it should be, and the damage is limited to that vertical.
Now look at Intel. Their processor designs are fine, their process technology is behind, so they're losing. The same thing happened to AMD -- Global Foundries fell behind -- but because they de-integrated, they could switch to TSMC. Not doing this sooner nearly destroyed the company. Now that they have they're back in the game.
The problem with vertical integration is that it requires all of your business units to be the best in the market. Because if any one of them falls behind, it pulls down the whole chain.
People like to ignore that when they're winning, but the result is that when you lose, you lose big.
Last year's (and somewhat, continuing) chip shortage will combine with a view in the US that chip-availability is a national security issue. At some point, the US will make a strong case for in-nation chip fabrication as a national push, whether that be favorable business conditions for companies like Intel, or negative business conditions for foreign chip providers. Intel already sees this, and is playing to that future game; maybe not necessarily win Apple back ( as they say ), but to play to a future environment where many companies that aren't 3T mega corps to not have very many attractive domestic options other than Intel.
Certainly I'm just playing a guessing game here, but that may be a potential future market they pitch to someone. "Apple will continue to play their own game, but here's a whole other market we see in the future".
As already described in several books, the structure of the company or area matters immensely for innovation [1] [2]. In this regard, a director can have great influence on this structure by knowing which buttons to push.
Prior to working for Apple, Wilcox was at Intel where he served as a principal engineer on PC chipsets
Edit: Sorry, low value comment. Seeing it written out, it just really struck me.
From the perspective of the people in the other side in that latter case everything is fine and everyone else are just whining. You can especially see this in places where status is uneven (e.g. the guys who went to the right schools call the shots) or where there’s some kind of hazing tradition (e.g. medicine or law have brutal startup loads for early career people but are resistant to change because all of the senior people have, by definition, been the ones who were able to deal with it and changes won’t benefit them).
The thing about Intel is, there has been a lot of bad press lately, specifically about attrition of good staff; so I am puzzled why a star would enjoy working with who's left. Maybe they have personal connections and know better, or are going to bring friends over.
I feel bad being so negative about Intel. I know if I was there and reading commentary like this I would be disappointed. So I hope this is all just overblown by the media and I apologize for buying it.
He likely had the power to negotiate all kinds of working conditions, budgets and general other power before he signed on the dotted line.
A high-level, high-demand executive has a very different hiring experience than the average engineer.
But he's also expanding his responsibilities, going from just Mac SoC design to being in charge of Intel's entire division.
Apple also had their fair share of Silicon Engineers leaving in the past 3 years. What happened to Apple being the best place to work?
Was it?
Ever?
It's much tougher to measure other metrics, as how much you value autonomy and "not spending your entire day triaging emails" is going to be a lot more subjective.
If someone thinks it’s a terrible place to work at, more money won’t make it “unterrible”.
I've been working at Apple for the last 17 years. At my last perf-review (which went very well), my boss was talking about promotion and where I wanted to go in my career: ICT6 (really hard to get, ICT5->ICT6 is based more on opportunity becoming available rather than ability IMHO) or Management. ICT6 comes with a $200k or so pay rise, a significant fraction of my comp, so it's a big jump.
My answer was simple enough: I plan to retire within the next 5 years (I'm 52 now); I'll go whichever path will make the most money before I retire. I'm willing to put up with a non-ideal work environment for a while in order to reap more financial rewards that might let me retire even earlier.
(I have an excellent working relationship with my boss, who is an adult and acts like one; we can talk openly and honestly about these things)
I make a lot of money at Apple, and I like working here, but it's still work and I work to live, not the other way around. I want to be able to enjoy the fruits of my labour while I'm still young enough to appreciate it, so my drive is simply money. It's evolved over time, and certainly when younger, "cool project" was more of a drive, but now that I'm approaching end-game ... things change.
There's a reason companies rarely reach 100 years old: they commit suicide.
If all of the engineers you work with move over to Company B and the leadership of Company A is recycled, what is left at Company A for you to be loyal to other than the name?