I love that idea. They really should. It’s a huge mistake to cede the high end just because they don’t have the resources or desire to make truly high end chips
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Five Years After Apple Broke Up With Intel, Intel is Begging for Money
- Thread starter MacRumors
- Start date
- Sort by reaction score
Yes, but the original point was why letting Intel hit the wall might not be good for the market, and attempting to prop Intel up to keep some semblance of competition is not just "socialism". Without Intel, AMD would be the only mass-market supplier of x86 chips, ( there is one other company with a license to make x86 chips, but they're not doing so on any scale), and if that were the case, AMD would have no motivation to significantly improve their chips or offer their chips at a competitive price (because there would be no competition).
I read through all the comments and it's amazing how some weren't alive or forgot that Microsoft had to bail Apple out with an infusion of cash/stock purchase.
That is when I was able to buy Apple for book value before Jobs came back. Apple was in a bigger pickle than Intel.
It had failed to develop a modern OS.
Failed in a partnership with IBM to develop an OS called "Pink"
Failed to acquire BeOS.
Finally went to Jobs and acquired NeXT and used next as the underpinnings for MacOS that we have today.
Apple almost didn't make it.
That is when I was able to buy Apple for book value before Jobs came back. Apple was in a bigger pickle than Intel.
It had failed to develop a modern OS.
Failed in a partnership with IBM to develop an OS called "Pink"
Failed to acquire BeOS.
Finally went to Jobs and acquired NeXT and used next as the underpinnings for MacOS that we have today.
Apple almost didn't make it.
I remember the PReP and CHRP hardware platforms. Kinda cool, if you wanted to run different OSes on the same hardware, or the same OS different hardware. Even with Intel chips, I liked being able to use Windows on my Mac when I needed to run Windows-specific software. Yes, I know that I can run emulators, but even then…
Now, Apple runs only on proprietary hardware. It has many advantages including thinner, more direct, and faster functioning, but I really miss being able to run other OSes, as well as being able to grab a new internal hard drive or Ram from the closest Best Buy or wherever, and it'll just work.
Now, Apple runs only on proprietary hardware. It has many advantages including thinner, more direct, and faster functioning, but I really miss being able to run other OSes, as well as being able to grab a new internal hard drive or Ram from the closest Best Buy or wherever, and it'll just work.
Zhaoxin also manufactures x86 chips so there would still be market competition for AMD without Intel. But by artificially propping up Intel, you actually remove any motivation for Intel to improve their technology so they essentially become a welfare corporation. It hurts the market and consumers, while Intel execs get paid with free money they did nothing to earn.
If you'd ask me, I'd say the culprit to intels demise is their clinging to the x86-Architecture. x86 doesn't easily scale any further without guzzling tons of electric energy which results in immense amounts of heat produced. The reason for this is one: x86 is CISC (while the ARM-Architecture is RISC), RISC scales easier since RISC instructions are simpler to decode, you can do this in parallel. CISC instructions on the other hand have different lengths so you don't know where the next instruction starts in your Byte-stream until you have decoded the instruction right before in the stream, so you can't parallelize this in CISC-processors. Reason two - and this might be the even more important one - is that x86 has the ballast of tons of legacy to carry around. Intel never had the balls to cut off old braids. The consequences are now visible: x86 can't compete anymore.
So if intel is really the best processor manufacturer there is (as some x86 fanboys claim), they would be able to produce the best processor of any given architecture. If I was the CEO of intel I would make the switch to RISC-V (not ARM, as this would impose license costs) as soon as possible, have a transition period of let's say two year and then ditch x86 for good. Of course I would go to Microsoft and convince them to port Windows to RISC-V and have them to develop a great emulation layer like Rosetta 2 for legacy software.
But I am not intel's CEO and nobody is listening to me …
Kind regards!
So if intel is really the best processor manufacturer there is (as some x86 fanboys claim), they would be able to produce the best processor of any given architecture. If I was the CEO of intel I would make the switch to RISC-V (not ARM, as this would impose license costs) as soon as possible, have a transition period of let's say two year and then ditch x86 for good. Of course I would go to Microsoft and convince them to port Windows to RISC-V and have them to develop a great emulation layer like Rosetta 2 for legacy software.
But I am not intel's CEO and nobody is listening to me …
Kind regards!
ARM is what it is today in a large part due to Apple's own contributions back in 1998, while Apple was still partnered with Motorola.
Acorn is an unsung hero in terms of the world we're living in now. If they hadn't embarked on developing ARM in 1986, we could be living in a very different world, likely with backpack-sized smartphones strapped to our backs ... "Intel Inside".
People clamoring for MORE RAM was always a fools errand. More RAM was never the ultimate solution. Better architecture — both hardware and software — is what brings better performance, and Apple knew this.
RAM headroom is important in many workflows, but not for performance reasons alone.
A bigger sky doesn't make a substandard plane fly faster. M-series was a brand new plane that could zip around in the same airspace, cooler and faster.
To some extent they did not ignore it.
"... “The most important role of managers is to create an environment in which people are passionately dedicated to winning in the marketplace. Fear plays a major role in creating and maintaining such passion. Fear of competition, fear of bankruptcy, fear of being wrong and fear of losing can all be powerful motivators.” ..."
Only the Paranoid Survive Quotes by Andrew S. Grove
64 quotes from Only the Paranoid Survive: How to Exploit the Crisis Points that Challenge Every Company and Career: ‘The Lesson is, we all need to expose...
www.goodreads.com
Deep fear of competition when pretty close to being a monopoly would. tend to push into becoming a bigger/badder monopoly.
There is a difference in not fearing being in a competition and fearing the process of competing (having to compete). It wasn't 'fear of the competition' , but competition itself.
Intel's other problem was fearing not shipping those high dividend checks ( to keep the stock price high).
This is. the chronic problem with using a negative emotion for a very long term crutch for motivation. It tends to b e corrosive and bleed into other dimensions.
Intel had a ARM chip (StrongARM that they had picked up from Digital (DEC) ) . The larger problem is that they have more dubious GPUs. And tried to push x86 into the GPU space.
It was far more so that they had what they needed , but didn't want to listen to the customers.
Timing also. Buying Infeon just as Apple dumped them as modem supplier did not help them long term.
Boards are more about guidance than management. Excessive paranoia and fear at that level is doom. Not diversified enough ... We have to diversify ( buy MacAfee (2010) , Wind River System (2009) ). Buy potential fab customer and make them eat the dogfood (fab) ... Altera (2015) Infineon (201). Buy anything that looks like AI , MobileEye( 2017) , Nervana (2016) , Habana Labs (2019).
fear and paranoia at board is doom. (see post above)
As for "knew GPUs". really????? baseline core processor for a GPU as a x86 core? Really? That was not a genius move. That was a "when all you have is a hammer everything looks like a nail" move.
Technically ARM is, by far, the largest computing platform in the world these days as far as ISAs go…
Pragmatically, yes. at 18:30. it would prolong the inevitable if Apple had not business practices.
Microsoft porting IE and Office to macOS are also basically an investment. In fact, the video talks about how Microsoft was using dropping Office as leverage in the video. The exact $150M amount , no. But Microsoft had walked away from apple and said "tough luck". If Apple had won the suit and gotten the money as the video projects it wouldn't have solved the problem. The settlement didn't solve it.
New products would halep if the major 3rd party software vendors walked away.
Part of the change was that "Windows has to loose for Mac to win". It wasn't just products.
An where are those chips made? Apple doesn't make them. Not technically 'in house' meal is didn't cook the meal at home. When pay someone else to make the meal it isn't "home cooked" at that point. Intel does make chips.
Early rumors was that Apple was working with Broadcomm on this sever chip. Broadcomm was reportedly working with Intel on a server chip ( not entirely happy with 18A progress though. But Intel could be trying to keep them for 14A. )
It could be server chip made from chiplets. The Broadcomm die could be Intel 14A/18A and Apple chip some USA made TSMC chip. Still have to package those dies together. Well , Intel has a packaging dies business also.
This doesn't have to loop in x86 Xeon. chips in the slightest. Intel does lots more than just design x86. cores.
Reportedly Apple has previously given TSMC money upfront for future manufacturing demands they wanted to make. Similar thing could be happening here where isn't so much 'begging' as asking to be the packaging vendor but need money upfront to handle the workload.
TSMC has some 12-18 month trailing fab processes in USA. Apple isn't really the leading bleeding edge TSMC fab process consumer anymore.
AMD is got the first N2 die.
AMD teases its first 2nm chip, EPYC 'Venice' fabbed on TSMC N2 node — also announces USA production of current-gen chips
First 2nm HPC design.
" ... Traditionally, Apple has been the earliest adopter of TSMC's leading-edge fabrication technologies, staying ahead of the industry by at least a year. ...This is not the case with N2, as multiple companies have either confirmed their N2 tape-outs already or are rumored to have reached this milestone ..."
Intel, AMD, and MediaTek reportedly among TSMC's 2nm early adopters — company said to have 15 customers lined up for new process tech
An unprecedented number.
Nvidia is on track to get the first A16 dies
" ...
Nvidia tipped to be TSMC's first A16 customer, ahead of Apple ..."
Nvidia tipped to be TSMC's first A16 customer, ahead of Apple — Feynman GPUs could make full use of GAA transistors and backside power
For the first time, an AI GPU will launch on TSMC’s bleeding edge.
TSMC doesn't do 100% of the chip package construction for Apple. Intel has USA packaged packaging factoring that could take a USA TSMC chip and put it together with other USA made chips. That wouldn't be the vast bulk of overall package prioduction that Apple needs across the whole procuct line. But Intel needs to walk before they run. Some decent volume to Intel would help them.
What Apple nees is chips that are packaged to be put into a product. Not simply just raw dies that are freshly cut from a wafer.
According to benchmarks, the fastest Intel processor is in 31st place behind AMD chips. It's been this way for years, and I'm shocked anyone still uses Intel at all.. and the gap is getting larger. Last year it was 22nd place.
How would Apple get zip? Priority scanner and manufacturing line priority. That’s nothing?And both companies got something out of that. Microsoft got Apple dropping a lawsuit and Apple got Office support continued. In this case, Apple would be getting zip.
And, that was for non-voting shares, and Microsoft sold all those shares by 2003 for a profit.
That has more to do with core count than anything else. Intel has had very large footprint core designs. That has kep their core count down. AMD chiplets approach was far more focused on scaling up core counts (and putting a cap on the core footprint to keep the chiplet smaller (even if that meant given up single thead drag racing crown. ) )
That really isn't true. What is being proposed above is close to what got covered in the "Mythical Man Month" book more than several decades ago. More money solves any problem didn't help IBM then . There are software install products every quarter at companies that have blown up after. millions in nstall costs have been applied.
$4-5 billion didn't get Apple a viable car for sale on the market.
Intel has a deeper need of customers, especially in the chip production and packaging space, than they do of just simple raw cash ( in and of itself.)
If Apple would sign up to make even a narrow range chiplet on 14A that would help save that process offering.
Similarly if Apple bought up some chip packaging ( on Intel Foveros ) that would help Intel rack up some 'wins' in that business also.
Intel has burnt a ton of trust as a partner/subcontractor who can get jobs done right and on time. If there is 'begging' it is to be given a shot at getting to do the work. They need 'outside of Intel' folks to buy/use the fab tech they have to continue forward making takings. ( the open date for the fab in Ohio keeps sliding in part because they don't have the business commitments to fill it up. )
Apple would need more than just TSMC. TSMC primary makes the dies for Apple, but actual chip packages go into products.
If made and cut the die from a wafer that TSMC made in the USA but couldn't get it into. package to put into a APple product then don't really have much.
There is some packaging contractors setting up in the USA now also near the AZ TSMC complex(es). But that likely isn't enough. Nor particularly likely to cover more advanced packaging techniques may need to shift to in the coming generation for the more larger/complex packages.
They did. There was somewhat of a cover up to mask it though. Apple consumed a healthy amount of the more profitable chips of the desktop line up. Apple didn't buy the bottom of the barrel options at the low end of the line up. Nor did they buy the hyper niche extreme gamer chips at the top ( Apple doesn't do extreme overclocking of their systems).
Intel's decision to delay shifting to buy EUV equipment was in part based on squeeze more proift out of the previous gen equipment . Lower upfront multi-year investments squeezes more profit out what was left. It wasn't just Apple piece though. Intel pushes fattest margins in the Data center space also. ( which gave AMD a pricing umbrella to keep their revenues/profits higher also). Mostly this is done to appease the 'next quarter' Wall St folks who care less about long term company health as long as they get the next couple of high dividend checks.
Intel should have cut their dividends years before then did. They just pretended ( and under invested) to keep the stock price higher .
The dubious decisions didn't start when Apple left, but Intel did double down on ''short sighted dumb' when Apple made their move. And it isn't like Apple did not LOUDLY telegraph their move YEARS ahead of time. (e.g.. Intel's quirk moves with Infineon modem product line up were also years before Apple decided to leave on macOS. )
Foundry customers is exactly what Intel needs, and they f***d that opportunity up to a high degree already with pulling out of 18A. Whether or not 14A will yield remains to be seen, but if they don't get a big customer on it, it's over.
Apple won't bet on the success of 14A, that wold be a stupid move, and they will not use Intel as a 2nd source to TSMC either, take way too much money in those nodes to double design, test, plus, no 2 processes are compatible so. there will always be performance differences ... like we've seen 10 or so years ago when Apple dual sourced that A chip from both TSMC and Samsung ...
All that current pour of money does for Intel is making the cashflow look good over the next couple years, the real problem is they need to be acting as a true foundry, on par with TSMC. Laying off all the engineers didn't help.
Intel itself needs to use its own foundries again, if they continue to outsource to TSMC - what does that tell you?
Intel's troubles started when Otellini became CEO, actually Barrett, that's when Intel started to become a spreadsheet driven company and no longer a operational excellence driven one ... and it got worse from there
If Apple signed up for what the US Government did , then it would be just non-voting shares. 'Begging' could include asking Apple to make the Trump/Besset move look good by also buying the same stuff. ( hence get the herd buying the stock so that can later cash out and look good before the herd changes direction again. )
But yesh, if Apple was buying fabrication slots and front loading the payments far in advance to help found equipment acquistion costs., then sure. Apple has done that for other players in the past. Intel shouldn't necessarily be excluded.
But it doesn't always work out.
Sapphire Glass factory in AZ went sideways.
GT Advanced Technologies Officially Announces Sapphire Settlement With Apple
Earlier this week, it was revealed that Apple and its sapphire supplier GT Advanced Technologies had reached an agreement to effectively dissolve...
www.macrumors.com
Eventually Apple flipped that building into a data center when that venture collapsed.
Intel didn't pull out of 18A. The high volume customers didn't show up. Intel dropped 20A to put more resources into 18A.
" .. Surprising news about Intel continues to emerge with the chipmaker vowing to use an external foundry in place of its own 20A process to make the upcoming Arrow Lake processors, amid talk that Broadcom has rejected Intel's 18A process as not ready for mass production. ..."
Intel cans 20A process node
Meanwhile, Broadcom reportedly displeased with 18A wafers
Arrow lake was suppose to be partially split on TSMC N3B and and 20A and they dropped 20A. In retrospect, that problem should have happened quicker ( might have had less disappointed customers for 18A if had got the design kits and the bugs worked out sooner).
What happened with 18A is that most folks dropped out. So it doesn't make sense to spend tons more marketing and external sales, external tool development tracking 18A improvements and following variants , if no one wants to buy it in volume. Intel is still doing 18A for themselves. If I recall correctly there are some relatively low volume customers still signed up, but they aren't 'chasing' anyone new at this point.
Likely not over , just slower. There would be an 18A refinement with another label slapped on it rather than '14'. 18AP. 18AE or something like that. Same thing as TSMC has a N3P after N3E and N3B.
The huge problem for Intel is that they can't particular sell the older stuff either ( mainly had their own internal design tool ecosystem for previous generations ). There is a substantive design infrastructure have to have outside of just the core fab process technology itself to make it reasonable for customers to use the tach.
There is no one size fits all node and a single tool that will work for a very diverse set of customers. Intel has been super slow on the update of that. Intel made noises about trying to find more fab customers back in 2013-2015 time frame and never did the real grunt work required.
Apple doesn't have to solely use 14A. The next Hn. Wn , chips are going to be. TSMC N2 or A16? Probably not.
Apple's silicon portfolio is more than just the bleeding edge phone chip now.
Different fab processes are going to excellent in different areas. The more different chips Apple has to build then the more different areas/opportunities there are.
Intel doesn't just make wafers then also do chip-on-chip packaging (foveros. That isn't solely Intel die only territory).
Intel 'need' to use. 20A for Arrow Lake ( which if recall correctly started out as N3 only) was part of the problem. The core problem here isn't using or not using it is weaving in at the rate points and at the right rate or evolution.
Boundary can't go from having. 1-3 customers ( PC-server, Altera , modem ) to having 15-20 in one iteration on a single fab process design.
Long term fab companies have customers spead out over multiple fab generations. Most of TSMC revenue doesnt come solely from the most bleeding edge process. Have to grow the business.
Intel had tons of overhead/infrastructure built up on interal only coupling between fab and design. The engineers assigned to that aren't necessarily who they need to get to a position where Intel design interacts not pariicularly differently than any other customer. Not going to get customers if all can expect is secondary crumbs that fall off the table after fab primarily just serves one customer. There is substantive about of trust that goes into signed up also. Intel had/has problem where design would ask fab to bail them out of certain things and vice versa.
Intel other issue was that they were trying to be a make everthing for everybody company. They had their hand in just about everything. When stop doing that some engineers are going to get dropped. '