And the original report was released on 01/12 before Intel appoints Pat Gelsinger as new CEO. So even if the report speculated correctly, which is unlikely, the circumstance already changed.
The only problem with this, is that its fairly well-known by industry insiders that Apple has TSMC's entire 5nm production locked up for a significant portion of this year, maybe even the entire year. M1 is just the beginning. New Axx chips are going 5nm too.
I would trust Charlie over at SemiAccurate before I trust TrendForce.
I'm not saying this isn't true or accurate, I'm saying that Apple's manufacturing demands are enormous just for iPhone / iPad chips, and now M1 has invigorated demand for desktop / laptop Mac products, at least amongst nerdier types like us, but wherever we go, the mainstream inevitably follows. Once "normal" people start using MacBook Pros and realize they now have 2-4 day battery life (or even longer for light users), then its only a matter of time until demand rises. I think Apple has anticipated this to some degree.
Not to mention, 2021 is supposed to be the Year of the Mx iMac / Mac Pro.
Ultimately, this is just a terrible time to be Intel. They're at least 2-3 years away from any worthwhile new product on their own nodes. 10nm is still a shitshow and 7nm isn't faring much better.
I was a super MBP user until a few years ago when I got into 3D printing. I started doing lots of CAD stuff and found Windows users... Lots of them. Then I got into CNC at a hobby level and I find most of that tribe is heavily invested into Windows. Looking at the non-hobby level of CAD users, that is, professional engineers who use CAD... I see lots of Windows users.
That is not to say Intel isn't in trouble, but I have exactly zero interest in running MacOS anytime soon. I don't care if the M1 MBP has 15 day battery life. I need a discrete GPU and loads of RAM and I want it in a Laptop, understanding that I need to plug it in after 1 hour. There are dozens of us!
Your issue there isn't so much the hardware, but the software. Any MBP with discrete graphics gives reasonable performance running Windows given the form factor.
Everyone I know wants an M1 laptop, has bought one, or is waiting for a larger laptop with it in it. None of these people are nerdy types, but the buzz surrounding the M1 is loud.
> Everyone I know wants an M1 laptop, has bought one, or is waiting for a larger laptop with it in it. None of these people are nerdy types, but the buzz surrounding the M1 is loud.
Now go to the most fashionable nightclubs that are the hardest places to get in, in America's largest 50 cities.
Ask them what they think about Apple's new M1 CPU / chip and if they are planning to get a MBP/MBA/Mac Mini.
Now go ask the top physicians in those 50 cities. The top psychiatrists. The top fashion designers. The top <anyone who isn't intrinsically involved in technology>.
I bet you less than 10% of them even know that Apple released a new MacBook and MacMini.
Performance isn't the draw for these people, its battery life.
If you want to sell 2 billion smartphones, give the people a phone with 75% of the performance of an iPhone 11 or Galaxy S21, but a two week battery life.
Most people that I know use some combination of Windows, Linux and older Macs. None of us have upgraded to M1, nor are we in any hurry to do so.
This comment is typed on an older Mac that will be replaced with either Linux or Windows when it dies. Because my experience with Apple has been going downhill, and I'm tired of dealing with the bugs.
So I work in tech in SV and at least a third of the engineers I know have ordered an M1. The ones that I know that have been using them sing their praises (with some teething around Homebrew and docker). I ordered a Mini a week ago and should receive it tomorrow. I am hoping to be as blown away as everyone I know was.
I work in tech in Orange County and nobody here particularly cares about the M1.
It may not help that I work for a company centered around manufacturing, and factories are deeply Windows centric. (For the Linux fans, it turns out that when the Linux box fails at 3 AM, and nobody in the factory knows what to do about it, it gets replaced by Windows ASAP. Doesn't matter why it failed, they want someone around who at least knows where to start on diagnosing it.)
I would guess that my experience is closer to average for a software developer outside of the SV bubble than yours. And developers of all kinds are going to be faster to upgrade computers than the general public.
So by SV I mean companies based in SV, however the dev team is all over the world. Anyway, I received my M1. Pretty blown away. I have 2 rsync checksum task running on 100K files each (one local between drives, the 2nd from a drive to NFS mount), Minecraft running (AFK), Music importing 90K files, a browser with 20 tabs open, a large compile job of a re-install of 100+ Homebrew apps going, the 1st Backblaze backup of the system, the 1st Time Machine backup of the system and , an Xcode job of a personal build, and a few other random task for setting up a new system running and there is ZERO slowdown in an responsiveness anywhere. The activity monitor shows the system is bored and the fan is not running. It is freaking impressive.
>The only problem with this, is that its fairly well-known by industry insiders that Apple has TSMC's entire 5nm production locked up for a significant portion of this year, maybe even the entire year. M1 is just the beginning. New Axx chips are going 5nm too.
Maybe I'm not following you, but if this is true, doesn't that just confirm parent's comment that Intel is likely not outsourcing to TSMC's 5nm process? Because TSMC has no 5nm process availability to outsource to?
They’re not even close to the bottom. Their current product line up isn’t the best and the future is bleak, but the market needs their fab capacity. The bottom will occur as competitor production capacity relative to demand improves. Part of this is pandemic related, too, as COVID has reduced production and caused a spike in demand.
Meanwhile, there's a shortage of foundry capacity. Automakers are pausing assembly lines because they can't get chips. So yeah, Intel isn't making the best chips, but they're still making ok chips, there's a chip shortage, and things could look different in 5 years.
Not saying I'd bet on them. I'd probably not bet either way. I just think the "future is bleak" narrative is overplayed.
They will recover, but we're not at the bottom yet.
I think the stock will hit rock bottom in early-to-mid 2022. It'll start to climb around Q2 2023 when new products on working Intel nodes are announced. Once those products are reviewed and performance is at parity or better than AMD / Apple products, it'll rebound.
And they will have to recover, one way or another. Intel is now a strategic asset. TSMC is too close to mainland China to risk losing access to, the Arizona fab won't be online for at least 3-5 years, and its capacity is tiny compared to the main fab in Taiwan. Samsung can't afford to share capacity because whatever portion of their fabs aren't used for Samsung products, the rest is locked up by NVIDIA, which literally every researcher needs because GPU-based compute is now dominant.
Intel has to succeed, and if they can't do that on their own, the United States government has to step in, in some way. They did it for a bunch of shitty bankers in 2008, they can definitely do it for an industry that produces an actual, tangible product.
Getting the government involved does not change the makeup of engineers or organizational antics; if anything, government involvement makes the entire thing worse. Intel has soft resource problems (managerial, talent), not hard resource problems (money, access).
If Intel is not shook from the top down, this story will not end well.
The thing that US government needs is to ensure that at least a reasonable part of modern chip manufacturing happens on US soil and not across the Pacific. The way they can step in for that is with subsidies for US-based fabs. However, do they need to be Intel fabs? TSMC might as well fulfil that need if they can be motivated to build fabs in USA.
If it has come so far that the government has to steps in than they are probably done. Once government starts to bankroll them, they'll have no incentive to be competitive and all the smart creative people will leave for better pastures.
Further back, modern computing / internetworking wouldn't exist without military and DARPA cash from the government. Read any book on the history of these.
There's a difference between the government paying for or subsidizing a needed service and the government bailing out a specific company after that company has failed to compete in the market. Successful industrial policy, like the sort South Korea practiced, requires backing the winners not rescuing the losers.
I don't know whether Intel will ever need government bailout. If however it does need it, it will probably be game over for them to ever become a leader again. There will simply be no incentive for them to innovate.
Lockheed and Boeing would both easily get government bailouts if it came to it, for the same reason I believe Intel would.
America can no longer cede ground in high-tech manufacturing to anywhere else in the world.
Hong Kong should have been the wake-up call for this. If China can subjugate Hong Kong, they can subjugate Taiwan. I'm not convinced the average American would support a war against China to free Taiwan, even if it did mean no more iPhones for them.
So the only other option is America has to return to being a bastion a high-tech manufacturing. The Wall Street types don't like this because it doesn't make them enormous margins. They're under the belief that offshoring everything will be fine, because to their mind, commerce and trade are the most important things in the world. They lack the mentality to understand that Communist China simply does not share this mindset.
That should be obvious with China "re-educating" their version of Elon Musk or Bill Gates or Mark Zuckerberg or whoever you want to compare him to - Jack Ma. No one is above "the Party". This isn't something that Wall Street types understand, because this is extremely bad for business, and they don't "get" why someone would sabotage profits for ideology. I would call it a Western failing, but frankly it was America and our Allies that failed in World War II. We should have listened to Patton and finished off Stalin, and then went to work on Mao, but we lacked the courage to do it.
If you really wanna go nuts, the stock split king for tech is Microsoft, with 9, yet its also at its highest point per share than ever.
Stock splits aren't a good reason for a low stock price, in fact, usually if a stock is hot enough in demand to split, after the split, the prices almost always eclipses the pre-split price.
> that its fairly well-known by industry insiders that Apple has TSMC's entire 5nm production locked up for a significant portion of this year, maybe even the entire year.
This is clearly incorrect. Qualcomm's Snapdragon 888 chip is made on 5nm, and they'll be in every Android flagship phone.
Why is it so hard to horizontally scale a working chip fab (i.e. copy-and-paste the entire plant to get more throughput)?
Is it a bottleneck on staffing people with the required expertise (i.e. are the only people who can run a given process node, the same people who designed/operationalized that process node)?
I read somewhere that scaling semiconductor fabs is incredibly difficult affair, as every minute detail have to be faithfully replicated - down to using the same color of lighting. All that is required to get consistent output quality.
Seems like the best idea would just be to try to plop the new plant next-door to the existing plant, then (and to keep doing that.) Would make the logistics of sourcing the same "everything" a lot simpler.
Intuitively, I'd expect that fabs would plan for this in advance, ensuring that 1. everything they buy, they have ways to get N of the exact same of that thing down the line — the same way that NASA missions contract to ensure they can get precisely-specced replacement parts decades down the line; and 2. that they build their first plant for a given process node in a huge empty expanse of land (much larger than what Plant No. 1 needs) where they either buy all that land ahead of time, or at least have a contract with the municipality to retain right-of-first-refusal on purchases of that land.
There's also a long lead up time to opening up a new fab. If you have just proved out your process, it might be tempting to duplicate that but when you're chasing the cutting edge as these markets are, there's a real opportunity cost to consider in scaling out your now "mature" process vs chasing the next big leap.
Maybe now that there are only two players left and new process costs continue to escalate we might see more of that
Despite the people, water, and electricity are another issues.
Taiwan is a small country with limited water resource due to the lack of land to have long river to keep the water on the ground.
For the electricity, nowadays, many BIG TECHs demand "green" electricity to manufacture their products, but once again Taiwan is not a big country. There is no enough area to hold the solar plates and the weather there is not suitable for wind power because of the typhoon.
These are the nature limitations. There are also other issues such as lack of engineers, the risk if expansion is not worth it in the future, etc.
IIRC a solar farm is being built in Australia which isn't going to touch Australia's own power grid, but rather is going to connect to Singapore's power grid using a pair of undersea HVDC cable runs.
Taiwan could probably get in on that action. If you're already running a 3500km cable, you could run a 4500km cable just as well.
(Kind of annoying that Taiwan can't just buy green energy from China — very easy logistically — but that's extremely politically untenable.)
Apple booked 80% of TSMC 5nm for 2021. AMD hasn't even started making 5nm CPUs this year while Apple has been shipping them to consumers since September of last year.
I don't know about volume, but Apple sure has more margin.
Outsourcing fabrication doesn't guarantee that you can choose the best fab, it just guarantees that you get to compete with Apple (and NVidia and, now, Intel) in a contest of "who can pay the most per chip."
The more credible speculation I've seen (although arguably once you're arguing with speculation is more credible, you've already lost) is that Intel is doing a multiyear deal to have TSMC crank out its discrete Xe graphics/compute cards. Which frankly makes a lot more sense, they need a leading process to compete with NVidia who currently has access to Samsung's 7nm and AMD who has TSMC's 7nm and will probably move on to TSMC 5nm sometime next year as Apple moves on to TSMC 3nm.
So supposedly Intel is guaranteeing enough Xe business to get TSMC to convert part of its Baoshan complex over to making Xe. Intel will keep trying to fix its own fabs for its CPUs and probably intends to bring Xe cards back in house at a later date, but they also want to go challenge AMD and nVidia for datacenter GPUs now and that means getting access to a better process.
Outsourcing the commodity i3 whose main goal in life is run MS Office on a Dell Optiplex is definitely a strange rumor. Out of all the things Intel has, that's the area that has the least need of a process upgrade.
Feels like we are quickly centralizing consumer chip fabrication into a single company. I guess the barrier for entry is so high and TSMC is just so far ahead.
TSMC only just pulled out in front. Rewind the clock a mere 5 years and Intel was in front, with Samsung and GlobalFoundries basically tied for 2nd, and TSMC in dead last (they had the weakest 16nm/12nm of that generation - the only one who couldn't hit 30MTr/mm2 of the bunch)
GloFlo then backed out entirely of the race and Intel slammed into a wall.
Since then Samsung and TSMC were on "equal" ground at "10nm" (both ~52MTr/mm2, both released 2017) and again at "7nm" (both ~96MTr/mm2). It's not until 5nm that TSMC was actually clearly in-front of everyone else, with their 5nm being 173MTr/mm2 while Samsung's is only 127MTr/mm2.
In terms of "TSMC is just so far ahead." Samsung's 3nm is supposed to use GAAFET while TSMC's 3nm will still be FinFET. So.. potentially Samsung re-claims the "crown" so to speak at 3nm. And Samsung does contract out their fabs - see Nvidia's RTX 3000 series. There's also no particular reason to believe that Intel is down for the count for good. They are a huge company with a huge amount of capital, they can fund a rough generation or two.
I remember the huge controversy around the iPhone 6S which had a processor that was either a 14nm Samsung or 16nm TSMC (https://en.wikipedia.org/wiki/Apple_A9). Apple had used Samsung for the iPhone 4 through iPhone 5S. The iPhone 6 was TSMC and the 6S was dual-sourced (20015/2016, 5 years ago).
I'd guess that some of TSMC's rise to prominence was partly driven by Apple not wanting to help Samsung. Apple was really pissed about Samsung copying the iPhone - not just by shipping an Android phone, but by copying icons to make their phones seem as similar as possible.
You can definitely look back at MacRumors articles from 2014 and see that there's a bunch of Samsung/GlobalFoundries/TSMC "who will be able to make it happen" talk.
In fact, re-reading through these articles, it seems that people thought that Samsung and its alliance with GlobalFoundries would be the winner of the iPhone 6S generation, but it's possible that Apple saw better yields from TSMC and saw the potential there. When you're as big as Apple, you're going to be really deep with your suppliers and you're going to have a lot of expertise to judge suppliers and their future potential. Maybe it was a combination of seeing TSMC over the iPhone 6/6S generation that gave Apple the confidence to move away from Samsung. Back in 2015, analysts were still expecting Samsung to be getting future business from Apple like the A10X processor.
Given that Apple is a buyer that can move mountains, how much of TSMC's ascendency is potentially Apple committing to a lucrative multi-year deal allowing TSMC to invest a lot of money knowing they had guaranteed orders? One of the hard things in business is knowing what to spend your time on - what do customers really want. Google, for example, has spent plenty of time on things that weren't good investments whether that's Wave or AppEngine or Google+. If you know "doing X will definitely make me a lot of money" it makes it easier to invest heavily in an area - basically, you kinda get the benefit of hindsight ahead of time with a long-term deal.
I hope Intel and Samsung continue to do well (or get back into the race as Intel's position might be) since more competition means lower-cost processors over the long-run. But I think it's definitely important that you point out that only a few years ago TSMC wasn't the powerhouse it is today. While I believe TSMC is going to continue to invest and improve, Samsung is producing Qualcomm's Snapdragon 888 on its 5nm process and if you're right about Samsung's 3nm process, that should provide a lot of orders there too - especially if Intel is willing to outsource manufacturing.
The job of a gate is to "hold open" / "pinch closed" a channel with an electric field. The closer to the gate, the better the hold/pinch. In ye olde days, you'd slap a gate on the top of a channel and call it a day. Every part of the channel was close enough to the gate get a good pinch.
Then everything shrunk and smaller channels wound up needing stronger pinches to completely shut them off. Instead of slapping a gate on top and calling it a day, they raised the channel into a fin and drizzled the gate over 3 sides so that it could pinch from the left and right, not just the top. Those are FinFETs.
The next step is to have the gate on the bottom, too, so that it can pinch from all four sides. The channel literally goes through the gate, which surrounds it on all sides. Those are Gate-All-Around FETs, or GAAFETs.
This is beautifully written; I stepped away from keeping up an in-depth understanding of silicon processing a bit before FinFETs took over and this feels like one of the best introductions to the basics of gate geometry of FETs.
I never quite got how electrostatic control was supposed to work with GAA, Fin and derivatives: without the bulk, how is your electric field supposed to work?
Unless the Gate itself is hollow, and the "bulk" is in the middle, like air would be in a hollow spaghetti? Or tha you use way higher voltages for the gate than for the drain/source (but then, how do you drive that?). I guess I just need to look it up, but most presentations gloss over that part.
In the 90s there were dozens of leading edge fabs. But while the barriers were high back then the capital investment needed to get into the next node has gone up exponentially, about 15% each node, since then doubling every 5 years. It took less than $1 billion to get in the game back then but over $20 billion now.
It's not clear whether this is cause or effect. The insatiable demand for silicon fabbing has arguably made a $20B plant today more economical than a $1B plant 25 years ago.
What’s tricky to understand is that once TSMC said, ‘we will build everything and design nothing’, everyone looked to them when there wasn’t another competitor willing to do the same.
The massive order volume they received (with low margin) let them experiment on process development 10x that of Intel.
Remember when Wall St. said America can outsource manufacturing because ‘other people’ aren’t smart enough to innovate? Looks like they were wrong :)
The American middle class is going to be further decimated over the next decade. Wups
On another side - has the software side done much to incentivize Intel to innovate over the interim? i.e. Windows has accrued a lot of cruft. It feels pretty outdated at times. It feels almost unprofessional at times coming from a unix-like OS user space and has since the 90s. But then unix-like OSes have been a thing this whole time.
Yeah, I think Intel is our latest, greatest example of the follies of not regulating markets properly. Properly regulated, Intel would have been smacked around or incentivized against dominating the chip market 'back when'. Given the factors required, this may have had to require state-led/funded chip research and fab production. i.e. Effectively making a government do what TSMC did years ago - and privatize the actual results of the efforts at key points/areas. Just enough public investment to push the market toward efficiency then back out. Proper regulation would have kept pressures in place to keep key technologies and manufacturing processes/abilities from entirely leaving the local market, too.
99% of the barrier is feature size. Producing the feature size starts with photolithography.
If you want to talk about centralizing concerns, look into the number of companies who can produce an EUV light source capable of supplying a photo tool with powerful & precise output 24/7/365.
The main problem is that there is no profit in chip manufacturing ( relative to software development). Apparently it takes 5-6 years for a node to make any profit.
> The main problem is that there is no profit in chip manufacturing
This isn't true. Check out the margins of TSMC, it tends to be 40% or so.
But to maintain this lead in the industry, they need to massively reinvest for the next smaller process. With that said, profits are great but they don't endure (software development is somewhat more 'sticky' especially since everyone is doing SaaS which provides more incentives for competition and many companies are growing even with covid19 changing the market landscape).
Maybe it is so with the monopolized ( or duopolized if we consider Samsung) foundry today. I visited Global Foundries just before they stopped investing in smaller nodes ( and Infineon which was nearby). Listening to the profits they made and the problems they had ( and my own experiences), I realised that the complete business flow for chip manufacturing is flawed somewhere. I did not have enough motivation to look into detail at that time.
Of course, most of the big foundries would still be open, if the profits were anywhere more than 10%.
I'd say it's weird because hardware like this is so much harder, but the benefit's also a lot more marginal. There was a time when you had to buy a new computer every 4 years or it would be cripplingly slow. These days, pretty good hardware that's 8 years old is good enough today if you have an SSD, 8GB of ram, and don't play AAA games.
There are a lot of other innovations possible, but the business flow for hardware manufacturing does not have the right motivations. I personally think that Apple is doing the right things on several levels.
While I understand the concept of "technology node" as a manufacturing process, where does the usage of the word "node" come from? From litography? Is it related to the "nodes" in the electric circuits?
> The technology node (also process node, process technology or simply node) refers to a specific semiconductor manufacturing process and its design rules. Different nodes often imply different circuit generations and architectures. Generally, the smaller the technology node means the smaller the feature size, producing smaller transistors which are both faster and more power-efficient. Historically, the process node name referred to a number of different features of a transistor including the gate length as well as M1 half-pitch. Most recently, due to various marketing and discrepancies among foundries, the number itself has lost the exact meaning it once held. Recent technology nodes such as 22 nm, 16 nm, 14 nm, and 10 nm refer purely to a specific generation of chips made in a particular technology. It does not correspond to any gate length or half pitch. Nevertheless, the name convention has stuck and it's what the leading foundries call their nodes.
Intel will never become fabless. They made the wrong bet on the type of technology to reduce transistor size and they overestimated what their engineers would be capable of creating and doing, but that won't last forever.
There's a lot of working going on with 10nm and 7nm to fix the production problems (because a lot of work is still required). It won't happen overnight, in fact, I predicted about a year or two ago that it would be around 2023, +/- 6 months, that Intel would have its node production problems mostly ironed out.
Sadly by then, TSMC should be at 3nm, and while Intel's 10nm is easily a match for TSMC's 7nm, it won't be enough to be competitive against TSMC 5nm and 3nm. I hope that Intel has their shit together for the 7nm node, or can at least break even on it. If they can break even (cost of wafer is equal to or less than what they can sell the chips for) on 7nm, they'll be able to hang in there until they get the 5nm / 3nm nodes up and running.
If Intel drops the ball though... and they aren't able to get 10nm yields over 90% by 2023 and if they can't get 7nm yields at a reasonable point... well that's a whole different ball game for Intel. It might not be out of the question for Intel to approach the US government about either subsidies, tax breaks, custom manufacturing for military applications, etc., in order to 1) keep Intel viable until the engineering challenges are resolved and 2) prevent the offshoring of all semiconductor fabs.
Put simply, Intel staying competitive is a matter of national security for the United States.
It seems a dollar of government investment now is likely worth $5 of investment down the line, once we've fallen further behind. The US ought to be investing on multiple fronts to regain/maintain the lead here.
Best company I worked at -- small startup -- would always do 2-3 R&D initiatives in parallel. One would be conservative (guaranteed results). One would be super-high-risk (huge payoff if it works). Sometimes there would be one or two more. It was all a big hedge. Some panned out, some didn't, and when we hit the market, our technology was like science fiction. Competition didn't know what hit them.
That would run a fair bit of coin, but fairly little on the national scale.
Nope. I don't identify myself online, and recommend the same to others.
I think the most detail I can give is general class of work it engaged in. Think of MRIs, sonar, radar, ultrasound, back-scatter x-ray, LIDAR, CAT scans, etc. That's the general sort of technology it worked on.
That should give enough detail for the purposes of this discussion...
No this is a move to protect market share and buy time their processors with the current Intel node are not competitive this way they are able to compete with AMD as well as restrict AMD supply. They will be able to hold market share with this till their own next generation fab is fully functional.
Clearly evidence of the "free market" working on human prosperity. Oh, wait, no it's the opposite. Innovation killed over a monopoly's power. Cool, cool, cool.
This sort of anti-competitive behavior should be illegal. How is it okay to just clog your competitors supply line, when they got the better tech you just can't up?
When it comes to fab, AMD didn't beat intel, they gave up and went with TSMC. Now they have to compete with TSMC's other customers, including Intel, for access to the kingmaking process. That's not foul play, it's the bed they made, and now they get/have to lie in it.
The current high end AMD parts are multi chip modules. The CPU dies are made at TSMC in 7nm but the IO dies that glue multiple CPU dies together is made in the Global Foundries 14nm process.
GF has basically given up on 10nm and smaller nodes.
Didn't AMD have contractual obligations to use Global Foundries for their high end chips from when they split, or something along those lines? I guess that turned out really well for them, though, since that might have led to the necessity of chiplet designs.
GF has given up on research and pioneering smaller nodes. Their current process is based on tech they licensed from Samsung. I would not be surprised if they licensed another process from TSMC or Samsung in the future.
Maybe not legally, but if their (partial) intention is to retard AMD's design success, it's at least what is considered "a dick move".
Intel is strong-arming AMD on many fronts AFAIK, I sincerely hope they vanish into insignificance for their dick-locomotion nature. And I also hope the EU succeeds at spinning up their own fabs. And that ancaps one day see the light of regulation.
I want fancy tech and scifi and wealth for everyone, and the free-market doesn't deliver. It's all monopolies and patent wars...
Intel (and NVidia for that matter) don't owe AMD uncontested access to TSMC's kingmaking process, no matter how much AMD fanboys wish it were so.
> AMD's design success
The market is showing us that the value center isn't design, it's fab. Which AMD gave up on. The lack of competition that's squeezing them is the very pile of dung they created by dropping their foundries.
Sounds like you have all this backwards. If the rumours are true, it's TSMC with all the power in this situation, not Intel. TSMC selling their services to the highest bidder because they have the best tech is the free market at work. It's the opposite of anti-competitive. Intel is the one going to pay up (to TSMC) for their mistakes. Incidentally, AMD outsourcing its fab came with inherent risks, one of which is this.
> This sort of anti-competitive behavior should be illegal.
You're assuming the legal system is intended to foster competition and prevent the formation of monopolies. That is not the case. The examples of this occurring are exceptions to the rule. And as evidence, you can examine the concentration of wealth, and investment capital in the USA (or other developed capitalist countries).
Why so? They have practically a monopoly as there is nobody that can actually compete, Samsung just isn't there, same for INTC's fabs, obviously.
In addition, more and more AI/ML accelerator startups show up with even more need for wafers and ASML has 2 year old orders on the backlog (half of their EUV's machines last year went just to TSMC).
We never know when intel or Samsung will catch up, or even get ahead. They have a monopoly today, but a little bit of luck on either part and that is gone...
Sure, assuming that their in-house manufacturing is more expensive than buying externally (quite likely, considering that it was grown for decades more on USPs than on pricing). I wonder how much the "cooperate to learn" effect has been part of the decision. You surely won't get outright trade secrets this way or technological details, but valuable insights in how they think, how they "do things on a high level"
If you want volume from TSMC you make a big order, you sign a contract, and then put down a billion dollars or two and TSMC will build you a nice new fab for your needs.
That's more or less how it works. TSMC doesn't allow itself to be in a position to have to screw one customer for another customer. It would destroy the trust that customers need to put the survival of their business in TSMC's hands.
I don't think so. They have a lot of things to produce in their fabs. Optane, Chipsets, NAND (if they still produce it), sensors, FPGAs, Ethernet Controllers, etc.
They're considering it as an interim solution IMHO. To buy time for their own processes.
It's confusing though because it's to be followed up with mid to high range CPUs as well. So this is a quite long interim period. If they just need a bit more time, outsourcing high end CPUs makes more sense.
Intel's selling off the NAND fabs and never owned a 3D XPoint fab; Optane products are built with 3D XPoint memory fabbed by Micron, with whom they co-developed 3D XPoint.
And it's not cheap or quick to refit a memory fab to produce processors.
Isn't it a bigger step than that, as they have to firewall the design team with access to confidential TSMC info now? If they got another process online in 2 years they'd have to have a different team on it or wait another two years or so.
TSMC could hand Intel everything they know and it wouldn't do them a bit of good before it's no longer relevant. Raw knowledge won't buy machines, train people to maintain them, build factories, tune the manufacturing processes, or make relationships with design partners. By the time Intel gets something running with something they pick up and starts thinking about making deals, TSMC will be off profiting from another breakthrough.
I've just heard that TSMC requires strict isolation for a period of years for any team that gets the detailed design specs that can lead to process knowledge, and that this now prevents several fabless companies from second sourcing without a second design team. I would think it would be even stricter with a non fabless client competitor like Intel.
I find it surprising that TSMC is going along with it, just because I would think it's hard to be sure some confidential information wouldn't leak. Unless this kind of institutional firewall works better than I think.
If I were to speculate, I would say that tsmc has a diversified enough customer base AND doesn’t see intel as a contract fab competitor. I would imagine if roles were reversed there would be a whole lot more concern.
Oh, transistors haven't shrunk that much. I don't know the exact process details, but integration density has been going up( putting gates vertical with FinFET), together with better electrostatic control (GAAFET, kind of a step uf from FDSOI), while gate length remains the same.
Think about it this way: you're doping semiconductors by adding a few other atoms. On the order of 1 dopant atom per 1M intrinsic (undopped) semiconductor. How thin can you make it before there's only a dozen dopant atoms? And then, how do you make sure they are evenly distributed? And in the same way across multiple transistors, to guarantee a consistent performance? Yield is key, at this scale.
Technically, they can still improve density by embracing 3D. A friend of mine works on multi-gate vertical nanowire transistors, for instance, but that's still a few years (or a decade?) away from the fab.
I'm not clear what Samsung makes for Apple (someone might know, but not me), but it makes sense to share the more complex parts. One team does all the expensive design work, and then both benefit because the design cost is over more devices, plus more hardware means better scale factors in manufacturing.
The above is on a case by case basis. So they may share some complex part, and some other (possibly more complex) they decide isn't worth sharing.
Weird they wouldn't put their high performance chips on 5nm first. Surely those have the most to gain? No-one's looking that carefully at Core i3 performance right?
And if it does make a big difference, couldn't they and up with i3s being faster than their own i5s? Which would be a bit embarrassing.
I guess they want to gather experience with TSMC's processes with a product nobody cares about, and then use that experience for the product that actually matter.
An Intel Core i7-6700K has almost two billion transistors while a GTX 1080 has 7.2 billion transistors.
Of course you may be referring to Intel's integrated GPU's, but they are still very complex. They are not merely number CPU's -- their design is very different.
Transistor count != complexity. A GPU is hundreds of simple ALUs stacked together. Minimal branch handling, no real speculative execution, no reordering buffers, etc... Very simple, very slow "cores", and just a whole shitload of them "copy/pasted" together. Some moderately complex management blocks then distribute work to all those ALUs, sure, but still nothing close to the complexity of a modern CPU core.
Which is also why GPUs can be built so large. It's much easier for them to handle defects than for CPUs. The loss of a single core cluster on a GPU isn't nearly as significant to the product's overall performance as the loss of a single core on a CPU, and the amount of transistors you need to turn off to handle that defect is much less.
As far as I understand, the 5nm process will yield lower power consumption[1]. As the i3 is on the lower end of power consumption this may be enough to make it viable for applications that it would not be viable for otherwise. While their powerful CPUs are fast enough (if you look at competitors) and the power requirements there is already so high that the reduced power consumption from the 5nm process won't make them that much more attractive.
Actually there are i9 chips with same TDP as i3 [2]. I would think that the i3 still sells more units and is lower cost, so maybe a lower power i3 will be more attractive to OEMs than a lower power i7 or i9.
Well the Apple m1 machines are fast and have incredibly low power consumption so maybe they’re trying to make their i3 compete with that?
Going to be interesting though, the higher tier Apple silicon machines will probably be shipping before any of these i3 chips so intel will likely be playing catch-up for years at least.
Eh, the cheapest M1 Mac is the Mac Mini at $699, $100 less than the outgoing model and offering better perfromance at some tasks than the iMac Pro. Keep in mind that there is the whole 'profit margin' factor. At $699, the mac mini offers much better perfromance than the outgoing model while costing $100 less. Why chop even more money off it?
Apple's cost in manufacturing has always been rooted in quality components and Software Engineering. They just don't build low end models and likely never will. The fact that you can buy a $300 iPad is actually kinda weird for Apple.
They are probably binning a lot of product right now. They'll eventually need to do something with all those chips that didn't make the cut. This is apple we're talking about. There's surely a 5+ year plan to develop this ecosystem fully. I'd be more surprised if we didn't see a complete line of custom chips across all apple products within the decade.
At least in the US market (I don't have experience in other markets), Apple is definitely not the cheaper laptop company. The average laptop sold in 2019 sold for ~$700.[0] The cheapest laptop Apple currently sells is $1,000. Apple only sells high-end luxury laptops. Most laptops sold are cheaper than even the cheapest Mac.
When you compare similarly spec'd machines from other manufacturers then yeah they're usually in the same ballpark numbers. I don't know I'd say Apple is always more expensive or always cheaper when looking at only the high-end price range.
The MacBook Air used to have an i3 in the $999 model which now has an M1. So in that sense the M1 is comparable to an i3. They both serve the same market.
There are also some high-end Windows laptops using i3s like the $999 Dell XPS 13.
Issue is that Intel will have an obvious problem on their hands:
Apple used the i3 in the Air because it was low cost and low power for good battery life. The M1 fits those needs but then blows the perfromance well beyond what we could call an i7 in this segment. (All in the comparable wattage i7 had 30% max less single-core perfromance and 1/3rd the multicore perfromance)
Intel sells chips like this on a sliding scale from i3 to i9.
If Intel made an i3 that compares favorably to the M1, it would be faster and cooler than any other CPU Intel makes, ruining i7 and i9 sales. It would pretty rapidly piss-off their vendors who seek to upsell expensive skus.
I could see a TSMC i3 core in a different name possibly, like Intel Evo Next or something like that and then them selling it as a premium chip for a high profit.
my thinking that this "i3 on 5nm at TSMC" is just to show that nothing good comes out of it. Basically faction fight inside Intel - fabless vs. owning fabs. The owning fabs faction is much more politically stronger inside Intel and they ultimately are forced to let that experiment, and they will make sure that it will have all the obstacles. Basically it like GM's EV1 or like Sun in the 200x when Linux/x86 was basically a guerilla effort oppressed by Sparc/Solaris, and the Sparc/Solaris faction allowed Linux/x86 only into a low end Sun's market.
Could it be partially motivated by competition? Maybe they are less concerned with performance, and just want to eat up TSMC capacity which would be used by competitors
Yes, they should charge Apple 30% of their revenue.
Also, they should insist to change "Made in China, designed in California" into "Made in China, High-tech from Taiwan, rest of the design from California".
If they don't comply, Apple should be thrown out of the FabStore.
ASML is making more than "just light sources". (indeed the "light source" part of their photolitography machines is something they bought a company from the US for - the entire industry has supply chains spanning the world, so attributing it to countries makes not that much sense IMHO)
>Also, they should insist to change "Made in China, designed in California" into "Made in China, High-tech from Taiwan, rest of the design from California".
Cambridge, England (Arm) should also be in this mix.
It does seem like TSMC has strong leverage since it would take Apple a lot of time and work to move to a competitor.
On the other hand, the incentives seem pretty well aligned. TSMC presumably makes a lot of money when Apple is successful and sells a lot of chips. Having a customer that is demanding and willing to pay for huge volumes of bleeding edge parts helps TSMC build on their lead. Having early access to the best process helps Apple differentiate, and their business model gives them the margins to afford all of this.
Anyway, I don't see why either side would want to really change this setup.
They can do it for a time.
Apple can then use some of the hundreds of billions they have on hand to prop up one of TSMC's competitors, as they did with the RAM suppliers.
Who, Samsung? Somehow I don't see Apple doing that. It's not a particularly populous industry; no other foundry (ignoring the blacklisted Chinese companies) but Samsung, TSMC, and Intel is involved in sub-10nm lithography, to my knowledge.
ASML sells the complete EUV lithography machine. Zeiss is one of their partners that supplies the optics. Trumpf makes the pulsed lasers which turn tin droplets into plasma which then produces the EUV light. They say they have many more research and supply partners, but those are the ones that I've found mentioned by name in press releases.
Do they design the actual tin droplet plasma chamber thing, or are they essentially a systems integrator from a bunch of suppliers putting it all together into their overall design to their specs?
This question doesn't make sense, both options involve designing the thing.
Anyway, it's not even about designing it. The technology had to be invented first. It would be crazy to use exploding metal drops in a vacuum chamber where you want to manufacture microchips if there were any other options but there were zero options before. There are few applications of EUV in general, and currently only one at those luminosity levels. So R&D is a big component here, not just manufacturing. And even if we only look at the manufacturing part, this isn't some "simple" vacuum chamber gadget either, only ~30 machines were built last year and they cost more than 100M$ each.
For comparison, four of these cost as much as one Wendelstein-7X. This is more of a big, barely productized science project shipped out as soon as it works just good enough for the chip manufacturers.
It could be some module within the vacuum chamber that has other uses and was already made by someone else for e.g. scientific experiment uses or something; I'm just asking if that part is made in house or is from another supplier.
Would it be internationally allowed for the US government to step in with billions in loans to Intel to get their act together? Slowly losing the last fab company left in NA is so incredibly depressing.
Is it surprising that the government of a nation would act in its nation's interests? My understanding is that the reason for supporting globalization, free trade, etc was that those policies _were_ in our best interest rather than being some lofty ideology that we should always strive for.
I think there are clearly good reasons to have a regional state of the art foundry available to all comers. It seems likely to be the equivalent of a national lab in terms of accelerating R&D even if they charge a modest profit.
I think having Intel involved would be fine, as long as yeah it was spun off and didn't discriminate between customers.
That is an important observation, but also it is worth noting that Intel reports earnings tomorrow (ie, they're in a quiet period) with the CEO role changing hands in less than four weeks, so presumably they would announce this move with the earnings report (or address it on the investor call) if true.
Not sure I understand your question. Between the end of a quarter and the announcement of that quarter's results, US public companies don't generally make major announcements outside of scheduled events (earnings call, press conference, industry showcase, etc). Even more so if there is uncertainty around the company's valuation -- ie, a leadership change.
Aside from the insider trading comment another user made, with which I agree:
Material information ought to be accessible by all investors in a fair and orderly way. This is the same reason that stocks get halted for impending news intraday, and earnings reports are released outside of regular trading hours (except perhaps for ADRs and other multi-market securities). Scheduled injections of information make it easier for market participants of varying sizes and geographies to receive and digest information at parity with one another.
Legal concerns around insider trading. People in the know of the company are reviewing the accountant reports. If there is a surprise they can trade on that and make a lot of money. It is easy to guess what will happen to stock prices if you real results are off from what everyone expects.
I was looking into TSMC worth and found that they had a patent war with Global Foundries, which resolved: "On 29 October 2019, TSMC and GlobalFoundries announced a resolution to the dispute. The companies agreed to a new life-of-patents cross-license for all of their existing semiconductor patents as well as new patents to be filed by the companies in the next ten years".
Isn't this some form of a loophole to fix the market?
Global Foundries was born from AMD splitting it out, and AMD learned through battle how to stop the giant and his patents from crushing you (or worse, keeping you out) by having your own critical patents important enough to force a cross licensing deal. I guess this is just a remake of x86/amd64 for them.
Someone has to fund Global Foundries with billions of dollars to plan and build out facilities, and that doesn't include the risk that they run into difficulties like Intel with scaling up. TSMC is reportedly spending $28b on capex this year, while GF is planning on spending $1.4b.
TSMC gets another bidder for its 5nm fab, which affects their pricing power versus bigger 5nm clients. In other words, TSMC makes more money both through increased fab utilization and through price elasticity of demand, while Intel gets to squeeze the gross margins of AMD and Apple.
It makes even less sense when you consider that all these i3/i5/i7/whatever designations are just marketing ploys that don't actually mean anything.
I have a number of computers with Intel processors of varying generations. I know by heart how many cores and threads each one has. I couldn't care less whether any given one is an i3/i5/i7 -- only people who have little clue about processors care about those brand name designations intended to command a higher price for about the same performance.
I think this makes sense in the format that Intel may make 5nm i3 CPUs branded as something else, the timeline for these chips is way far off though, 2022 at least. The M1X ravaging will have already happened to the high-spec SKU industry. Intel needs a response framed as 'we are there at that level'.
That should be an i3-like (in design, not name) ultra low power but clocked to the moon CPU. Possibly breaking 5Ghz by a LOT.
Is there a connection to the global shortage of semiconductors having already caused car manufacturing plants to shut down, in the sense that Intel re-purposes their existing fabs to make money in these demand-driven markets? Read an interview with GF's CEO just this week where he says their older (22nm, 45nm) processes see full capacity right now.
No, there aren't any on supply side. It's just overwhelming demand for every kind of ICs across the industry. From 200mm fabs to mid-tier to cutting edge.
Electronics, and semiconductor industries just had the best year on record.
It seems just everything electronic related saw huge sales.
Even the cheapest retail computers aren't running on processors with 22nm microarchitecture.
The legacy semiconductor demand isn't coming from CPU/GPU for PC, servers, or mobile. It's coming from other applications and industries (automakers especially).
Realistically, how easy is it for Intel to switch like this? Is there really some industry standard interchange format they can use to ship their i3 design off to a different fab? I would have expected them to use proprietary formats from top to bottom.
Whilst you can easily point your synthesis tool at a new cell library (this is the tool that takes the hardware description in a language like Verilog and produces a circuit that implements it) there is significant back-end work in getting the best out of any library.
Intel will probably have to rework their layouts, deal with new memory compilers that have produce memories with different characteristics and adapt memory interfaces that what they need. Their implementation engineers will take time to understand all of the new design rules, quirks and best optimisation strategies that come with an entirely new library (given this is an entirely different fab there could be significant differences).
I suspect that's why they're going for Core i3 first, they can get away without lots of detailed work and optimisation to really push the process. A straight-forward (ish) port what you've got and see how it goes will be good enough and will give their implementation engineers experience with the new library that they can then use to work on the higher end.
It's not as bad as you think. From a high-level: Modern Synthesis tools turn your RTL code (which is coded in an HDL or Hardware Description Language) into gates, and then map them to a library of "Standard Cells". These foundry-specific cells are physical plans for an AND, OR, XOR, gates, flip-flops, etc. Once the code is mapped to these cells they are run through a Place&Route tool, which lays out all the mapped standard cells onto a plane, and then wires them together in 3D following a set of design rules from whatever foundry you are using. Finally after verifying the physical properties of the output design, you ship it to your foundry using a industry standard format called "GDS2" which is basically a series of 2D layers for turning into actual lithography masks. Doing this process (commonly called "RTL to GDS2") is non-trivial, but could be done to target a new foundry in <6 months. Now, Intel is known to use some custom layout methods rather than this Synthesized flow I've described, but that's pretty out of vogue and is a vestige of their early days.
At a high level, Verilog, a widely used hardware description language, is portable to almost any fab.
There will be all kinds of layouts for specific subcomponents which are harder to move between providers, but I would guess that the move from 14nm Intel to 5nm TSMC will more than outweigh all the layout based optimizations.
Considering the doubts already at the top of this thread and that I have a distinct memory, though I'll be damned if I can remember where I read it, that TSMC would generally not work with Intel because they prefer long term partners not ones who are using them strategically in the short term and since they have their pick of partners they didn't feel any pressure to work with Intel either. I'm not sure I trust this report yet.
My guess would be this is just a test to start ramping up on the high end... My guess is takes a lot of work to move manufacturing processes and get good yields.
There is zero point in moving a midrange part like the Core i3 to 5nm when you want the performance improvements at the highest end.
If you can't beat them... join them. Unfortunately, instead of Intel beefing up their manufacturing business it looks like they want to focus on design.
In my opinion, Intel will not be able to lead on design alone. Not with Apple, Nvidia, ARM, AMD all very competent on design currently.
How is intel getting all this capacity from the other players (AMD / Apple?)? That is some fantastic negotiation! I thought apple made capital commits to allow factory build outs on leading nodes so they had first dibs on capacity. Amazing to hear that Intel is getting to take over the 5nm and future production from Apple / AMD. Not too long ago they were a fab competitor, now TSMC rolls out the welcome mat!
Risk. Outsourcing all the manufacturing process has several a lot of uncertainty that they need to start figuring out. By the time they decide to manufacture their flagship chips, all the uncertainty will be already in the past lowering a lot the risk. It is better to screw it up with low-end, cheaper chips, than with the ones that represents the brand (high-end)
The long lead time between plan and production means he will be looking 5+ years out. Basically, he is going to need to science the shit out of this situation he will be in in ~5-7 years where AMD, Apple, QC, and Samsung won and everyone else in their client list is jumping ship to ARM Fabs (QC, Samsung) with more quantity small-node Fabs. You're looking at Good Arm Chromebooks, a Surface Pro X that doesn't suck, ETC.
“I’m rich and went to an Ivy, I know that engineers are just as effective when we put the building machines in Asia and we save on tax money! I get a bonus to send my kids to private school too, this is great!”
Someone I knew who went to Harvard Business School had a cynical take like this: the whole system is basically set up to ensure high-paying jobs for the children of the rich and political elite... regardless of whether they are truly qualified.
I don't know if the elite university name brand bias had anything to do with Intel's story but it is a major anti-pattern in American institutions. Don't get me wrong... these are very good schools. It just doesn't follow that all people who went to Harvard are better than all people who went to some state school.
I wouldn't travel on a ship steered only by shipwrights, and I wouldn't travel on a ship built by sailors alone.
An engineering company needs engineers at the top levels, but we all know at least one brilliant engineer who is puzzlingly naive in other aspects of life. The business of engineering is more than "make good devices and sell them at for a profit." A good engineering business needs shrewd businesspeople to keep it running. You need solid management and solid engineering throughout the entire hierarchy of the business.
> It just doesn't follow that all people who went to Harvard are better than all people who went to some state school.
I totally agree. The capability of a student transcends the imprimatur of the institution where s/he learned.
However, business school is also about connections, and learning in an enlightened group often gives rise to a sort of intellectual critical mass that begets even greater learning through discussion and the interchange of ideas. If given a choice, I'd rather study engineering at Cal Tech than some middle-tier school, and I'd rather study business at Harvard than some less reputable school.
I think of it this way: to run a business, you must understand what the business does at least well enough to "grok" how it creates value and to tell the difference between good ideas and bullshitting. At the very least you have to be knowledgeable enough about the details to spot other people who know more than you do about said details and distinguish them from bullshitters.
In technical companies, engineers are often (but not always) the people who really "get it." Yet not all engineers have the other skills you need to lead a business like good communication skills, people skills, decisiveness, etc.
Depends if you want to a) actually practice engineering and b) fork out 70k a year (which if you’re shrewd you can make back by more lucrative SWE opportunities). HBS I totally get
When I was in college it became very apparent what the fraternity and sorority system was- a way for the elite and upper class to ensure that their children met other rich children and kept their money and power incestuously consolidated. Kids and hearts don’t care about class, the only way they can ensure a union is make it where pretty much the only people the kids meet are other elites and 1%.
Yep. If you look at the global semiconductor industry it has been consolidating rapidly into very few organisations from day one. And that’s not even fabrication.
Doubt Taiwan is unhappy about that. The more TSMC provides the world with compute, the more incentive the US and other countries have to keep the island out of China’s hands.
Seems like a great opportunity for a post-Brexit UK. Invest a few dozen billions in establishing a nationalised fab. They have the money, skillset and reputation to be a contender.
Does the UK have good electronics companies right now? (Well, there's Dialog.)
They're mostly known for building cars that catch fire and having some of the worse and siller audiophile companies that sell you special gold-plated power cables to make your MP3s sound warmer.
That's an unfair characterisation. When it comes to cars UK is known for:-
Running Europe's most productive car plant[1].
Building the world's fastest cars since 1983[2].
Being the base for most of the Formula 1 constructors[3].
When it comes to electronics innovation even ignoring companies like ARM and Imagination the UK has a large military electronics manufacturing base.
I don't think that current government is likely to do anything to challenge TSMC but to reduce all that to "gold-plated power cables" suggests that you're not really paying attention.
UK has neither the skillset nor reputation for manufacturing. If anything, it may have the reputation of being even more skilled at financial engineering than Americans are.
1. Nationalized means poorly run. Effective companies emerge through competition amongst many companies, and are guided by profit-motivated shareholders, not barely invested and highly distracted politicians.
2. A semiconductor manufacturer needs to be based in a region with a substantial semiconductor supply chain to be successful, and for the UK to develop such a supply chain, it needs a broad-based shift that makes it less-social-democratic/more-capital-friendly, i.e. a more performance oriented economy.
This is only true if you do it wrong. Read "How Asia Works" to find out how SK and Japan built good nearly-state-owned companies through export discipline.
The article notes much more foundational properties of the booming Japanese economy, like a smaller portion of private sector output being taxed to support the public sector, as more likely causes of its growth.
Instead of being the ones who other vendors outsource their fabs to ... intel disappears into oblivion. Will there be an intel in a couple of years? Maybe not.
Well I guess, now would be a good time for the CCP to mobilize for a sea invasion. Easily defused by letting SMIC breathe some fresh ASML.. I meant air.
And the original report was released on 01/12 before Intel appoints Pat Gelsinger as new CEO. So even if the report speculated correctly, which is unlikely, the circumstance already changed.