▲ 106 points
back
47 comments
“As the world is no longer peaceful, TSMC is gaining vital importance in geostrategic terms,”
-Morris Chang, former chairman and CEO of the world’s largest semiconductor foundry Taiwan Semiconductor Manufacturing Company (TSMC) in 2019.
And he said something similar in 2017 as well. Turns out he foresee all sort of problems when TSMC became the leading Fab for the world.
The 2013 piece [1] on Intel is also interesting. That was when Intel were about to launch their delayed 14nm. It is rather unfortunate Intel decide not to follow through their Custom Foundry.
The long interview [1] with Morris Chang is interesting. When asked why he went to Taiwan to found TSMC (at a reduced level of compensation) he replied: it was the newness that appealed to me, the difference from what I had been doing for several decades!
[1] https://www.semi.org/en/Oral-History-Interview-Morris-Chang
IMO TSMC has been geopolitically significant for at least a decade when they started becoming the dominant supplier in terms of volume for mobile and compute between mediatek, qualcomm, nvidia, ati/amd and others.
I think Intel tried pretty hard for custom foundry but didn't get much business between IP concerns from competitors and their 10nm failures resulting capacity shortages.
I haven’t been following the nanometer rat race, but SMIC just announced it achieved 14-nm. This means they are only 7 years behind Intel, according to your statement.
Now, TSMC is at 7 nm, and while the smaller nanometers are important for energy efficiency, and especially for mobile devices with a battery, it doesn’t really increase the clock cycle. Actually, nobody seems to talk about gigahertz clock cycles anymore. It’s all about multiple cores now.
Then, doesn’t this mean that SMIC is now competitive against Intel on desktop computers and laptop computer CPUs? And to be honest, I don’t even know the specs of the chip in my laptop. I don’t think I even purchased the fastest chip anyways.
>> the lesson in 2020 should be that technology is inseparable from geopolitics.
I cannot agree more. If we learn one thing this year this is it.
The thing is this used to be a well known truism, for a long time. Not sure when we forgot it ... maybe after the USSR collapsed and everyone read "The End of History" and actually believed what Fukuyama wrote.
It wasn't only Intel making microprocessors in the US. Samsung has one of their largest fabs in Texas. Global Foundries has their best fab in New York. As I understand they're both equipped to make 14nm-class processors. Not that it detracts from the general point of the article.
I assume that multi-chip packages will soon develop as a new source of modularity within chip fabrication itself. There will be common standards for how multiple chiplets can be wired inside of a package, and the OEM will no longer be restricted to having a single supplier like TSMC for the entire "package" component but be free to choose among multiple, freely-competing suppliers ala TSMC, Intel, Samsung, SMIC etc. Then the "geopolitical" need for an advanced, billion-dollar "fab" will essentially be disintermediated away - only a few tiny, perhaps optional components will ever be manufactured at anything like a 7nm or 5nm process. Everything else will simply end up as a basic commodity.
IANoWayAFabExpert but I don't think so. I understand the cost of transferring signals between chiplets is high, so I don't see this as a way forward except for when the cost of such a transfer can be amortised by handing over a smallish packet of work (in terms of data size) that needs a large quantity of CPU, and a smallish amount of data to transfer back. So coarse grained only. Not suitable for general computation anyway.
IMO only. Not my area. Pinch of salt etc.
Perhaps I'm misunderstanding your point, but hasn't AMD proven this wrong at this point? With a correctly architected interconnect you can absolutely do chiplet based general computation processors.
Again, I may be misunderstanding your point...but I think the chiplet future is here on some level.
And I may be wrong, and for sure chiplets are proven (for now) to be a very good way of doing things, but those chiplets in AMD chips are relatively decoupled[0]. They're very large and connected within themselves already. They can be expected to operate substantially independently[0]. While they do have to talk to each other to get memory, probably a lot, DRAM access is so slow that it swamps the overhead of inter-chiplet communication[0]. Shared L3 cache, that's another matter I dunno.
I took it you meant like much smaller components, like ALU from this supplier, DRAM controller from that supplier, cache from another supplier etc, all on separate chiplets, each having to go off-chiplet to speak to the others. I am sure[0] that would cost much power and performance.
[0] serious risk of n00b arse-speak happening here, beware.
It’s not like chiplet was invented yesterday
Typo: 80286 (not 80826), and 80386 (80836), in the opening paragraphs.
I gotta say my brain auto-correct those kind of things.
I would love to see Intel get into contract manufacturing. We need more competition in that space and Intel's position no longer makes sense in the current environment.
They did, It is called Intel Custom Foundry and had customers from Nokia to Ericsson. In the end they failed. And it is no longer a path they want to follow.
Why did they fail?
They are extremely secretive:
One of the funniest stories I heard was about the first copy of the Intel 14nm design rules Altera got from Intel. They were heavily redacted, which is something I had never seen in the foundry business.
https://semiwiki.com/semiconductor-manufacturers/intel/7912-...
Additionally, a contract fab tries to be flexible, easier to design for. They make their libraries & design rules work well with commercial design software. A private fab on the other hand can boost yield & performance by pushing increasingly complex design rules on the design team, and the design team can use customized or internally developed software to make it all work.
So in short, they had the best fab tech in the business but they weren't used to servicing anyone other than themselves.
Surely 'themselves' is a huge enough market, and 'boost[ing] yield & performance' is exactly what they want to do, as the world's largest CPU maker? It sounds perfect. I guess maybe financially the extra perf etc. just wasn't worth it.
(thanks for answer though)
There is no reason why you cant optimise for both. To simply put they wasn't ready to be a contract manufacture, and TSMC set the standard of what is expected to be a contract manufacture in the Fabs industry and Intel failed miserably to meet even the pass mark.
Yes, it make perfect sense to optimise for you own yield and performance until TSMC overtake you and you no longer has the volume advantage. In manufacturing volume is everything.
It is also worth pointing out Altera is actually a Subsidiary of Intel. If that is what they were going through you can imagine about the others.
Maybe they do know how to optimize, but not exactly why anymore?
I've seen a tv-documentary about 15 years ago about their fabs. Therein some reporter asked why there was a bend in some bundle of pipes for no apparent reason. They answered that was there because in another factory there was a column there which they had to route around, which didn't exist at this factory. They explained further to not wanting to risk process variation, by building that bundle of pipes straight again. Since then i have been unimpressed by all that high-tech, being so complex, having to cargo-cult around, instead of being able to understand it from first principles and build according to the situation. Not blindly copying something which worked at another place but doesn't apply anymore, because the place and its conditions changed.
Altera was already working with Intel before they were acquired.
Well, exactly, they are their own best customer. Contract fab was just a way to fill capacity in a lull period. So there was never a strong motivator to make sacrifices to build a great contract fab program.
I guess the software analog of this is that the quality of code usually goes up when it’s open-sourced. Not just because of the contributions from outsiders, but because the original developers feel apprehensive about putting out low quality code.
For a contract to be mutually profitable it has to somehow be better than off-the-shelf. I got to imagine coordinating custom chip development is expensive for both parties.
The problem is that they would be chasing a competitor (TSMC) that is much better at it. They'd have to invest many years to catch up.
Last time they did that they ended up buying half their customers.
Amazing read, but I disagree with this simplistic political conclusion:
> Second, at some point every tech company is going to have to make a choice between the U.S. and China. It is tempting to blame the tension between the two countries on Trump, but the truth is that China, particularly under Xi Jinping, has been significantly hardening its rhetoric and actions since before Trump was elected, and has been committed to not just catching but surpassing the U.S. in technology for years. There is a fundamental clash of values between the West and China, and it is clear that China is interested in exporting theirs. At some point everyone will be stuck in the middle, like TSMC, and Switzerland won’t be an option.
If Cold War has taught us anything, it’s that display of force didn’t yield anything. Dialogue and exporting values did. So yes, absolutely: the escalation is Trump’s fault, because of the insufferable rhetoric of his warmongering hawks. Xi Jinping is a way more tactical and dialogue oriented leader (not a positive judgement on official PRC values, of course) that is being defeated by the proverbial pigeon shitting on the board during a game of chess. Decoupling from China is easily sellable as the anti-communist propaganda in the 50s, but the harsh reality is that in today’s world it means recklessly hammering pieces off of the global economy jenga tower, and hoping it will stay up. Dialogue and severely regulated exchange with the potential enemy is the real weapon that could weaken China. Putting human rights on the table when it comes to trade treaties, for example. Instead, there’s a resurgence of scary nationalism that is exactly what brought the world on the brink of collapse 75 years ago. If I had to choose two words, right now, I would go with “brace brace”.
> Putting human rights on the table when it comes to trade treaties, for example.
Hasn't this been tried before? What happens if they refuse? We do nothing and just keep talking? The absence of any nationalist impulse is even scarier than your 'scary nationalism'. I would note that China has none of this sort of self doubt.
Any recommendations on a pcb assembly house in Taiwan?
To get something made or to work at?
To get the electronics made