Teardown confirms Huawei's Pura 70 contains SMIC 7nm process node A teardown of Huawei's Pura 70 smartphone by an IC research firm revealed the Chinese tech giant is relying on Semiconductor Manufacturing International Corp's (SMIC) HiSilicon Kirin 9010 processor, likely because US sanctions mean the Chinese company can't buy from other sources. "Following the technical analysis of the …
SMIC's (notional) 5nm node still isn't close to TSMC's N3 and Intel's upcoming 20A (supposed to be in production this year).
As I've noted before, Conway's discussion of chip manufacturing in Material World describes the numerous hurdles to producing recent-node ICs. It's not just a matter of "smart". The silicon wafers have to be extremely pure. There's only one known source of sufficient-grade silicon in the world, and it's in the US (in Maryland), and there are only two companies mining it. The variant of the Czochralski process for pulling the cylinders of refined silicon of sufficient purity is only implemented by one Japanese company, in a US facility, and involves numerous closely-guarded trade secrets. So does the wafer slicing.
And then there's ASML and EUV, the part that everyone knows about. Also necessary, of course. And, again per Conway, there are dozens of other specialized manufacturers providing unique equipment that SotA fabs require.
Can China duplicate all of it eventually? Well, yes, if they can find a suitable silicon source, or create another process for refining less-suitable silicon stock. But it isn't going to happen overnight. The amount of basic research and engineering effort that went into the latest process nodes is enormous, even relative to China's population of (domain-relevant) scientists and engineers. And even if we allow for a significant amount of industrial espionage and sanctions-avoidance smuggling.
I would also say that even if complete plans plus the machines to make the sub-components were stolen and setup in a factory in China, it still will not matter because a lot of know-how is just not written down properly. For evidence I point to any company where management fire the older, skilled workforce, in the misguided belief the young new hires will be able to carry on. And typically those companies are not world leaders at *anything* before the layoffs...
By the time all the issues are finally figured out, the machines will be obsolete and the next versions will be out.
The EUV machines are the best ASML can make, incorporating the best lasers Trumpf can make and the best mirrors Zeiss can make - at least three tech world specialist leaders had to collaborate on making that. There are probably more ridiculous niche products made by only one company that I just don't know about, as I assume that the moving parts are of insane quality, etc.
[A disturbing thought is that if the only factory burns down for one of these niche products taking out a percentage of the workforce, humanity may never be able to figure out how to recreate what is lost. The idea that we can always recreate past inventions does not hold universally true - sometimes the idea, or trick, is lost.]
This post has been deleted by its author
> markings on the Kirin 9010 "technically new but remarkably similar" ...
The markings on this newer Motorola IC are remarkably similar to *this* Motorola IC, hmm, sounds fishy.
> exposed Huawei's use of Kirin 9000s
Ah ha, we have exposed you! So much cleverer of us than just, you know, looking up the part number.
> TechInsights referred to as a "repackaged"
Just like the sneaky Chinese, repackaging as though it is cheap knockoff sneakers in a fancy new box. You won't catch us changing the IC packaging and giving it a fancy number that is obviously in the same series. Nope, DIL now, DIL forever!
Who will the almighty land of the free and home of the brave lean on now to try and stop chips being made.
They already meddle in ASML and despite their fuckery have accomplished nothing but setting the rest of the world backward in futile attempts to stop the sun rising and setting.
Whatever will the pearl clutching, bible beating, gun nuts do now.
"Won't somebody think of the children", for good measure.
There are a lot of holes dug on the ground in resource compromised parts of the USA.They will ultimately get built as chip fabs that are still much more expensive go produce in ghan in Taiwan.
They will need to expand the nearby airports for the large cargo planes so the finished chips can be shipped to China anyway (and Vietnam, India , Malaysia etc) for end product manufacturing… as all of that was locally shut down too.
As I have said before; China needs to send a huge "thank you" note to Trump and Biden, all those calling for China to be cut off from American technology..
Sanctions and a trade war were the kick up the backside China needed to force them to stand on their own two feet, take back control, and regain their sovereignty.
It was incredibly generous of America to surrender her soft-power influence over China to secure that.
They can do 7nm, and probably 5nm with the newer DUV machines TSMC was able to sell them before they were cut off last year. But they won't be able to go any smaller without EUV, which they cannot get, and which would take them many years to develop on their own (even if they had stolen the blueprints from ASML and had a working model to examine) The only other way to go lower would be to find a novel method, or make existing methods that aren't fast enough (like e-beam) or reliable enough (like nanoimprint) to be economically feasible for mass production.
For military applications though where cost is not important they may be able to go smaller using DUV and even more exposures, or using those alternative methods not suitable for mass production. For military applications you need a much smaller number of chips, and if they cost $10,000 each because of lengthy processing and/or extremely low yields they'll pay it. The economics for smartphone SoCs and chips going into weapons systems that cost hundreds of thousands or millions of dollars each are quite different, obviously.
With China it's probably better to avoid saying what they can't do because in 2019 people said they would take a decade or more to hit 7nm and they got there last year. Just two years after hitting 14nm. This year supposedly they will do 5nm.
Tjen people said that quality was low and so they can't produce in volume. Mate 60 etc. Obviously that was totally wrong and the Pura series is set to increase volume over the Mate 60.
Tjat the Pura series uses a variant of last year's chip is no surprise. It's always been that way. Even when using Qualcomm chips.
I think it's important to try and imagine what isn't completely visible right now. The common sense stuff.
Huawei committed to smartphones. Qualcomm reported last year that it would see no further 'material revenue' from Huawei. Obviously they felt they had a viable roadmap. We just don't know what it is right now, but the simple fact that these situations are known, leads one to believe that they have enough in the pipe to be competitive. It is just a question of waiting it out and the chances of that being years away are slim so my guess is we'll see regular advances in all areas of achieving self sufficiency.
... why don't you all apply for a job there?
You'll enjoy a mobile app that spies every minute of your life (location @home or in the office).
This is also a company that had one of the highest suicide rate among its young engineers a few years ago... Try it!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
