It's definitely an "on the shoulders of giants standing on the shoulders of giants" thing. Insane breakthrough technologies on top of other insane breakthroughs. Firing lasers at microscopic molten drops of metal in a controlled enough manner to get massively consistent results like what??
It’s a mind blowing achievement, nothing below sorcery if you think about it.
ASML machines are hitting tin droplets with 25kW laser 50,000 times a second to turn them into plasma to create the necessary extreme ultraviolet light, and despite generating 500W of EUV, only a small fraction can reach the wafer, due to loses along the way. I believe it was like 10%.
One thing I am curious about - how many generations of process shrink is one of these machines good for? They talk about regular EUV and then High-NA EUV for finer processes, but presumably each machine works for multiple generations of process shrink? If so, what needs to be adjusted to move to a finer generation of lithography and how is it done? Does ASML come in and upgrade the machine for the next process generation, or does it come out of the box already able to deliver to a resolution a few steps beyond the current state of the art?
If you’re interested in this stuff Asianometry has lots of great videos. They’re not all on semiconductors, but he’s done a lot on this history, developments, and what’s going on in that world.
Maybe the high water usage is at some other stage? Or intermediate preceding stages? I'd love to understand more end-to-end, as surely it isn't as easy as popping a wafer in a semi-truck trailer sized lithography machine.
Check out the Branch Education channel, they have a series of videos that explain how the underlying transistors are made in 3d space with multilayer exposures etc.
One thing to understand is that you’re seeing an accumulation of over 50 years of incredible engineering and cutting edge science, these things were invented incrementally.
Lithography is one of many steps, but probably the most important one. You use it to expose a photoresist to create a mask for further processing. After exposing the photoresist you need to develop it, remove either the exposed or unexposed photoresist. The remaining photoresist then is the mask and you either etch or dope the surface that is not covered by the mask or you deposit material on top. And then you need to remove the mask and start all over again for the next layer. The high water usage comes from repeatedly needing to clean the surface to remove chemicals and photoresist.
I think this clockwork-in-a-vacuum was preceded by eidophors: a projector with a spinning disc of oil that has an image drawn by an electron gun, that is then illuminated by an arc lamp.
The interesting part is the E-Core on A19 Pro are nearly as good as the previous ARM Big Core while only using half the power. I would love to know the die size of the cache and E-Core.
ARM were catching up to Apple in terms of big core, now Apple has leapfrogged in E-Core again. But competition is good. ARM should have some announcement coming in next few months.
Fascinating, what surprises me is that it looks more "fractal" like than other chip dies I've looked at. Perhaps all the newest ones do, however it makes me wonder if part of Apples secret sauce is more sophisticated design tooling.
There is no backside anything here. This may be a photo of a thinned die; silicon is somewhat transparent so you can often see the die structures better from the back because it isn't blocked by the metal stack.
Anand was involved in a scandal recently there [0]. Is he still in that job?
The driving force behind the reviews for mobiles and related topics at AT was Andrei Frumusanu. His reviews had a level of depth very few even on AT could touch. But he left to work for Qualcomm so that ended his reviewer stint.
The “scandal” was unsubstantiated assertions in a Nuvia legal filing. Basically Apple accused Nuvia of poaching employees. In responding trying to show they were good guys, Nuvia said Anand sent them powerpoint slides marked confidential, but they responded that the communication was inappropriate. The case ultimately went to nowhere: https://www.theregister.com/2023/05/01/apple_nuvia_lawsuit/.
This doesn’t mean anything other than some lawyer thought it would be a good optics. I bet Apple’s default powerpoint slide template has confidential headers. And Nuvia would be right to cover its ass by responding the way it did—they don’t want anything marked Apple confidential in their possession even if the actual content of the slide is public information. Inclusion of the correspondence in the Nuvia legal filing could even have been a prophylactic measure, intended to get out in front of the evidence before Apple seized on it to show Nuvia did anything wrong.
Just don’t believe stuff in legal filings. It’s not that they’re untrue, it’s that they’re definitionally self-serving and selected to paint the other side in the worst light possible. That’s a byproduct of our adversarial system.
They might be referring to Substack having a very open policy of having writers of any political (and other) affiliation on the platform, but I don't necessarily see it as a drawback to be so neutral.
Oh, you're so close. The fascists are parasites: they only tolerate liberalism insofar as it allows them to spread their ideas. They desperately want to--and will--silence their opponents as soon as they wield power. They view tolerance as a weakness, and the social contract as a game to be played only until it can be rewritten.
I do not dispute that Nazis can, legally, say what they say. I don't even necessarily think it's bad that their speech is protected. But their speech should have consequences, and platform owners can and should tell them to fuck off.
> But their speech should have consequences, and platform owners can and should tell them to fuck off.
Sure, some platforms do, and some don't, like Substack, that is their prerogative and if people don't like it they can choose not to read them. But we cannot force them to do anything, nor should we.
This. Sorry you're getting down votes at this moment for speaking the truth, but there really is no excuse for anti society anti intellectual illiberalism from the world's literal biggest losers.
Might be the allegations such as those made by David Farrier this August[1]
> Then, last month, Substack sent out a push notification to some users of the Substack app encouraging them to subscribe to a Nazi newsletter. Like, full-Nazi stuff. Not subtle:
followed by a screenshot of a Substack profile full of swastikas describing itself as news for the "National Socialist and White Nationalist Community"
It’s still my strongly held belief that things like this are one of humanity’s grandest achievements.
It's definitely an "on the shoulders of giants standing on the shoulders of giants" thing. Insane breakthrough technologies on top of other insane breakthroughs. Firing lasers at microscopic molten drops of metal in a controlled enough manner to get massively consistent results like what??
It’s a mind blowing achievement, nothing below sorcery if you think about it.
ASML machines are hitting tin droplets with 25kW laser 50,000 times a second to turn them into plasma to create the necessary extreme ultraviolet light, and despite generating 500W of EUV, only a small fraction can reach the wafer, due to loses along the way. I believe it was like 10%.
Here’s an incredible, very detailed video about it: https://youtu.be/B2482h_TNwg
That is a very high quality video.
One thing I am curious about - how many generations of process shrink is one of these machines good for? They talk about regular EUV and then High-NA EUV for finer processes, but presumably each machine works for multiple generations of process shrink? If so, what needs to be adjusted to move to a finer generation of lithography and how is it done? Does ASML come in and upgrade the machine for the next process generation, or does it come out of the box already able to deliver to a resolution a few steps beyond the current state of the art?
If you’re interested in this stuff Asianometry has lots of great videos. They’re not all on semiconductors, but he’s done a lot on this history, developments, and what’s going on in that world.
https://www.youtube.com/c/Asianometry/videos
I've learned so much from this channel. One of the best out there.
Even the video about zippers was fascinating.
https://www.youtube.com/watch?v=9d6eNmtHFQk
That is a really cool video, thank you!
Maybe the high water usage is at some other stage? Or intermediate preceding stages? I'd love to understand more end-to-end, as surely it isn't as easy as popping a wafer in a semi-truck trailer sized lithography machine.
Check out the Branch Education channel, they have a series of videos that explain how the underlying transistors are made in 3d space with multilayer exposures etc.
One thing to understand is that you’re seeing an accumulation of over 50 years of incredible engineering and cutting edge science, these things were invented incrementally.
Lithography is one of many steps, but probably the most important one. You use it to expose a photoresist to create a mask for further processing. After exposing the photoresist you need to develop it, remove either the exposed or unexposed photoresist. The remaining photoresist then is the mask and you either etch or dope the surface that is not covered by the mask or you deposit material on top. And then you need to remove the mask and start all over again for the next layer. The high water usage comes from repeatedly needing to clean the surface to remove chemicals and photoresist.
I've been trying to find that video!! Our professor showed it in class but I was half-asleep and I wanted to rewatch it so badly.
Thanks for this! This must be the best video on EUV lithography that has ever been made.
I think this clockwork-in-a-vacuum was preceded by eidophors: a projector with a spinning disc of oil that has an image drawn by an electron gun, that is then illuminated by an arc lamp.
https://www.youtube.com/watch?v=5783jPTKzjk
Nah, NVIDIA GPU has way more raw computational power & smarts.
See how fast it is running on 8 watts.
GPUs are far more regular in structure, lots of "copy-pasted" blocks, because they are a collection of many relatively simple processors.
Anyone interested in more info and detailed videos should check out Geekerwan: https://youtu.be/Y9SwluJ9qPI
English subtitles are recommended, unless you are better at Chinese than I am.
The interesting part is the E-Core on A19 Pro are nearly as good as the previous ARM Big Core while only using half the power. I would love to know the die size of the cache and E-Core.
ARM were catching up to Apple in terms of big core, now Apple has leapfrogged in E-Core again. But competition is good. ARM should have some announcement coming in next few months.
Is there some magic with AI that lets you watch those videos dubbed in English?
There's a YouTube feature but the creator needs to enable it per-video.
Is there a link to a version that isn't almost entirely compression artifacts?
This definitely isn't the original, since the blue text at the bottom right isn't even legible.
Seems like the full resolution images are for sale ("contact us").
+1, it's pretty fuzzy. Maybe chipsandcheese or hotchips will do an independent imaging?
The A19 appears to be remarkably intricate chip.
Yeah if you scroll down they tell you how
I scrolled down, and there is only 1 paragraph of text and two blurry images. What did you scroll down to?
Next to the paragraph of text is this:
"High Resolution Floorplan images available here"
With some contact info below that
It is moved to the end of the page on mobile it seems
> High Resolution Floorplan images available here
> Call Us
> Write to us
Didn't think I'd be confronted with a covert Factorio map today. I just deleted the damn game to try and regain control over my life!
Fascinating, what surprises me is that it looks more "fractal" like than other chip dies I've looked at. Perhaps all the newest ones do, however it makes me wonder if part of Apples secret sauce is more sophisticated design tooling.
Very organic looking, like they passed it through an optimization algorithm.
Looks like an aerial shot of farmland to me
Do TSMC current node sizes allow for Backside Power Delivery (BPD)?
As someone who knows nothing about PCB, from those images it appears that double side printing of some sort is happening.
Please correct me if I'm wrong.
No backside power delivery on TSMC yet, even at 2nm. It's supposed to come on TSMC's A16 node.
I think what you're seeing is the silicon layers visible from the back through the bulk substrate, and the metal layers on the front.
There is no backside anything here. This may be a photo of a thinned die; silicon is somewhat transparent so you can often see the die structures better from the back because it isn't blocked by the metal stack.
It will be a shame when backside power delivery makes that not work as well.
Does anyone know what the block dot at 12 o'clock on the image is? https://chipwise.tech/wp-content/uploads/2025/09/A19_SoC_die...
Well I see the 'die shot,' but not the 'analysis'
Unfortunately I think think you have to pay for the analysis and high resolution pics.
It’s tragic that those die shots are so tiny, you can’t make out anything.
Any idea of transistor count?
Are there also already analyses of these die shots, explaining what be concluded from them?
https://xcancel.com/highyieldYT/status/1970533400151818256
Ken Shirriff it's your time
I'm afraid that chip is too complicated for me. I'm still trying to figure out the 386 :-)
But I did colorize the A19 die photo with Apple's M1 rainbow gradient, just for fun: https://oldbytes.space/@kenshirriff/115256179526128051
If the Apple A19 could be made in the 386's process (1000-1500nm), it would be about 300-500x bigger than its current size in each dimension.
"too complicated", what an understatement! Even the 386 wasn't exactly simple. The complexity of these modern chips is mind-bendingly immense.
The A19 is a SoC so it's doing a lot more than a 386 would have.
Haha no sweat I'm be waiting for your 386 series
The only thing more impressive than this technology to me is possibly the machines that manufacture these chips.
I think I see a bug.
A bug as in a MEMS microphone?
Missing Anandtech.com right about now
Just in case you didn’t know, dude works at Apple now!
(For about ten years now..)
Anand was involved in a scandal recently there [0]. Is he still in that job?
The driving force behind the reviews for mobiles and related topics at AT was Andrei Frumusanu. His reviews had a level of depth very few even on AT could touch. But he left to work for Qualcomm so that ended his reviewer stint.
[0] https://www.tweaktown.com/news/70653/anandtech-founder-sent-...
He’s still at Apple as of last year. https://forums.anandtech.com/threads/an-anand-has-been-sight...
The “scandal” was unsubstantiated assertions in a Nuvia legal filing. Basically Apple accused Nuvia of poaching employees. In responding trying to show they were good guys, Nuvia said Anand sent them powerpoint slides marked confidential, but they responded that the communication was inappropriate. The case ultimately went to nowhere: https://www.theregister.com/2023/05/01/apple_nuvia_lawsuit/.
This doesn’t mean anything other than some lawyer thought it would be a good optics. I bet Apple’s default powerpoint slide template has confidential headers. And Nuvia would be right to cover its ass by responding the way it did—they don’t want anything marked Apple confidential in their possession even if the actual content of the slide is public information. Inclusion of the correspondence in the Nuvia legal filing could even have been a prophylactic measure, intended to get out in front of the evidence before Apple seized on it to show Nuvia did anything wrong.
Just don’t believe stuff in legal filings. It’s not that they’re untrue, it’s that they’re definitionally self-serving and selected to paint the other side in the worst light possible. That’s a byproduct of our adversarial system.
I wonder if Anandtech would have lasted longer if that hadn't happened.
Probably not. The market moved on. Giving such involved in depth articles for free was no sustainable compared to clickbait random articles.
/me pours one out
Chips & Cheese is a good current alternative.
https://chipsandcheese.com/
[flagged]
We detached this comment from https://news.ycombinator.com/item?id=45354959 and marked it off topic.
Eschew flamebait. Avoid generic tangents.
Please don't use Hacker News for political or ideological battle. It tramples curiosity.
https://news.ycombinator.com/newsguidelines.html
What’s wrong with Substack?
They might be referring to Substack having a very open policy of having writers of any political (and other) affiliation on the platform, but I don't necessarily see it as a drawback to be so neutral.
[flagged]
If you want free speech, you'll have to have people who have views you don't like, otherwise they might silence you too.
That would include people with views about Substack too, though, right?
Of course, why wouldn't it?
Not views that espouse violence.
Oh, you're so close. The fascists are parasites: they only tolerate liberalism insofar as it allows them to spread their ideas. They desperately want to--and will--silence their opponents as soon as they wield power. They view tolerance as a weakness, and the social contract as a game to be played only until it can be rewritten.
I do not dispute that Nazis can, legally, say what they say. I don't even necessarily think it's bad that their speech is protected. But their speech should have consequences, and platform owners can and should tell them to fuck off.
> But their speech should have consequences, and platform owners can and should tell them to fuck off.
Sure, some platforms do, and some don't, like Substack, that is their prerogative and if people don't like it they can choose not to read them. But we cannot force them to do anything, nor should we.
[flagged]
This. Sorry you're getting down votes at this moment for speaking the truth, but there really is no excuse for anti society anti intellectual illiberalism from the world's literal biggest losers.
[flagged]
Might be the allegations such as those made by David Farrier this August[1]
> Then, last month, Substack sent out a push notification to some users of the Substack app encouraging them to subscribe to a Nazi newsletter. Like, full-Nazi stuff. Not subtle:
followed by a screenshot of a Substack profile full of swastikas describing itself as news for the "National Socialist and White Nationalist Community"
[1] https://www.webworm.co/webwormisleavingsubstack/
Agreed! Check out Beehiv or Ghost!