You Didn’t Build your PC… This Did.

[Sveeno]

Building a computer is difficult and building a CPU is even tougher.. But building the machines that are used to make a CPU? It takes literally the most bonkers engineering and physics on the planet, and ASML were kind enough to show us how it all works.

 

 

[filpo]

1971 POV: You built a machine to build something that doesn't even exist

[Bob__]

This is dope

[Crunchy Dragon]

This is awesome, but I'm pretty sure this super cool machine did not assemble my PC parts into a working PC.

 

/s for semantics

[manikyath]

"Truly annihilate some tin"

 

i need that on a shirt..

[Jerakl]

8 minutes ago, manikyath said:

"Truly annihilate some tin"

 

i need that on a shirt..

I would buy a shirt with that quote.

Good vid.

[ultimatebob]

Where is this fab located? Is that a secret?

[XNOR]

I enjoyed Alex coughing up the Intel logo when talking about being stuck on a node.

[Vishera]

It's one of the most advanced machines humanity has ever made, very few people know how to make something like that.

[WolframaticAlpha]

2 hours ago, filpo said:

1971 POV: You built a machine to build something that doesn't even exist

The first lithography machines were literally just lasers aimed through a microscope

[XNOR]

10 minutes ago, Vishera said:

It's one of the most advanced machines humanity has ever made, very few people know how to make something like that.

Which makes it even more amazing you can just... buy these cutting edge node chips for very attainable amounts of money. Black magic fuckery on the very edge of technology, factory sized clean rooms, mind boggling investments of (quite literally) unimaginable amounts of money, and you can just go on Amazon and order one with a couple of clicks so you can look at cat pictures faster.

 

It amuses me no end that desktop processors are just about the fastest a CPU gets. While there are plenty of chips with more raw processing power in the enterprise world and elsewhere, we put some of our fastest, most cutting edge chips in computers to make games go brrr.

 

29 minutes ago, ultimatebob said:

Where is this fab located? Is that a secret?

In the video they said it's the San Diego branch of ASML.

[williamcll]

42 minutes ago, XNOR said:

Which makes it even more amazing you can just... buy

See if this was the case you would see them everywhere especially in developing countries.

[XNOR]

1 minute ago, williamcll said:

See if this was the case you would see them everywhere especially in developing countries.

Are you talking about desktop chips or EUV machines? I feel you may not have read my comment in its entirety.

[williamcll]

2 hours ago, XNOR said:

Are you talking about desktop chips or EUV machines? I feel you may not have read my comment in its entirety.

Both actually. EUV manufacturing remains a monopoly and desktop chips affordability has made little improvement with rising prices. 

[idmb]

New favourite LTT video. 

[dogwitch]

very neat.

had the funniest bad take on how difficult it is. on  a reddit comment.

 

they thought billions are difficult ... hardest thing in human history.

 

another field has to deal with An estimated 10 nonillion (10 to the 31st power) difficult.  that on top of another power factor...

but i do commend the people to be general smarter on ltt ... then  average reddit user!

i love the nerd video on science.

 

[GodAtum]

I just want to know why they blurred out the tip?

[Kid.Lazer]

58 minutes ago, GodAtum said:

I just want to know why they blurred out the tip?

To make it legal in Japan.

[Vishera]

1 hour ago, GodAtum said:

I just want to know why they blurred out the tip?

Trade secret.

[XNOR]

On 5/26/2023 at 2:32 AM, williamcll said:

Both actually. EUV manufacturing remains a monopoly and desktop chips affordability has made little improvement with rising prices. 

EUV manufacturing will always remain a monopoly, unless someone else figures it out independently. The machines they're building now are literally the result of decades of research and many, many billions invested. Not only from the point of the business would it be a terrible idea to share the technology, even from a geopolitical point of view there are major interests to keep the technology under wraps.

 

I don't really agree with desktop chip affordability making no progress. Compared to 10 years ago, you can get insane amounts of computing power for little money. This might still represent a lot of money for people in developing countries, but the performance you get per buck is immense.

[ImorallySourcedElectrons]

On 5/25/2023 at 11:28 PM, WolframaticAlpha said:

The first lithography machines were literally just lasers aimed through a microscope

We didn't have lasers at the time, I-line mercury lamps were the primary source of intense UV light, and those are in fact still used on cheaper production equipment. 

But some of the first lithography machines were contact aligners, and those only used microscopes for alignment of the mask versus the wafer. You bring the mask into direct contact with the photoresist, ensuring that you get a pretty good exposure. Of course, this requires a 1:1 mask (not so interesting at today's 15-20 nm feature sizes) and runs the risk of damaging the mask with each exposure. On the other hand, contact aligners do have a few advantages, like being dirt cheap and being able to take pretty much anything as a mask. We ran the lab one using transparencies that came out of a laser printer at one point. But then you have further development of lithographic systems, which coincided somewhat with the development of modern colour photography and general printing technology, whole mess that no one really knows the story behind I think.  

 

On 5/27/2023 at 7:23 PM, XNOR said:

EUV manufacturing will always remain a monopoly, unless someone else figures it out independently. The machines they're building now are literally the result of decades of research and many, many billions invested. Not only from the point of the business would it be a terrible idea to share the technology, even from a geopolitical point of view there are major interests to keep the technology under wraps.

Sharing the technology would also not do too much, you could have the schematics and detail drawings of this machine today, and it'd still take you decades to figure out how to make it without the people who actually run the production helping you directly. Just grinding those mirrors to the right specifications is a an achievement that's hard to grasp.

 

And the economic benefit of these machines is also quite often misunderstood, I'm not sure if the market demand would even be there to really carry a second manufacturer, because it's unlikely the price of the machines would come down that much if there were a second source. While I'm not privy to the details, I would imagine a lot of the cost goes towards maintenance, spare parts, production supplies, etc. 

 

And there are a few specific process steps for which EUV machines have major advantages, for most other things though major fabs are still running their old steppers, even with the reduced yield they often still come out cheaper. The economics of chip manufacturing and which equipment you use can get quite screwy, since a lot of it depends on the volume you can run, and EUV machines tend to have a major disadvantage there. If you run a thousand wafers an hour on the EUV machine with a yield of 99%, or you run 20000 through your older steppers with a yield of 95%, the latter might be better for your bottom line.

[XNOR]

On 5/29/2023 at 11:48 PM, ImorallySourcedElectrons said:

And the economic benefit of these machines is also quite often misunderstood, I'm not sure if the market demand would even be there to really carry a second manufacturer, because it's unlikely the price of the machines would come down that much if there were a second source.

The point wouldn't necessarily be to have a competing EUV machine manufacturer in the regular free market kind of sense. China can't buy these machines and can't make them, so it's dependent on foreign manufacturing if it isn't to fall behind in terms of high end computing. Having independent domestic access to EUV machines would free China from a couple of political and technological dependencies, allowing access to high end computing without bending the knee.

 

Of course, China could also use any independent access to undercut the existing market with the help of subsidies. If this were to be done successfully enough for ASML to go under, it would mean the world is now dependent on Chinese EUV manufacturing, giving China leverage.

 

Any sufficiently advanced technology is indistinguishable from magic politics.

[ImorallySourcedElectrons]

5 hours ago, XNOR said:

The point wouldn't necessarily be to have a competing EUV machine manufacturer in the regular free market kind of sense. China can't buy these machines and can't make them, so it's dependent on foreign manufacturing if it isn't to fall behind in terms of high end computing. Having independent domestic access to EUV machines would free China from a couple of political and technological dependencies, allowing access to high end computing without bending the knee.

 

Of course, China could also use any independent access to undercut the existing market with the help of subsidies. If this were to be done successfully enough for ASML to go under, it would mean the world is now dependent on Chinese EUV manufacturing, giving China leverage.

 

Any sufficiently advanced technology is indistinguishable from magic politics.

China already spent billions trying to build EUV equipment, with little to no success. And I think we can all assume they have full access to the engineering files from ASML through industrial espionage. There are many reasons they're falling behind, but I think a big one is the lacking educational system in China. Engineering and science in general is a creative activity, and should be taught as such. The cramming style education you frequently see in many Asian countries isn't very conducive to producing individuals with the scientific and technical know-how and that touch of creativity to side-step the issues they might run into. Repeatedly running into the same physics wall isn't very productive, but that's pretty much what they've been doing when tackling electronics manufacturing processes and equipment.