9 ms·
As he says here, the decision not to take risk and jump to EUV with TSMC and Samsung was the mistake that started this downfall and it snowballed. The decision
by Nokinside 6y ago
As he says here, the decision not to take risk and jump to EUV with TSMC and Samsung was the mistake that started this downfall and it snowballed. The decision happened somewhere in the pathfinding stage into 10nm.
When Gerald Marcyk was running Intel components research back in the
2001 Intel was already considering using EUV lithography for the 45nm technology node that would come out 2007. It turned out that the switch to EUV could be delayed and delayed again. Intel thought they could get away with it this last time and it was a wrong decision.
EUV is super hard. There is only one company that can make EUV lithography machines, Dutch company called ASML. TWINSCAN NXE:3400B costs ~$120 million
https://www.asml.com/en/products/euv-lithography-systems/twinscan-nxe3400b https://www.asml.com/en/products/euv-lithography-systems/twi...
Intel, Samsung and TSMC invested billions into ASML in 2012. In exchange Intel got 15% ownership, Samsung 10% and TSML 5% if I remember correctly.
They don't have any significant ownership anymore.
- baybal2 6y agoWell, not to forget the 157i debacle. Intel first started cheaping out as far back as 2003-2005. 157i could've been almost as good as EUV, at much lower cost, but Intel fully consciously decided to forego it because "nobody can beat us on 193i yet," and of conern that 157i consortium will give superior litho to potential competitors.
- rowanG077 6y agoWhat is 157i? Googling with various helper terms doesn't turn anything up.
- trishmapow2 6y agoAs someone with no background knowledge, I first looked up "193i" intel which lead me to the term 193nm immersion lithography. After that doing the same for 157nm was trivial.
- baybal2 6y ago157nm lithography process using fluorine excimer laser light source.
- rakah 6y agoEUV is hard, but isn't part of the reason why only ASML can make those machines because of the CRADA they have with DOE as a member of the EUV LLC? I believe this agreement was why ASML wasn't allowed to sell their solution to China recently. I bet Intel will have no trouble buying EUV to bring in-house - especially if that serves US strategic interests.
- Taniwha 6y agoASML only makes a few machines every year, you can't just buy them off the shelf IMHO banning sales of ASML's product to China is chest thumping protectionist nonsense - if you're worried about the rise of China's military then you should be tangling their economy up more with that of the West so that they have more reasons not to go to war rather than driving them to be self sufficient.
- totalZero 6y agoI don't think they're worried about China's military so much as China's independence from the rest of the semiconductor supply chain. When you give China a machine, it's only a matter of time before they reverse engineer it and make something slightly worse but much cheaper. This kind of theft is backed by the government. So the expectation that they will remain forever dependent upon ASML is wishful thinking.
- Taniwha 6y agoBut surely that should be ASML's choice?
- someperson 6y agoASML machines include a large amount of technology from the US, and so can be controlled under US export control laws. A country is free to develop all those technologies locally and not face US export control, but that's such a difficult task that nobody has achieved it.
- Animats 6y agoA lot of people were hoping someone would come up with a better way to make "extreme ultraviolet" (soft x-rays, really) light than the horrible kludge used now.[1] Syncrotrons were tried. (Whatever happened to Lyncean?[2]) What's used now looks like a physics experiment, not a production process. Drops of tin are hit with lasers to emit X-rays, which are then focused with mirrors on the mask and wafer. This takes a setup the size of a house that produces very little power at the output end. Much of the industry was hoping something better would come along. But so far, it hasn't. [1] https://strobe.colorado.edu/wp-content/uploads/STROBE_ASML-EUV-Sources_Purvis_25-Sept-2020-1.pdf https://strobe.colorado.edu/wp-content/uploads/STROBE_ASML-E... [2] https://lynceantech.com/ https://lynceantech.com/
- keanebean86 6y agoAn option would be to replace the laser and tin with some tape. [1] I know tape is a consumable item but is considerably less professional compared to a laser. [1] https://www.nature.com/news/2008/081022/full/news.2008.1185.html https://www.nature.com/news/2008/081022/full/news.2008.1185....
- godelski 6y agoA laser is not just light. It is a specific wavelength. X-Rays are a spectrum. The EUV/soft X-ray is a part of that spectrum. There's also other properties that you want from the laser. But this tape-x-rays won't be easy to control or uniform in wavelength. Making it not a laser, even if you could collimate it. https://en.wikipedia.org/wiki/Laser https://en.wikipedia.org/wiki/Laser
- keanebean86 6y agoThanks for this info. I assumed there was a specific reason they used this setup. I was joking about using tape. Too bad we can't make x-ray leds.
- Animats 6y agoIt's easier to generate hard X-rays than EUV. An X-ray tube is just an electron gun and a copper target.
- dannyw 6y agoIntel invested lots and lots in novel EUV alternatives and they didn't work out. They took risks, they spent big on R&D, and I think you can certainly blame mismanagement but it's wrong to imply Intel wasn't taking risks.
- JohnJamesRambo 6y agoI would have been more wowed by the $120 million if I didn’t know B2 bombers cost $2 billion each. I wonder how much more safe and secure we would be with lots of Twinscans 3400B’s and a few less B2’s?
- mercer 6y agoI was so curious to see if the website would have a shopping cart, and what would happen if I added one or two EUV lithography machines to it...