EUV Lithography Makes Good Progress, Still Not Ready for Prime Time

[ShintaiDK]

http://www.anandtech.com/show/10097...-good-progress-still-not-ready-for-prime-time

Seems EUV is the ever fading into the distance thing we so desperately need :/

It is now more and more looking like a 2020 tech at best.

 

[Qwertilot]

Wrestling the laws of physics isn't easy

 

[jpiniero]

It's going to be an ugly 4 years in GPU land.

 

[JDG1980]

It's going to be an ugly 4 years in GPU land.

According to the article, EUV won't be needed for 10nm:

Intel, TSMC and Samsung have publicly said that they plan to start producing chips using their 10 nm manufacturing technologies in 2017. These process technologies have already been developed without EUV, and unless the chipmakers decide to introduce new versions that involve EUV for critical layers, EUV will continue to not be used for 10 nm.

The reason why GPUs stagnated on 28nm for so long was that 20nm planar was simply unsuited (apparently, AMD and Nvidia's testing indicated that there would be no meaningful performance improvements on 20nm at all, since the extra leakage would eat up all of the gains). Thus, a GPU process node advancement had to wait for the foundries to transition to FinFET, which is why we aren't getting them until mid-2016. That shouldn't be the case with 10nm, which (as far as I can determine) is just a die-shrunk FinFET process. We could conceivably see 10nm GPUs in, say, 2018.

Getting the foundries onto EUV will be the next technological leap. But I don't think we'll see the kind of 4-to-5-year stagnation we did with 28nm. 2-3 years to get from 16/14->10nm, and then another 3 years (maybe a bit more) to get from 10->7nm with EUV, seems more like it.

 

[StrangerGuy]

I'm gonna chuckle when they gonna say the same thing like 4 years later.

 

[Fizzy Coca-Cola]

Quatum computing is only 15-25 years away, right

 

[videogames101]

It's doable if you guys want to pay $10k a chip for 7nm today.

Sadly the tech has been there for awhile, we just can't do it cheaply enough yet.

 

[jhu]

Will EUV be viable before the end of silicon advancement?

 

[ShintaiDK]

According to the article, EUV won't be needed for 10nm:

No EUV at 10nm just means cost spirals yet more out of control.

Will EUV be viable before the end of silicon advancement?

I am starting to doubt. Because the end of silicon advancement is quite close due to economics by lacking EUV.

 

[ClockHound]

Quatum computing is only 15-25 years away, right

Quantum computing is here now. And then. Or will be. Or was. And will be again. Or not. ;-)

 

[witeken]

I'm reading that ASML apparently can reach 200W, but I don't see that in wafers per day processed numbers and the information from Intel/TSMC.

At the SPIE Advanced Lithography conference, Intel confirmed that there were four running EUV tools with 80 W light sources around the world as of early 2016, up from one a year ago.

ASML also said that it could operate a 200 W light source for one hour with full dose control.

So I guess that explains the "EUV isn't a matter of if anymore, but of when". They "just" need to make sure the availability becomes like 95% instead of 1 hour.

EUV is coming, but it's painstakingly slow and with lots of challenges. In hindsight they should probably have invested more and earlier in the technology. I mean, everyone knew this time (without better lithography) was coming. Without lithography, you can't really do much to advance Moore's Law.

My bet is that EUV will probably be used from the start (by Intel) at 7nm.

7nm is coming in 2019 at the earliest, so that should be long enough.

(I mean, we're hearing about steady progress here, but it suddenly seems like people are doubting EUV will come even at 7nm, while just one year ago, people were still thinking it would/could be inserted at 10nm.)

 

[Ajay]

I'm reading that ASML apparently can reach 200W, but I don't see that in wafers per day processed numbers and the information from Intel/TSMC.




So I guess that explains the "EUV isn't a matter of if anymore, but of when". They "just" need to make sure the availability becomes like 95% instead of 1 hour.

EUV is coming, but it's painstakingly slow and with lots of challenges. In hindsight they should probably have invested more and earlier in the technology. I mean, everyone knew this time (without better lithography) was coming. Without lithography, you can't really do much to advance Moore's Law.

My bet is that EUV will probably be used from the start (by Intel) at 7nm.

7nm is coming in 2019 at the earliest, so that should be long enough.

(I mean, we're hearing about steady progress here, but it suddenly seems like people are doubting EUV will come even at 7nm, while just one year ago, people were still thinking it would/could be inserted at 10nm.)

Seems like semicon companies backed the wrong horse. Of course, they've dumped too much time && money into EUV to back off now. Sadly, we all pay the price.