IBM Claims 5nm Nanosheet Breakthrough | Hacker News Reader

IBM Claims 5nm Nanosheet Breakthrough

eetimes.com

75 pointskungfudoi21 comments

Related discussion from yesterday: https://news.ycombinator.com/item?id=14485960

Queue extreme UV lithography. Where the light source costs $100M, took 8,000 engineers 20 years to develop, and consumes 1MW of power to shine a 1kW beam.

https://www.google.ca/amp/spectrum.ieee.org/semiconductors/d...

From the Ars article a couple of days ago [1], it sounds like EUV improvements are actually at the core of the work announced here.

[1] https://arstechnica.co.uk/gadgets/2017/06/ibm-5nm-chip/

Just FYI:

https://en.oxforddictionaries.com/definition/cue

https://en.oxforddictionaries.com/definition/queue

Same pronunciation, different meaning.

At this point isn't this basically a fab attached to a particle accelerator?

This uses tin plasma generated by laser pulses. No particle accelerators involved. One could build an EUV source based on a free electron laser too, which involves a particle accelerator, but I assume that has some limitations that make it less suitable for lithography.

Have you heard of Rayton Solar? They are supposedly slicing nanometer thin silicon with a particle accelerator. I don't know how hard it'd be to extend this to lithography though...

Bill Nye the Science Guy is their PR person: https://www.youtube.com/watch?v=JA8fKlfcA7E

Progress has to begin somewhere. In another 10 years, this kind of lithography will likely have evolved into something much more viable.

The problem is that EUV was originally expected in 2007. It is already more than a decade late, and it still has issues.

If it is a short burst, perhaps 1MW is not a big deal?

You can use a bank of capacitors and the power itself is not a huge deal (still not trivial, but it's been done for 60+ years).

The problem is also that at 1kW for every 1Mw you still only use 0.1% of the energy you pumped in. And the energy requirements are still large because you are firing this thing repeatedly. Assuming the same amount of energy is required to ablate the atoms as with conventional UV (I'm not in the field, but its probably not a bad assumption), you have to multiply the ratio of the efficiencies of both methods to get at the energy required. Could be several multiples more if not an order of magnitude.

Also where does the remaining 999kW go? Eventually heat, so you have to spend 2Mw more to cool it (assuming 50% AC efficiency, I guess you could pump cold water instead). Add in the energy costs of opening and closing a vacuum lock and this is a fun project to be working on :D

Why is IBM still making all these breakthroughs?

I thought they got out of the chip business.

They left the semiconductor manufacturing business. That's a different segment. The lithographic tools and the technology used to make those tools are both different market segments that IBM is involved in.

They haven't gotten out of research in the Area at all, their labs are still among the leading ones in many different subareas (the director of the Zurich Lab gave a talk at my University just recently). They also continue to develop both their Power 9 and Mainframe processors.

This is a few years old now, but sharing to illustrate that they've done some really cool stuff in that space: http://www-03.ibm.com/press/us/en/pressrelease/38653.wss.

Not even close. They still help design, test, and fabricate a huge number of high performance ASICs.

patent licensing?

Getting close to quantum mechanics problems... https://youtu.be/rtI5wRyHpTg

Actually most nodes are phasing out EUV lithography research and moving to plasma-enhanced lithography. Pretty sure you can't really go below 15nm without PECVD/Plasma enhanced etch steps.

Reply on news.ycombinator.com