Feeds

back to article New transistor tech could beat silicon and save Moore's Law

Boffins at MIT's Microsystems Technology Laboratories have developed the world's smallest transistor made of indium gallium arsenide, a substance they say could replace silicon as the go-to material for building tomorrow's ultra-fast, ultra-small microchips. The tiny transistor is just 22nm in length, according to a report by …

COMMENTS

This topic is closed for new posts.

Page:

Paris Hilton

"Should they succeed, however, there's still one hitch to overcome before chips based on the new material go mainstream; namely, that indium gallium arsenide – which is composed of the elements indium, gallium, and arsenic – is currently as much as 10 times as expensive as the equivalent amount of silicon. Those chips may be small, all right. But they'll cost you."

10 times as expensive as sand, eh?

3
2
Silver badge
Facepalm

I'm pretty sure that is not the way it works.

Naturally occurring silicon always exists as silica or silicates, never as a stand alone element. Sand is not silicon nor or 'chips' silicon.

5
1
Silver badge

"Sand is not silicon"

Indeed it is not. Further the silicon is highly purified which I think probably dwarfs the raw material cost.

4
0
Silver badge

Re: "Sand is not silicon"

" the silicon is highly purified which I think probably dwarfs the raw material cost"

Undoubtedly true, but there's more. Even allowing for purification, I'd guess the end to end process costs including design, manufacturing plant, operations etc will all dwarf the purified material costs of either.

3
0
TRT
Silver badge

Re: "Sand is not silicon"

Mmmm... "Indium Gallium Arsenide Valley" just isn't as catchy.

3
0
Facepalm

Re: "Sand is not silicon"

Indium, gallium & arsenide.

Yes, these sound plentiful and cheap to process...10x more expensive you say? Bravo...

0
0

Re: "Sand is not silicon"

Mind you, "InGAs Valley" isn't too bad.

1
0

This post has been deleted by its author

Anonymous Coward

Is that a cost comparison wafer to wafer or gram to gram? If it's the latter, materials really aren't the dominant cost in semiconductor fabrication. If it's the former then I'd imagine the massive economies of scale involved in silicon production and the related technological advancements will be skewing that somewhat.

2
0

May I be the first to say....

... thank you, Jesus!

3
2
Gold badge
Coat

Re: May I be the first to say....

I thought it was St Paul who wrote the Letter to the Arsenides containing the parable of the 22nm wafer?

4
0
Silver badge
Happy

Re: May I be the first to say....

"...for I believe in the Epitaxial growth, and the diffusion of Gallium. For ever and ever, Arsenide."

4
0
Silver badge

Re: May I be the first to say....

For I do not come to destroy Moore's Law, but to fulfil it.

4
0
Silver badge
Meh

Gallium arsenide has been talked up for decades now

But it's always been too expensive in the end. As I recall it is also less environmentally friendly than silicon.

I think that ultimately we'll see lots of it used on chips, but after so many years of vaporware I will believe it when I really see it with a big share in the marketplace.

1
0
Silver badge

Re: Gallium arsenide has been talked up for decades now

Indeed. It's been around so long - I remember it being talked up in the 80s - that if it coud have been productised it would have.

The issue isn't resolution, it's yield and cost.

1
0
. 3

Re: Gallium arsenide has been talked up for decades now

Researchers were building circuits with up to 6 transistors on a gallium arsenide substrate when I was in uni 15 years ago, also using MBE. I think no matter how small they are, drawing the individual transistors on is not feasible when you need billions per device.

0
0
FAIL

Re: Gallium arsenide has been talked up for decades now

Isn't that the crystal that you have to grow in microgravity, using quickly reusable spacecra.. oh, bugger.

0
0

Re: Gallium arsenide has been talked up for decades now

When I was in integrated circuit design 20 years ago we had Gallium Arsenic ASIC chips with up to 30,000 logic gate, made using photolithography not using MBE. To draw on each individual transistor would be unfeasible. The yield rate of the chips was terrible, and that was one reason why they were so damn expensive.

2
0
Bronze badge
Boffin

Re: Gallium arsenide has been talked up for decades now

Indeed, the fact is that up until now silicon has always kept up with its more exotic competition due to taking the lion's share of the development budget and the market, so the economies of scale have mounted up.

If, however, we are now reaching the point at which the material physics advantages of the exotic materials even up the playing field then perhaps there will be a shift away from silicon for some volume uses at least.

That depends on how much of a "brick wall" there is at sub-10nm resolution.

1
0
Boffin

Re: Gallium arsenide has been talked up for decades now

Huh? MBE - Molecular Beam Epitaxy is used to lay down the active layers on the wafer. AlGaAs or InGaAs or some such. The gates are made using either photolithography or electron beam writing. The e-beam writing is for the short gate lengths. It is a serial process, so it is slow for large numbers of FETs. Photolithography is a parallel process, but starts to have problems below 20nm.

0
0
Bronze badge
Coat

Where are you going to get the equipment to run the wafers.

Silicon wafers are generally eight inch in diameter with twelve inch becoming common and sixteen being talked about. It has been awhile since I worked with Gallium Arsinide but I don't recall anyone running any thing bigger than six inch. The new tools move wafers a lot faster than the old tools. I don't think Gallium Arsinide can handle the stress. Currently Gallium Arsinide is run on older, slower tool sets. Then you have die size issues. Your yield is going to be terrible on the smaller wafers.

All in all I can see some of the current Gallium arsinide fabs running specialty chips but I can't see the big production houses doing it.

0
3
Facepalm

Re: Where are you going to get the equipment to run the wafers.

I hear that's exactly what Intel do when they're considering a new production line. Get on ebay and find some old cutting edge chip making gear.

2
0
Bronze badge
Coat

Rice crackers?

Someone was eating them in the office. Smelly, noisy... a real irritation unless they could be put to good use as an alternative substrate for chip manufacture.

Has anyone tried making chips out of them? We ought to be told!

1
0
Anonymous Coward

Re: Rice crackers?

Dude, I fear all those arsenic fumes emanating from the foundry may have sent you the way of Napoleon :(

1
0
Anonymous Coward

Alternatively...

Maybe we can stop the software bloat and develop better mainstream languages/compilers to make good use of the chips we can actually make cheaply?

I mean, WFT is a web browser doing these days needing GB of RAM just to run?! Is it really doing so much more than the browsers _and_ OS of a decade ago that were usable in 64MB and a 200MHz CPU?

10
1
WTF?

Re: Alternatively...

You're right! I mean, 640K is more memory than anyone will ever need!

Bring back the 386DX is what I say!

1
4

Re: Alternatively...

There is huge merit in making the most of what you've got. The megahertz wars were like the UK housing market - people were too dazzled by the embiggening of numbers and superficial gains to see the multitude of sins being hidden by them. I still think there is going to be a big market for 'good-enough', single board, very cheap PCs in the near future.

3
0
Happy

Re: Alternatively...

re: mhz wars...

Humans lost this years ago (we run at around 15hz*, but damn, we still parrallel process *way* better than silicon**. So hz/mhz isn't really that important.

* though it is possible to run an entire Country on as low as 1-2hz...see "George Bush".

** parrallel processing is not number crunching. See "financial crisis 2008" for evidence of that.

5
1
Boffin

Ah, but which wall?

"Researchers believe we are fast approaching a "brick wall," "

Yes, but which one? there's been about 3 now haven't there?

1
1
TRT
Silver badge

Re: Ah, but which wall?

I think they're playing a 1st Person 3D version of a sideways scroller, a'la "Tron".

1
0
Silver badge

We have enough power

We just need to learn to use it properly.

From what I've seen, the vast majority of code is bloated and lazy, with people lifting libraries from other programs and 'tweaking' them.

If you could make a program that had the coding efficiencies of Elite with modern equipment, I don't think we would need to worry about the power of our computers.

Unless you are running the latest 3D games and triple 24" 1600*1200 screens and absolutely won't accept anything less than 60fps then I think we could get by with what we have*.

*I realize this is, of course, blasphemy. I expect to be punished.

8
0
Pint

Re: "*I realize this is, of course, blasphemy. I expect to be punished."

Please hand in your PCs, laptops, tablets and smartphones, and collect the components to build your own Atlas. Once completed, you may use that to access the interweb.

0
0
Silver badge

Re: "*I realize this is, of course, blasphemy. I expect to be punished."

"Please hand in your PCs, laptops, tablets and smartphones"

Hmm, I'm sorely tempted :P

2
0
Bronze badge
Boffin

Re: We have enough power

Whilst this is true the basic advantage of resource waste is that it has allowed cheaper development - i.e. the cost of development is much more than the cost of hardware therefore throwing hardware at the problem was the better choice. So we'd have less development and hence less software to run in general.

Although of course if one is working from the mindset that only the software of peak efficiency should be allowed to survive perhaps that's not a bad thing.

3
0
Silver badge

Re: We have enough power

Kill off the high-frequency-trading scam too.

1
0
Silver badge

Not bad

I did some work experience in an R&D fab developing gallium arsenide components a few years back (well, about 13 to be precise) and the cost of a raw wafer was a LOT more than 10x that of silicon. So we've come a long way already. If it starts to be used, costs will come down.

Interestingly, they told me back then that there was research going on to try to grow GaAs transistors on a silicon wafer. I guess they never got it to work well enough, but it was a very interesting process (involving depositing many layers of different compounds on the silicon wafer to match the crystal structures).

1
0
Anonymous Coward

old style gallium arsenide asics...

had two major drawbacks:

1) They used an nmos style logic gates the speed up was only for n-type transistors.

Holes still propagate at silicon transistor speeds so no point for a CMOS style gate.

Hence very power hungry.

2) Wafers were physically more fragile.

The talk about weather it's more eco-friendly than traditional CMOS is rubbish:

Standard CMOS processes use some very nasty/dangerous chemicals and compounds already.

Arsine anyone?

2
1

Molecular beam epitaxy

For a second there I thought we'd accidentally got a bit of a Star Trek: Voyager script pasted in.

Can I have an upvote for Seven please?

1
1

Re: Molecular beam epitaxy

Nice one

0
0
Silver badge
Devil

Re: Molecular beam epitaxy

Indium Gallium Arsenide-ey implants?

0
0
Meh

gallium arsenide used to be used in IBM mainframe processors

IBM took a hit and for a while HDS made the fastest S/390 mainframes. The rational then was that more CMOS cores packed close together would provide higher performance at lower lost with greater MTBF.

gallium arsenide might be faster for high-frequency-trading, but unless they address the heat problem, you won't be able to co-locate in exchanges.

0
1

This post has been deleted by its author

Molecular beam epitaxy

the MBE bit is the REALLY expensive step here - the wafers are so pricey it's just laughable.

Really fast transistors have been made with similar methods for a while (though production people avoid MBE because of the cost and use faster processes like MOCVD by preference). You can find one of those in your sky dish.

Nice that they can make small ones now but hey, I wouldn't bank on the silicon era ending just yet.

When I was in semiconductor physics at university ~15 years ago it was like that and it still is.

1
0
WTF?

NPN transistor?

NPN devices don't have a Gate, Drain and Source. Could have at least used a FET symbol, or described the transistor as having a Base, Collector and Emitter.

Just saying...

4
0

Ahh Gallium Arsenide

All the original work on Gallium Arsenide was done at UWIST (now part of Cardiff Uni) working with INMOS before they were flogged off to the French. I was an Electronics undergrade at the time and most of my professors where working on it, describing how it was going to replace silicon. They were just years ahead of their time.

0
0
Coat

Ahh Gallium Arsenide #2

My old Pentax ME Super had a Gallium Arsenide Phosphide Photo Diode (GAPD).

Mine's the one with the Zorki 4k in the pocket!

3
0
Silver badge

How much smaller can these things get!

Back of an envelope calculations give them only 3906 atoms to work with at 10nm.

1
0
Gold badge
Boffin

The 80's joke was "GaAs, the material of the future"

"Always was and always will be."

Back then DARPA was running the "Very High Speed IC" Programme and targeting 32bit MIPS (for the low transistor count) processors at the unbelievable speed of 200Mhz.

GaAs was tough to mfg (IIRC the process was originally developed out of RSRE Malvern, now a bit of Quietq), fragile, expensive and dangerous (Arsenic vapourises readily at the operating temperature). As noted the NMOS process was power hungry. IIRC "holes" which are important in PMOS, that make up the other half of "complimentary", move slower in in GaAs than Silicon, something perhaps someone should have picked up on sooner.

Oh and the oxide layer is a pig. But hopefully InGaAs will be much better.

This is V 0.1 technology. No one will push this unless their backs are really against the wall.

0
0
Boffin

Re: The 80's joke was "GaAs, the material of the future"

No one is seriously considering using GaAs or InGaAs for processors. SiGe (Silicon Germanium) is the current best hope for very high speed digital devices. It is not too expensive as it is grown on top of standard Si wafers. It has better speed than Si and works pretty well for mixed signal applications.

0
0
Unhappy

Editor: Transistor symbol is wrong....

The Register main page shows the symbol for a bipolar transistor.....should be for a field effect transistor as here >> http://www.google.com/imgres?imgurl=http://www.interfacebus.com/JFET-N-Channel-symbol.png&imgrefurl=http://www.interfacebus.com/Glossary-of-Terms_J.html&h=106&w=103&sz=2&tbnid=u4OWZwlZ2VT86M:&tbnh=84&tbnw=82&prev=/search%3Fq%3Dfet%2Btransistor%2Bsymbol%26tbm%3Disch%26tbo%3Du&zoom=1&q=fet+transistor+symbol&usg=__ayrb0Oa0WlWlp6II6Nv9pVM_lMM=&docid=7vUgV1eI3SBFCM&hl=en&sa=X&ei=2GXHUJ3ENIvC9gTCi4CIAQ&sqi=2&ved=0CE8Q9QEwBQ

0
0

Page:

This topic is closed for new posts.