Graphene Won't Replace Silicon In CPUs, Says IBM 81
arcticstoat writes "IBM has revealed that graphene can't fully replace silicon inside CPUs, as a graphene transistor can't actually be completely switched off. In an interview, Yu-Ming Lin from IBM Research (Nanometer Scale Science and Technology) explained that 'graphene as it is will not replace the role of silicon in the digital computing regime.' Last year, IBM demonstrated a graphene transistor running at 100GHz, while researchers at UCLA produced a graphene transistor with a cut-off frequency of 300GHz, prompting predictions of silicon marching towards its demise, making way for a graphene-based future with 1THz CPUs. However, Lin says, 'there is an important distinction between the graphene transistors that we demonstrated and the transistors used in a CPU. Unlike silicon, graphene does not have an energy gap, and therefore, graphene cannot be "switched off," resulting in a small on/off ratio.' That said, Lin also pointed out that graphene 'may complement silicon in the form of a hybrid circuit to enrich the functionality of computer chips.' He gives the example of RF circuits, which aren't dependent on a large on/off ratio."
Congratulations (Score:3)
Can't be switched off? We finally found perpetual motion. It's huge!
Re: (Score:3)
You can't switch off silicon either. You can turn it down quite a bit so it's almost off, but it's not like it's a magic box that shuts off completely and turns on instantly when you supply exactly 0.7V across the base.
Besides, anyone on here with an ounce of hardware knowledge knows that on and off are a range. They don't switch from +- rail or to 0 instantly and accept nothing else.
Re:Congratulations (Score:5, Informative)
Graphene actually can be made to have a bandgap so this problem may only be a temporary one.
Here's a paper which discusses graphene's band gap: Direct observation of a widely tunable bandgap in bilayer graphene [nature.com]
Here's a free article discussing this: Tunable Graphene Bandgap Opens The Way To Nanoelectronics And Nanophotonics [sciencedaily.com]
In fact, here's an article about IBM doing research on this very topic: IBM opens bandgap for graphene [eetimes.com]
Note that the date of the article discussing graphine with a bandgap is after the date of the article linked in the slashdot summary discussing that graphine can't have a bandgap. Sounds like the authors of the articles need to talk to some more people and get their facts ironed out.
Re: (Score:3, Interesting)
The problem is that bilayer graphene doesn't have the same linear dispersion that a single layer has, so it might not be possible to reach such high clock speeds with bilayers. The point with single layer graphene is that the conduction electrons behave as though they are massless and so have a very high mobility. This feature is diminished somewhat in a bilayer.
Re: (Score:2)
This feature is diminished somewhat in a bilayer.
So, what, I'm going to get a lousy 300fps in Crysis, and not 500, as I hoped for? Meh!
Re: (Score:2)
To be fair, even a massive reduction in clock speed would still result in something between 5 and 10 times higher than what we currently have.
Re:Congratulations (Score:4, Insightful)
You can't switch off silicon either. You can turn it down quite a bit so it's almost off, but it's not like it's a magic box that shuts off completely and turns on instantly when you supply exactly 0.7V across the base.
You are confusing MOSFETS with BJTs.
BJT transistors have small threshold voltages such as 0.7V, however you cannot say exact for any transistor. The threshold voltage fluctuates with many material properties, temperature, and the voltage applied across the main current path (collector to emitter for BJTs, drain to source for MOSFETS) of the device.
Aside from that, its nice to take the engineers view of the world. If the 'off ' state is several orders of magnitude less conductive than the 'on ' state, then it is pretty much off. In the case of graphene, the conduction ratio for on/off is much smaller than that seen in silicon.
Final note: nothing in the world happens instantly! (except in quantum physics, where magic is just a statistical possibility)
Re: (Score:2)
Re: (Score:2)
Yes. Apparently the spacing between "on" and "off" is a lot smaller, however, which makes noise a prohibitive problem.
Re:Congratulations (Score:4, Funny)
That is not the real issue. We will run out of silicon! We need to preserve silicon for future generations.
Re: (Score:2)
Sahara called and it wants a bucket of sand back from you!
Re: (Score:2)
Sand Nazi: "No sand for you!"
I missed the first post.. (Score:1)
Re: (Score:3)
Yes, but now I cannot stop posting. :(
Re: (Score:2)
Psst. Someone tell this guy about "noise" and "filters" and the like. On or off. That's a good one, lol.
Quantization noise (Score:2)
Someone tell this guy about "noise" and "filters" and the like. On or off. That's a good one, lol.
All right. A digital signal can be thought of as an analog signal plus a "quantization noise" that's the equivalent of the difference between the digital and analog signals.
When you filter out the noise from the digital signal you have the analog signal back. The interesting part is that it you can shape the noise [wikipedia.org] so that it can be perfectly filtered out of the digital signal.
In other words, that old "wisdom" that analog signals have infinite precision, while digital signals are limited by their resolution
Re: (Score:3)
"Analog" is a bit of a simplifying assumption anyhow - if you look closely enough, everything is quantized.
Re: (Score:2)
Re: (Score:2)
if you look closely enough, everything is quantized.
That's really not of concern to anyone building electronics though.
To someone building a modern CPU it absolutely is. They are
approaching the point where single electrons count, it can't
get much more quantized than thant.
Re: (Score:3)
Re: (Score:2)
This is only true if delta-sigma modulation is performed with infinite time-resolution
Only if you need infinite bandwidth. The magic of Delta-Sigma is that you can make the quantization noise as low as you like over a limited bandwidth.
Re: (Score:2)
This is a problem? (Score:3)
So why not use it in a current-switching logic like ECL? Yeah, the power consumption might be a bit high, but you only need to make the cache and cores run faster.
Ever hear of leakage? (Score:5, Informative)
Most of the freaking chip is cache. Have a look at the floorplan sometime.
Intel engineers sometimes joke that they're the biggest memory vendor that nobody heard of.
The fundamental problem with "doesn't turn off" is that leakage current (IDS(OFF)) is already a major component of chip dissipation, even when we use all sorts of tricks to reduce it. With graphene, that goes from "problem" to "useless." Except for analog, where the transistors don't turn off anyway.
Re:This is a problem? (Score:4, Informative)
We're not talking "a bit high". ECL logic is *insanely* power hungry. If we implemented all of a processor's logic in source-coupled logic, it would consume 10-100 times more power than it does currently.
Re: (Score:2)
Considering that the energy density (as dissipation per square inch) in a CPU chip is already higher than the burner on an electric stove, that would make for some smokin' logic.
The thing that strikes me about this (Score:2)
The thing that strikes me about this is that when something new comes out that has different strengths and limitations, at first we are at a loss for how to make it useful, but then we work around the limitations and are much better off.
The things I'm aware of that make the contrast between high and low important is measurement of those highs and lows and short term memory. So those are potential areas for improvement that could make the technology viable.
Just because they haven't worked out how to do it ye
Whoah (Score:2)
100GHz? 300GHz? 1-fucking-THz?
I started drooling just a bit. Talk about a jump in speed. That's like going from floppies to thumb drives. Maybe even portable hard drives.
Re: (Score:1, Informative)
> 100GHz? 300GHz? 1-fucking-THz?
> I started drooling just a bit. Talk about a jump in speed.
The US government has had 100 GHz for around ~20 years ago using graphene. It is only "new" to the civilian world, who can't afford the multi-million dollar cooling system that goes along with it.
But still, yeah, dam cool. :-)
Re: (Score:2)
Please do post some links. In general the government has the worst and oldest gear. 20 years ago no one had made transistors out of the stuff.
Re: (Score:2)
Re:Whoah (Score:4, Informative)
Re: (Score:2)
[citation needed]
Re: (Score:2)
Oddly enough, it's hard to provide links into the nsa.
Re: (Score:2)
I question the validity of the claim on the basis that graphene hadn't been manufactured 20 years ago (it wasn't until 2004 meaningful quantities of the stuff was produced), let alone turned into transistors.
Now I can buy that the NSA may have access to some applications of science (algorithms, possibly some hardware) that is not accessible to civilians; but I do not buy the notion of the NSA being some 15 years ahead of the scientific community in theoretical physics.
Re: (Score:1)
This isn't as big of a jump as you might think. The core clock speed of your processor is significantly slower than the maximum switching speed achieved in silicon. (The first page Google provided me for "highest frequency ring oscillator silicon" didn't provide me with any truly promising links, and I'm too lazy to search more)
Ya well don't get too excited (Score:5, Informative)
The other problems with graphene mentioned aside, you start to run in to speed of light issues with extremely high frequencies. At high frequencies the wavelength is so short, that it can't travel across a chip in a single cycle. That has some real design issues. For example at the full speed of light a 5GHz signal has a wavelength of 6cm. Ok, not a problem. A core in a CPU is smaller than that so the signal can travel anywhere in a single clock, even taking in to account that wire runs could be longer. However at 50GHz, well then you are only talking 6mm. That's a potential problem. Current chips are larger than that, never mind the wire runs. Maybe if cores are kept small and simple it is fine, but it is getting problematic. At 1THz you are talking only 300 micrometers wavelength.
So even if graphene becomes practical, speeds that high may never make it in to CPUs. That a transistor can operate at those speeds doesn't mean a whole CPU can.
Re: (Score:2, Insightful)
You just need to think fourth dimensionally.
Just because stuff can't get across the chip in one cycle doesn't mean it's useless, it just means we have to rethink how we do computing at that level.
Whether this means your "pipeline" is actually just a wire, or whatnot. The fact that there is a new problem doesn't mean we throw our hands in the air and give up.
Re:Ya well don't get too excited (Score:4, Informative)
This is a problem with or without graphene. Chip design usually doesn't have signals travel all the way across the chip. In fact, good place and route engineers will keep signal nodes as close as possible and logic designers will try to keep high-connection nodes to a minimum, separating out logical clusters to be as isolated as possible.
Re: (Score:2)
IIRC there's been a fair amount of work lately on making chips asynchronous, where different parts of the chip don't have to wait for a cycle to propagate across the whole chip. I would expect that to be an essential part of any solution. Maybe the chip becomes a teeny tiny cluster with queues for each different function (*obligatory mention of Beowulf goes here*). This would make languages like Erlang more important.
Re: (Score:3)
That is, if I remember correctly how to calculate maximu
Re: (Score:1)
So wait, as the chips get faster, they (by necessity) get smaller, too?
That seems handy/annoying.
Re: (Score:1)
Surely you mean 30 cm in a nanosecond. Light travels 299,792,458 meters in one second...
Re:Whoah (Score:5, Funny)
Light travels 30 cm/sec? You sure about that? If that were true, I expect Einstein could have tested Special Relativity with a watch and a quick trot round the block.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
30 cubic meters unravels to a lot of linear meters when you break it down into planck volume pieces.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
it still losses its state when you remove power ... what it means is in a traditional transistor there is a fairly wide tween nearly off (you cant really switch them totally off either) and fully saturated (totally on) for the sake of simplicity its the difference tween ~microvolts and ~0.7 volts. This style of transistor that gap is very close together and transistors cause quite a bit of noise when zipping around even at low speeds and low quantities due to current being constantly switched on and off
bott
Re:SSD application? (Score:5, Informative)
Here's the electronics 101 version: Transistors have three ports. Electrons flow from one (the source/emitter) to another (the drain/collector). The amount of electrons that make it across depends on the mode the transistor is in, and the mode is controlled by the voltage applied to the third port, (the gate/base). There are three modes: off, linear, and saturation. In saturation, the electrons are flowing as fast as they can and small changes in the gate voltage don't matter. In linear mode, the current is directly proportional to the gate voltage - this mode is key to analog circuits. When the transistor is off, very little current gets across (on the order of femtoamps). When they say graphene transistors can't be completely turned off, they mean the amount of current that gets through when it is off is much larger than for normal transistors. It can still be "turned off" in the sense that if you take away all of the electricity, it loses its state, so there's no particular reason that it would be useful for storage.
As the article notes, a likely use would be in combination with more traditional transistors, wherein you could take advantage of graphene's speed, and then have a silicon "boot" to turn off the circuit when it's not in use by cutting off all of the power to that block.
IBM is both right and wrong (Score:5, Interesting)
Why they're right:
Graphene is a metal (or semimetal, whatever). Capacitive effects cause your current gain (ratio between input current and output current) to drop with frequency. The highest practical frequency of operation is where the gain is 1. IBM demonstrated a year ago a transistor whose current gain reached 1 at 100GHz (also known as fmax or unity gain frequency). However, that's just current gain. Digital circuits require a voltage gain as well. You can have high current gain but not have a high voltage gain by having a low output resistance.
Why they could be wrong
Without giving a crash course in electronics, computer (not all) transistors these days operate by raising/lowering a potential that either "gates" electrons and keeps them from passing through or allows them to go through freely. How big the gate is depends upon something called the band gap size, i.e, the energy required to move an electron from an atom (valence band) into participating in macro-scale conduction (conduction band). Graphene does not have a band gap.
However, you can artificially create a band gap by excluding the low energy, long wavelength electron states that exist at the bottom of the conduction and top of the valence bands. You do this by patterning your graphene into strips below about 30nm in width. In this way, no electron state with a wavelength longer than the width of the strip can exist. In a few years (right when graphene starts to hit its stride outside academia), such patterning will be possible (Intel is at 32nm now, if you recall).
Re:IBM is both right and wrong (Score:5, Interesting)
And this is what exactly I'm doing right now. Believe me, its darn hard to go below 10 nm even with ebeam lithography. But I achieved 100x on/off ratio on room temperature (10^6 on 4K). I saw IBM results on bilayer graphene with 300x on/off 1.5 years ago. There RF graphene research (on epitaxial) and DC device research on exfoliate graphene has quite different goal. You don't want to combine those into one device, because you won't achive any of those goals in time. In a long term: yes, you can achieve both. There will be fast, high on/off ratio graphene transistors, I promise.
Re: (Score:1)
Meanwhile... (Score:5, Insightful)
Re: (Score:2)
dam (Score:1)
and I was looking forward to drawing my own cpu's
diamond (Score:5, Interesting)
Diamond, on the other hand, has a band gap of 5.5 eV, and has _excellent_ thermal properties.
Re: (Score:1)
Re:diamond (Score:4, Interesting)
Unfortunately, it's a pain in the ass to get other materials to interface with diamond. I admit I haven't looked into it in a few years, but last I heard, diamond transistors looked like a dead end.
Re: (Score:2)
Re:diamond (Score:5, Interesting)
Yep, with the 5.5eV band gap you surely can turn a diamond device off. And it does have really great (the best, actually) thermal properties. But...it's really hard to get a useful n-type material (gotta boron-dope it first - which is easily done and makes a diamond p-type semiconductor - then expose it to a deuterium - yes, deuterium, not hydrogen - plasma, to passivate the boron acceptors and form some shallow donors); and there's no convenient native oxide like Si has; and etc....Diamond will take as much R&D to be useful in common electronics as will graphene. But diamond is sexier than graphene, that's for sure.
Re: (Score:2)
Yep, with the 5.5eV band gap you surely can turn a diamond device off. And it does have really great (the best, actually) thermal properties. But...it's really hard to get a useful n-type material (gotta boron-dope it first - which is easily done and makes a diamond p-type semiconductor - then expose it to a deuterium - yes, deuterium, not hydrogen - plasma, to passivate the boron acceptors and form some shallow donors); and there's no convenient native oxide like Si has; and etc....
You know, for a while there I was in doubt if this was a serious post or a "you just need to invert the shield polarity of the flux capacitator" post...
Re: (Score:3)
No, it's for real...People have been making active devices from diamond (usually CVD synthetic diamond film) for quite awhile.
By and large, the devices live up to the potential implied by diamond's electronic and thermal properties. Around a decade back, I saw a diamond FET that demonstrated the material's potential for high temperature devices. It was running at about 750C, and you could read by the emitted thermal glow! Pretty impressive.
There's been lots of work on diamond detectors for hard rad enviro
Silicone's demise? (Score:2)
Re: (Score:3)
somehow they failed to check google (Score:5, Informative)
Just curious (Score:2)
My understanding is that analog computing is gaining ground in the research field (after being dormant for decades) and this seems like perfect timing.
Bilayer graphene can do this (Score:1)
Transistor Speed Is not the Issue (Score:2)
Wow... A lot of this is missing the main issue that has now limited the speed of processors for a number of years - Limitations of the interconnect.
The average microprocessor is not limited by the capability of the transistor, but rather the RC time constants associated with the connections between them.
Thats the reason, aluminum as a metal interconnect was dropped a while back in favor of copper. Lower R for the same C.
Analog computing? good luck with that!