IBM demonstrates 100GHz graphene transistor
By Donald Melanson 
posted February 7th 2010 8:16AM
posted February 7th 2010 8:16AM
It's just been a little over a week since IBM researchers announced that they managed to open up a bandgap for graphene-based field-effect transistors, but they're now already back to show off what that's made possible: a 100GHz graphene transistor. What's more, this latest record-setting transistor (which IBM hopes will one day replace silicon transistors) was made using processing technology that's compatible with that currently used in advanced silicon device fabrication, which should no doubt help speed up its eventual commercialization. Of course, any widespread adoption is still quite a ways away, but IBM says that this new transistor "demonstrates clearly that graphene can be utilized to produce high performance devices and integrated circuits." For those keeping score, this first-of-its-kind transistor already beats the frequency performance of current state-of-the-art silicon transistors of the same gate length, which now top out at a mere 40GHz.
now can we create a super advanced race of robots? or just faster netbooks so steve just shuts up
Let's hope this one makes it to the public unlike all these other breaktrough discoveries we hear of and never see again..
@Karim the chip in your netbook/notebook/tv/mobile/washing machine...all of them were news & inventions some years ago :-)
@Karim
Yes, we need to realize that inventions like these do not come to the mass market in a few months. That said, this look very promising because of the fact they used the existing fabrication techniques.
Moore would be proud.
@Invader Par
He got moore than he bargained for!
*insert drums and recorded laughter here*
@TMC
www.instantrimshot.com
@Invader Par Actually he wouldn't. Moore's law is all about adding more and more transistors. It's not about making the transistors faster, or even about making the processors faster, but about having MOAR TRANSISTORS!
It's a pretty stupid thing to say, really. It's an even dumber philosophy.
@KarlW Moore jokes that he's credited with inventing exponential growth.
@Spiraling Shape NOpe! Ray Kurzweil did!
@Invader Par
What noise does running over a boy holding a cymbal make?
Ba Dum Tish*
I don't care what the hell this bs is about, I just want my 100ghz processor. Now.
@Exbloder If only it were that easy, where's my 40GHz PC?!!?
I thought optical computing would be the next big thing due to "limitations" of this type of processing, but this happily keeps measly electrons relevant!
@Exbloder
And the scientists don't care about you either. :)
Before anyone asks, I'm quite sure it'll run crysis, and crysis 2.
@aero I'm sure It can run 50 crysis processes.
@aero
Yes but what about Crysis 3.
@Federaly Not without fermi
@aero u kiddin? this could host a crysis streaming server for more than 100+ crysis games
iBM is so awesome
It's always so hard to imagine the future, thinking of looking back at overclocked Core i7's and laughing at their measly performance compared to our 100ghz graphene based processors, and whatever other tech we might have, quantum computing perhaps?
Its a faster switch. There are still many problems plaguing the graphene switching capability.
This doesn't mean we get 100Ghz. It is one transistor, and they already said that silicon gets 40Ghz, which it doesn't in real life situations. This wont make our chips 100x faster, maybe 2-5 times.
The bonus here is that now we can increase sequential calculations. For the last few years, companies have basically just been adding more cores to increase calculations, rather increasing clockspeeds substantially. This is good, but still limits the speed of our applications, as not all instructions can be operated in parallel. This will lift that bottleneck, hopefully.
@jol
Hopefully this doesn't deter the shift toward multi-threaded programming in general, however. While faster clock speed will always give us linear performance increase; each time intel/AMD doubles the number of cores in their CPUs, performance can potentially increase exponentially. Unfortunately, few development firms have adopted the use of multi-threaded programming needed to take advantage of dual/quad/hex core CPUs.
@Jyncus
So true, which is why i'm so glad I haven't upgraded my dual core PC. Quad and Hex cores are coming out but it always takes a long rime for the software makers to support them.
@Plazmic Flame
And what about power? With 100GHz, we need a water plant to cool down the system.
Off course 100GHz will not be used in the average computer as well.
@Jyncus: These days I think most development teams working on software with high CPU requirement have adopted it, but there are cases where it's really difficult to do well so you have to look for specific tasks that you can farm out to other processors to make it faster, even if the general thing is single-threaded. (For instance, a web browser might basically be single-threaded, but when it plays video, the video decoder will run in a different thread. That kind of thing.)
A specific area that was holding things back was games, but thanks to the current HD consoles requiring it (assuming the game has to do any challenging cpu work at all), game developers have had to adapt too.
There are also cases where it's easy and has been done for years (servers).
So anyhow I don't think we have a particular problem in the next year or so; software that requires a high degree of performance will probably all be able to make use of a reasonable number of processor cores. The picture might be a bit different if we ever move to a situation where the number of cores is really high (hundreds) because then the challenges involved start to look more like those currently associated with supercomputer applications...
@sourav - Could be wrong on this, but I thought one of the reasons for the eventual switch to graphene was a better heat tolerance?
@Jyncus
Unfortunately, your program will run as fast as the slowest linear process.
See http://en.wikipedia.org/wiki/Amdahl%27s_law
You need this speed to run Flash on a Mac, say Apple.
@Sarcasme And in standard def, and the video can't be more than 5 minutes
@tvick47
thanksfor completely unrelated juvenile comments guys! Plz read the guide. Stay on topic.
I have a quad core q6700 that while running flash video can get up to 70% CPU. It's not macs, it's flash.
@The Dead Marxist Trio
I don't know what your problem is but when I run a flash video on my Vaio with a T7500 processor, the CPU usage doesn't go above 50% and that's at 720p resolution..
@Karim Well that's because you're on a PC. Flash on Mac is way more buggy than on Windows. If I watch a 480p YT vid on my MacBook, my CPU goes up to 80%. If I watch the same vid on Windows. I get around 67% (on a VAIO). And HTML5? Up to 30%. Even though Flash is a platform, HTML5, which is a language is superior.
@Vincent but HTML makes the browser more boggy, where as the flash plug in's only load when they are needed. In terms of making an efficient browser the switch from Flash (optional plugin) to integrated HTML is a design compromise. Adobe obviously needs to improve Flash~
Sounds like the beginning of Skynet. And so it begins... Anyway, I always like speed -- but the trend towards fully multi-threaded applications needs to continue as well.
I'm going to overclock mine to 180GHz.
See THIS is the kind of electrical engineering I want to do, not programming. Only EE jobs I've been finding lately are programming centric. I want to do research stuff like this. If anyone there is reading this, I'm a graduated EE and I WILL work for free if it's on something spectacular like this. I salivate at the notion =D
@deciBels
unfortunately, many engineers end up stuck in cubicles coding all day. You could possibly do work like this, but I think most of the engineers and scientists who do work like this have MSEEs and PhDs in EE specialties specifically related to semiconductors, mems, materials, etc.
I'm sure there are also plenty of physics PhDs there as well. It's a cool field, but tough... so I guess the question is, are you willing to go the research route, and get your PhD as well? It could be rewarding if you're good at it - but otherwise you might just end up hating your life, which is never good.
@deciBels, this goes more into the domain of physics and chemistry, than it does in electrical engineering. When they stabilize it and start applying the new tech, you should consider sending your CV.
I read about this on Ars, they only demonstrate it running at 30ghz, with data showing (extrapolating) that 100ghz operation would be possible - albeit not great.
IBM also created a processor prototype with integrated water cooling channels going right thorough the actual chip, how cool is that? Imagine shedding all that weight off a laptop and also removing the need for a fan!
@Nitesh so if there's no fan, how are you going to cool the water/liquid? or are they just going to have laptops with a huge burn-your-hands heatsink on the back? or the laptop only runs when you connect it to your cold-water tap? :)
don't get me wrong, it's neat technology, but specifically for laptop cooling I think the real solution is chips with lower power consumption - batteries aren't going to get much better any time soon, so reducing power usage solves both the 'heat' and 'battery life' problems.
While I do think this was awesome work, there are two things to keep in mind:
a) they only measured up to 10GHz and extrapolated towards a possible cut-off frequency of 26GHz
b) they are assuming that, according to the theory, a change in feature sizes would allow frequencies up towards the THz-regime
And, as it often is, it is not trivial to turn theory into operational devices.
It great to see you guys reporting on this but you are dead wrong. The 100 Ghz figure is NOT clock speed. It is cut-off frequency. This is a figure of merit for a transistor used in RF/analog circuits and not digital circuits. It determines roughly the maximum theoretical bandwidth when the device is used in amplifier configuration. For the purists, higher order poles are located some fraction of the cut-off frequency of the device.
Until the graphene develops a bandgap and can be doped both n and p type, we won't be seeing it used as a digital switch.
@ethanhunt They already did manage to open a bandgap - about 130mV.
Whats the power consumption?
@odin607
My thoughts, exactly.
Paving the way for Moores law to continue: where many core scaling becomes physically limited by footprint/ die size issues, then this will enable Intel and others to pump up the throughput of the existing poly-core frameworks of future IC's.
Word up IBM^^ are they going to bust out with something that can give the 6core gulftown a run for its money? A revival of PowerPC? Should be interesting to see how this technology unfolds into application.
Likewise 40GHz in state of the art silicon is really something considering that most all current dual and quad core architectures speed step peak at under 4.5GHz. Looks like there is lots of head room for future GPU/ CPU/ APU/ SOC's to gain performance boosts; lets hope that we can see these speed increases roll out with efficiency improvements; since the better the performance per watt = better battery life (From laptops, to the inverters in battery electric and fuel cell vehicles) improvements to the core material science governing our ability to switch, compress, scale, or otherwise manipulate electrical energy and the information and power it carries; will be crucial to pushing the continued evolution of science and technology.
Since the transistor has been widely hailed at the most pivotal and ubiquitously influential technological development in history; any improvement made to transistor technology has the ability to push forward the boundaries of what is possible into application what currently resides in the realm of imagination.
Onward with innovation ^^ Two thumbs up IBM
100 GHz at the same fabrication process? like 100 GHz on 40nm die?
yea tat in theory probably would work but wouldn't that fry up the chip?
Or graphene has very low thermal envelope?
This probably might need a new fabrication process, maybe even lower than 25nm, to operate under a typical TDP