bit-tech.net

Graphene transistor hits 300GHz

Graphene transistor hits 300GHz

Graphene - a clean, safe future alternative to silicon in CPUs - has been used to make a transistor which runs at a whopping 300GHz.

It seems to be the week for age-old game-changers to dust off their research and hit the headlines again: as well as HP announcing a deal with Hynix to bring the long-awaited memristor to market, scientists have announced the development of an ultra-fast graphene transistor.

Graphene, a form of carbon in the shape of a honeycomb lattice, first came to our attention as a future replacement for silicon in processors back in 2007 following research by Princeton University and then promptly disappeared, as do so many other "killer" technologies that are trumpeted by universities.

Scientists aren't willing to let go of graphene that easily though: according to an article on IEEE Spectrum, a team at the University of California at Los Angeles has created a nanowire-based graphene transistor that operates at speeds silicon-based equivalents just can't touch: 300GHz.

Such performance is around twice as fast as that achievable from equivalent silicon-based transistors, and similar to what can be achieved with those made from expensive and dangerous materials such as gallium arsenide. Better still, the technology shouldn't cost more to produce than existing silicon-based transistors .

With the technology believed to scale into terahertz, graphene could be the new silicon - but a single transistor does not a CPU make. It's clear that teams working on graphene technology have a long way to go before we start enjoying the benefits in our PCs.

Are you pleased to see that graphene research hasn't hit a dead end yet, or will you only get excited when Intel, AMD, or ARM announce their first 1THz graphene-based chips? Share your thoughts over in the forums.

32 Comments

Discuss in the forums Reply
Andy Mc 2nd September 2010, 10:42 Quote
It would be nice to see speed of 4ghz+ without any overclock needed.
shanky887614 2nd September 2010, 10:52 Quote
i think that it would be nice to have this but it would require a major overhaul of motherboards and the technology that transfers data becasue at the moment becasue of how fast optical drives and hardrives are, there wouldnt be much benerfit from this at this time and that is if you could find a motherboadr that would be compatable
rickysio 2nd September 2010, 11:00 Quote
Next time instead of GHz we'd be talking about how many THz your CPU is clocked at.
adam_bagpuss 2nd September 2010, 11:05 Quote
OMG it sover 9000 THz !!!!!

i haz uber pwnd u
Plugs 2nd September 2010, 12:32 Quote
these transistors are twice as fast as current silicon ones, and can (possibly) scale to another order of magnitude up
which means approximately 10x - 20x faster than current CPU offerings

30GHz-60GHz (stock)
doesnt sound too shabby
shanky887614 2nd September 2010, 13:06 Quote
i dont know about you, but i dont much care for the clock speed

it is how efficent they are is what really counts

for example a core i3 is more efficent and faster than the same speed core 2 duo

thsi is the kind of thin i want not just more speed
Mraedis 2nd September 2010, 13:30 Quote
I might be way off here, but I think it's a logical guess.

If you replace every transister in mobo/cpu/ram etc, will they all operate faster, if they operate at all?
BRAWL 2nd September 2010, 13:33 Quote
There be only one response for such an incredible speed boost...

"BY THE EMPEROR IT SHALL BE SO!" I hereby demand that these things are used excessivly in future technology... of course said demands will fail, however... I can deal with a 60GHz CPU....
mecblade 2nd September 2010, 14:25 Quote
im quite interested in the response to this, from the companies like ARM and Intel though. If they manage to pull off a Graphene based CPU, they could have a massive advantage against the rest of their rivals. However, im sure Intel would manage with the amount of money theyve been spending last month. i wish they could spend 9 billion on good ol R and D :(
dyzophoria 2nd September 2010, 17:12 Quote
as long as I dont need a 20,000 watt PSU to handle a 1THZ, that would be cool, honestly, even 1kw systems now are ridiculous,lol
perplekks45 2nd September 2010, 18:44 Quote
That'll suck the socket out of your wall!
cgthomas 2nd September 2010, 20:00 Quote
I just got rid of my 64-core 46Ghz cpu, got quite old tbh. Planning to upgrade to the latest Intel 128-core 80Ghz plus a pair of Corsair DDR8 128GB ram. Though not sure wether it can run Crysis 2?
Teq 2nd September 2010, 20:08 Quote
I'm still waiting to release the kill bots, we need more power!
borandi 2nd September 2010, 20:11 Quote
Manipulating graphene is very tricky, compared to silicon. Bulk silicon, can be cut, etched, eroded, in a very regular way. Graphene, SWNT and MWNT can be a bitch, and don't offer the same categories of manipulation that silicon does. Thus the equipment to make a decent transistor array will need to be invented before graphene becomes the de facto material.
thehippoz 2nd September 2010, 20:17 Quote
just the leap in what you could do on the software end with that kind of speed.. huang needs to switch to graphene and then yell at everybody
ChainsawBunny 2nd September 2010, 20:18 Quote
Finally something that can run Crysis at 60 FPS?
Pete J 2nd September 2010, 23:53 Quote
Hmm.

What about the heat generated by one graphene transistor? Just because one can operate at 300GHz, it doesn't mean that a whole load put together will work at this frequency.

*goes off to look at the thermal properties of graphene versus silicon*
cgthomas 3rd September 2010, 01:10 Quote
Quote:
Originally Posted by ChainsawBunny
Finally something that can run Crysis at 60 FPS?
That's right, but what about Crysis 2?
I'm thinking about adding another 2 GTX 470 so I can have 3-way SLI. I would then be really happy if I can run everything on max at 23 - 36 FPS. I would be a happy man
Pete J 3rd September 2010, 11:21 Quote
Well, the thermal conductivity of graphene seems to be quite amazing, so heat will be transferred far better out of a chip compared to silicon. I can't find any thermal expansion numbers though (though I've found it has a negative CTE between 30 and 300K).
Showerhead 3rd September 2010, 11:48 Quote
Would the efficiency of the transistor not have an effect on heat output? Or is all the energy lost as heat?
Elledan 3rd September 2010, 12:06 Quote
Quote:
Originally Posted by borandi
Manipulating graphene is very tricky, compared to silicon. Bulk silicon, can be cut, etched, eroded, in a very regular way. Graphene, SWNT and MWNT can be a bitch, and don't offer the same categories of manipulation that silicon does. Thus the equipment to make a decent transistor array will need to be invented before graphene becomes the de facto material.

As we're moving towards stacking dies and 3D designs, though, it seems likely that we'll need a better way to conduct the produced heat away from the die, and that silicon isn't going to work there. Miniature watercooling has been proposed, but seems quite complex to get right :)

Whether graphene is the answer here I'm not sure, but it's most definitely an interesting material.
Mraedis 3rd September 2010, 13:35 Quote
Synthetic diamond is getting cheaper and easier to produce, and conducts heat much, much better than copper and still better than graphene.
Elledan 3rd September 2010, 15:27 Quote
Quote:
Originally Posted by Mraedis
Synthetic diamond is getting cheaper and easier to produce, and conducts heat much, much better than copper and still better than graphene.

Well yeah, diamond is an even more perfect molecular carbon matrix than graphene, so that makes sense :)

On a sidenote, I still find it funny to see people express disbelief at the fact that diamonds can easily be burned :D
SpeedLegion 3rd September 2010, 18:46 Quote
Quote:


"but a single transistor does not a CPU make"

When did Yoda start writing for Bit-Tech???
SWEET!
somewhereoveryonda 3rd September 2010, 23:43 Quote
Incase anyone is wondering what this actually is... its very similar to graphite. Where graphite has 3 bonds to other carbon atoms giving a 'spare bond' if you like. This leaves an unbonded electron which becomes free (delocalised). These bonds form sheets of one atom thick... Graphene is basically this however the sheets form tubes. If your interested check out "Buckminsterfullerene"... These are small balls of 60 atoms of carbon. They are formed in the same way and have been around since 1985! I hate to alarm you folks but this is still in research stages at Unis and its taken this long since they were discovered to produce anything useful! This is about 10 years from making production yet!
Zinfandel 4th September 2010, 00:20 Quote
Quote:
Originally Posted by somewhereoveryonda
Incase anyone is wondering what this actually is... its very similar to graphite. Where graphite has 3 bonds to other carbon atoms giving a 'spare bond' if you like. This leaves an unbonded electron which becomes free (delocalised). These bonds form sheets of one atom thick... Graphene is basically this however the sheets form tubes. If your interested check out "Buckminsterfullerene"... These are small balls of 60 atoms of carbon. They are formed in the same way and have been around since 1985! I hate to alarm you folks but this is still in research stages at Unis and its taken this long since they were discovered to produce anything useful! This is about 10 years from making production yet!

Noble Prize for Chemistry awarded to Bolton's first Nobel Prize winner (I'll be its second, obviously) Harold Kroto!

Buckyball <3
Zurechial 4th September 2010, 00:57 Quote
The problem here is that people are thinking "300GHz? Wow, my CPU only runs at 3.2GHz!".
Note that this graphene, according to the article, can achieve speeds twice as fast as normal silicon.

That means someone out there has been running silicon at 150GHz, or could if they wanted to (for single transistors, presumably) - Right, so why haven't we got CPUs at 150GHz yet? Because it ain't that simple.

That means in turn that just because a single Graphene transistor can run at 300GHz doesn't mean CPUs made from it will too.

Sure, it's great to see technological improvements and I'm not slagging off the article or the news itself at all here, but my excitement usually fades when I read these articles closely and see that it's yet another thing that won't leave the lab for another decade.
metarinka 4th September 2010, 01:59 Quote
like every newcomer technology, it might not ever get developed going against such a well embedded and vested technology like silicone.

Simply put now silicone transistors are relatively well understood and the scientific and engineering hurdles to increase speed more palpable. Trying to go from nothing to a full graphene product would take decades of research and 100's of millions in research and development for something that might not even wash out.
Fizzban 4th September 2010, 08:09 Quote
Quote:
Originally Posted by Zinfandel
Buckyball <3

http://www.nanotech-now.com/nanotube-buckyball-sites.htm

Pretty interesting.
dangerman1337 4th September 2010, 14:30 Quote
Woudln't a Processer (unless the graphene can somehow negate the effects) at that speeds bloody hot and very power hungry?
Timmy_the_tortoise 6th September 2010, 18:40 Quote
Quote:
Originally Posted by dangerman1337
Woudln't a Processer (unless the graphene can somehow negate the effects) at that speeds bloody hot and very power hungry?

According to the article, current Silicon based transistors can switch at roughly 150GHz.. Obviously these frequencies aren't used in ICs because it would get ridiculously hot... so I'm guessing that these Graphene based chips would not run at 300GHz... but could still probably run a lot faster than Silicon ICs, I'd imagine.
Elledan 6th September 2010, 19:07 Quote
Individual transistor speed is largely useless in determining the clock speed of an entire circuit. That depends far more on the slowest component in the entire circuit, which limits throughput, and of course there's always communication latency for on-die traffic. A zero-latency, zero-leakage circuit could maybe run at 100 GHz or so. Of course, that's just a pipe dream :)
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums