Sign in with
Sign up | Sign in
Your question
Closed

FAST! IBM Develops 100GHz Transistor Device

Last response: in News comments
Share
February 8, 2010 8:02:51 PM

All my wildest dreams will come true.
February 8, 2010 8:11:49 PM

AND YES OF COURSE THIS WILL RUN CRYSIS
February 8, 2010 8:16:37 PM

Hmmmm I wonder how long before they have a fully working cpu developed on this new transistor that can run modern X86 code as well power PC code.
February 8, 2010 8:21:48 PM

that data means nothing if they do not supply the electrical and thermal requirements
February 8, 2010 8:25:21 PM

Is the source of graphene (graphite) as cheap as silicon? I'd be glad to have more speed, but if it doubles prices I might have to cry.
February 8, 2010 8:27:08 PM

It seems a long way for it to reach our life yet.
February 8, 2010 8:28:49 PM

grillz9909Is the source of graphene (graphite) as cheap as silicon? I'd be glad to have more speed, but if it doubles prices I might have to cry.


Well how much does pencil lead cost? Its pretty cheap stuff. Although I would hazard a guess that the first few batches of CPU's would cost much much more then your Core i7.... Honestly this is exciting and will allow for whole new kinds of software to be developed! The things that take hours to do now would take seconds...
February 8, 2010 8:32:27 PM

Will it ru-

DAMNIT AMABHY.
February 8, 2010 8:36:35 PM

OMFG!!!!

I just read the other IBM article about the Power 7's

8 core's, 32 threads, now just imagine if you could combine the above 100Ghz into a package like that.
Want to encode the extended edition of Return Of The King in 4 seconds?
No problem.
Want to open 1000 photo's from a 20MP camera in CS3?
No problem.
Want to run 10 separate instances of Crysis at 2560 x 1600?
No problem.

OK, we aren't there yet, but the future is looking very bright.
February 8, 2010 8:43:51 PM

But will it run Crysis 2?
February 8, 2010 8:52:15 PM

Forget about "LUDICROUS SPEED", when do we get Plaid speed devices?
February 8, 2010 8:54:14 PM

Keep in mind a 100 GHz transistor will NOT make a CPU with a clock speed of 100 GHz. For a single clock cycle, each transistor may have to flip SEVERAL times; there needs to be enough done that each logic gate can at least process a cycle ONCE.

As the article noted, top-end silicon transistors currently top out at ~40 GHz, which would be about what Intel uses for their sub-4 GHz CPUs... So that means perhaps 10-12 transistor cycles per stock clock cycle.

Still, this is a very good piece of news; it's STILL 2.5 times faster than the currently-used tech, which means we could see CPUs with it in the 8+ GHz range. Furthermore, the fact is that they're building these out of graphene, a material basically similar to the so-called "nanotubes." (the difference being that graphene is a flat sheet, while nanotubes are fixed in a hollow tube-shape)

While the current mentioned "size" is listed as 240 nanometers, I'm not sure how that equates to actual process size equivalent, since logic gates tend to be longer than they are wide (they consist of a lot of transistors) and the "feature size" for fabrication processes typically goes on the distance when placed side-by-side. However, since this is basically nanotechnology, there's tons of room to shrink them down in size, potentially into the sub-nanometer range. (the distance between carbon atoms being about 140 picometers, aka 0.14 nm)
February 8, 2010 9:12:21 PM

back_by_demandWant to run 10 separate instances of Crysis at 2560 x 1600?No problem.OK, we aren't there yet, but the future is looking very bright.

One problem: Crysis is for Windows only and the CPU is most likely to be PPC.
February 8, 2010 9:13:46 PM

HeadlineThis could one day replace silicon to give us... LUDICROUS SPEED!
Quake
February 8, 2010 9:16:02 PM

4treesAll my wildest dreams will come true.

When your 60!
February 8, 2010 9:51:14 PM

amabhyAND YES OF COURSE THIS WILL RUN CRYSIS



No, a single transistor will not run anything. A few million of them together might be able to.
February 8, 2010 9:54:49 PM

well with this achievement we are 1/100 of the way to terahertz speed on a single chip.
February 8, 2010 9:59:44 PM

Impressive. It's good to know some worthwhile work is being done to advance clock speeds, at least in terms of the future.
February 8, 2010 10:21:40 PM

This was reported a few days by other sources but in the details it said that IBM hadn't operated it at 100Ghz -- they had operated it more slowly but extrapolated their data to predict that it COULD operate at 100Ghz. Either they did a lot of work in a couple of days or some important details got dropped out of this article.

Nonetheless an exciting advancement.
February 8, 2010 10:27:03 PM

JonathanDeaneWell how much does pencil lead cost? Its pretty cheap stuff. Although I would hazard a guess that the first few batches of CPU's would cost much much more then your Core i7.... Honestly this is exciting and will allow for whole new kinds of software to be developed! The things that take hours to do now would take seconds...

this is not the case, graphite and graphene are NOT the same price. graphite is a naturally occuring compound composed if sheets of graphene bonded together. graphene is hard to extract and there is no current industrial process that can mass produce this substance efficiently. it is currently surpasses gold as one of the most expensive materials known. making a main stream chip with this is years away and it will only happen when the price comes down.
February 8, 2010 11:07:08 PM

amabhyAND YES OF COURSE THIS WILL RUN CRYSIS


On one transistor? doubt that...
February 8, 2010 11:28:29 PM

xaiosMethinks you might want to check your math. We're actually 1/10 of the way to a single chip operating at 1Thz, if this article is to be believed.


yea your right. wish toms had a edit button
February 8, 2010 11:39:21 PM

Overclocking anyone? :p 
February 8, 2010 11:46:05 PM

intel craps its pants... and will offer ridiculous bundle pricing that will force manufactures to stay intel only
February 8, 2010 11:48:50 PM

cheepstuffthis is not the case, graphite and graphene are NOT the same price. graphite is a naturally occuring compound composed if sheets of graphene bonded together. graphene is hard to extract and there is no current industrial process that can mass produce this substance efficiently. it is currently surpasses gold as one of the most expensive materials known. making a main stream chip with this is years away and it will only happen when the price comes down.


Well sand is silicon the same thing they make chips out of.... The price difference is of course huge. I doubt that after they start making an honest stab at making the stuff the price will be very much different. Sure right now its expensive as hell as was the first batches of doped silicon. When talking about new things the price is always absurdly high.

The point is when your buying computer chips the raw material is hardly something to be worried about (your buying a very small item to begin with so no matter how the material costs its still going to be a small percentage of what your paying for)

February 9, 2010 1:21:26 AM

"Want to encode the extended edition of Return Of The King in 4 seconds?
No problem."

"Want to open 1000 photo's from a 20MP camera in CS3?
No problem."

"Want to run 10 separate instances of Crysis at 2560 x 1600?
No problem."

I think to be able to do the things above, we need to have faster storage options... we need faster hard drives.


February 9, 2010 1:22:28 AM

Hopefully, this new one would produce something way better than the current existing solutions.
February 9, 2010 1:52:53 AM

cheepstuffit is currently surpasses gold as one of the most expensive materials known.

Actually, gold never was the most expensive material by a long shot. while it made big news after passing $1,000US an ounce a bit back, Platinum has well passed $1,500US... And it's ALSO used in computer production as well, as it's one of the three main metals used for the magnetic coating in hard disks, along with Palladium, (which currently runs $400US/oz or so) and Cobalt. (which isn't that expensive) And to say nothing of the price of other, more-restricted materials, like plutonium-238, (the material for "atomic batteries") where the price of a single GRAM can run well over $4,000US, making it over 100 times as valuable as gold. Many radioactive isotopes run higher, like Californium-249, which can be well over $100,000,000US per gram.

But at any rate, the good news is that exfoliated, pure graphene may run at about $28-44US million per gram, these chips are not using pure stuff, but just graphene bonded to silicon carbide, which is a millionth the price. (i.e, $28-44US a gram) Further, since the cost of the materials is almost zero by comparison, refinement in production techniques can bring this down further, making it as cheap as traditional silicon fabrication.
February 9, 2010 2:25:36 AM

This starting to to sound like really big bucks are at stake.
Anonymous
February 9, 2010 2:28:07 AM

Now my endless loop will run in under 2 hours.
February 9, 2010 3:08:16 AM

icedeocampo"Want to encode the extended edition of Return Of The King in 4 seconds?No problem.""Want to open 1000 photo's from a 20MP camera in CS3?No problem.""Want to run 10 separate instances of Crysis at 2560 x 1600?No problem."I think to be able to do the things above, we need to have faster storage options... we need faster hard drives.

intelligent comment fail
February 9, 2010 5:38:26 AM

Shadow703793One problem: Crysis is for Windows only and the CPU is most likely to be PPC.

Hence the further quote
back_by_demandOK, we aren't there yet, but the future is looking very bright.

We're not, but it is.
February 9, 2010 6:19:14 AM

Will it bottleneck with the video card?
February 9, 2010 7:28:56 AM

i understand
February 9, 2010 7:41:11 AM

i understand that the benchmark of 100GHz is eye catching but is this really news? carbon is the best and most versatile element in existence. it can do almost frigging anything. graphene is old news and it is strange that many companies have not been pouring millions into R&D. wanna make some for cheap? take a #1 pencil and draw on something or shade a large area. instant(poor quality) graphene! im sure some of you must remember manually increasing(or attempting to) your nVidia GPU V-core voltage buy meticulously shading resistors directly on the card or oc'ing old school athlons or durons in a similar way...
February 9, 2010 7:45:49 AM

jonnyzati understand

Tom's Hardware software fail
February 9, 2010 8:01:31 AM

This has me excited. =D
February 9, 2010 8:05:05 AM

notthekingActually, gold never was the most expensive material by a long shot. while it made big news after passing $1,000US an ounce a bit back, Platinum has well passed $1,500US... And it's ALSO used in computer production as well, as it's one of the three main metals used for the magnetic coating in hard disks, along with Palladium, (which currently runs $400US/oz or so) and Cobalt. (which isn't that expensive) And to say nothing of the price of other, more-restricted materials, like plutonium-238, (the material for "atomic batteries") where the price of a single GRAM can run well over $4,000US, making it over 100 times as valuable as gold. Many radioactive isotopes run higher, like Californium-249, which can be well over $100,000,000US per gram.But at any rate, the good news is that exfoliated, pure graphene may run at about $28-44US million per gram, these chips are not using pure stuff, but just graphene bonded to silicon carbide, which is a millionth the price. (i.e, $28-44US a gram) Further, since the cost of the materials is almost zero by comparison, refinement in production techniques can bring this down further, making it as cheap as traditional silicon fabrication.

Well, now my question is, what could they do using pure graphene?
$100,000,000 CPU anyone?
Also, interesting to know just how valuable a little Californium-249 is.
February 9, 2010 8:11:48 AM

That something can cycle 100 million times in one second boggles the mind...
February 9, 2010 8:13:55 AM

Whoops, I meant billion. 100MHz is still amazing when you think about it though.

Ahh... there's really no way I can get out of looking like a moron just now is there.
February 9, 2010 8:15:46 AM

MrMaestroThat something can cycle 100 million times in one second boggles the mind...

Kilo = 1,000
Mega = 1,000,000
Giga = 1,000,000,000
100GHz = 100,000,000,000 (100 billion) cycles per second.
It's a big difference. =D
February 9, 2010 11:11:21 AM

"Hey Vegeta, what does the software say about the gigahertz level?

"IT'S OVER NINE...TY!!!"
February 9, 2010 11:24:17 AM

Two comments...

#1 - just because IBM got a chip up to 500GHz does not diminish this article in the slightest. They had to use an exotic solution to get to 500GHz. This 100GHz transistor is probably at or near room temperature. If you used this 100GHz transistor to make a chip and then put it in the same freezing solution, that 500GHz barrier would probably look like nothing.

#2 - Yes, that current transistor would only make a chip in the 8+ GHz range for now. But remember that it says a gate length of 240nm. Imagine what it could do at say 32nm or even smaller. That won't be immediate, but long-term it very well could be.

Truly a good find by IBM!
February 9, 2010 11:41:28 AM

amabhyAND YES OF COURSE THIS WILL RUN CRYSIS

I think cutting people off before they even say this should count as just as bad as saying it. How different is it to ask whether it'll play crysis vs. just saying that it will? Useless and annoying either way.
February 9, 2010 11:45:41 AM

Also, one extremely important thing they failed to mention in this article:

Lin cautioned against thinking of graphene as a substitute for the silicon-based microprocessors used in today's computers, at least at anytime in the near future. One major roadblock is that graphene does not work easily with discrete electronic signals, he explained. Because graphene is a zero bandgap semiconductor, meaning there is no energy difference between its conductive and nonconductive states, transistors made of the semiconductor cannot be turned on and off. In contrast, silicon has a bandgap of one electron volt, making it good for processing discrete digital signals, Lin said.

Instead, graphene is better suited for making analog transistors, such as signal processors and amplifiers. Today, such circuitry is largely made from GaAs (gallium arsenide), though GaAs offers nowhere near the same electron mobility, Lin said.
February 9, 2010 11:58:03 AM

hmm maybe phase changing cooling will be the standard for this kind of speeds :) 
February 9, 2010 12:12:09 PM

I thought the transistor gates in a CPU opened at 1 per hertz, so a 4GHz would open gates at 4GHz, not multiple times per GHz. The frequency is what causes the gates to open, no?

Yes, it takes more than one cycle per instructions, say 22 steps in the cpu core for what ever type of chips, but that means 22 frequency cycles, not 22 times in 1 hertz.
!