Print this article  |  E-mail this article  |  Comment on this article

500 GHz IBM Chip Breaks Speed Record

By Ed Oswald, BetaNews

June 21, 2006, 4:24 PM

Researchers at IBM and Georgia Tech said Tuesday that they had created the first silicon-based chip capable of operating at speeds of over 500 GHz. The processor beat the previous speed record by 125 gigahertz, and is some 125 times faster than today's fastest chips.

To accomplish this feat, the researchers froze the chip to near absolute zero -- minus 451 degrees Fahrenheit. IBM said that computer simulations seem to indicate that the silicon-germanium (SiGe) technology could eventually support near-terahertz operation, even at room temperature.

At room temperature, the chip operated at 350 GHz, slightly slower than the record but still over 87 times faster than the fastest commercially available chips. According to IBM, SiGe chips operate faster at very cold temperatures.

"This groundbreaking collaborative research by Georgia Tech and IBM redefines the performance limits of silicon-based semiconductors," said Bernie Meyerson, vice president and chief technologist at IBM. "IBM is committed to working closely with our academic and industry partners to deliver the insight and innovation that will enable a new generation of high-performance, energy efficient microprocessors."

The two organizations said that these ultra-high performance chips could have several important uses, including commercial communications systems, defense electronics, space exploration, and remote sensing.

It also shows the industry that silicon-based chips are far from reaching their performance limit, which means they could be produced using conventional methods and at a reasonable cost. The two companies plan to continue work in SiGe technology in order to spur innovation in processor design.

"Understanding the basic physics of these advanced transistors arms us with knowledge that could make the next generation of silicon-based integrated circuits even better," said John D. Cressler, Byers Professor in Georgia Tech's School of Electrical and Computer Engineering.

Add a Comment (63 Comments)

BetaNews reserves the right to remove any comment at any time for any reason. Please keep your responses appropriate and on topic. Foul language and personal attacks will not be tolerated.

Name (required):

E-mail (required):

Enter Your Comment:

By hrxy

edited Jul 7, 2006 - 10:39 PM

Why do you need to use SiGe why not pure Ge like the first generation of transisters were?

Score: 0

By knightsquest99

edited Jun 22, 2006 - 7:25 AM

Reading these news reports one has to wonder when these so called marvels of computer processing power will ever make its way down to the consumer. If they have the chips that will do all this processing power why have we not seen really big improvements in processor speeds.

I know AMD and Intel are touting thier dual and quad cores as the lastest and greatest thing on the market but if you really look at it the speed has not increased. It would be nice if the chip makers would really give us a product that would be fast and not break our bank accounts

Score: 0

By jstroke

edited Jun 21, 2006 - 10:33 PM

Get over it! It's a breakthrough technology!! These chips won't run on your most brand new chipsets that you just bought. Congrats that you have those, bottle necks and all, but they're designed for the chips you put in them. When this research breakthrough does find it's way down to the consumer level, you fiends you, there will be enough hype and rage surrounding the technology that will support it at that time to satisfy your appetites for faster window switching. I'm sure there's alot of you that have no idea where to start utilizing cpu's to their potential, that are also complaining how useless this is.

Score: 0

By M.Sweazey

edited Jun 26, 2006 - 4:58 PM

Score: 0

By PC Rat

posted Jun 22, 2006 - 10:33 AM

...

Won't help desktop users ...sub-systems couldn't
begin to keep up.

They don't call it a "computer ~system~" for
nothing !

...

The Computer Rodent

...

Score: 0

By flake

posted Jun 22, 2006 - 9:49 AM

I believe that the correct news is that IBM created a *TRANSISTOR* that oporates at .5THz, not a chip. That's a pretty big difference.

Original story here: http://tinyurl.com/zus5w

Score: 0

By Muggs

posted Jun 22, 2006 - 11:33 PM

They were clearly talking about a " chip ":

"For the first time, Georgia Tech and IBM have demonstrated that speeds of half a trillion cycles per second can be achieved in a commercial silicon-based technology, using large wafers and silicon-compatible low-cost manufacturing techniques," said John D. Cressler, Byers Professor in Georgia Tech's School of Electrical and Computer Engineering, and a researcher in the Georgia Electronic Design Center (GEDC) at Georgia Tech. "This work redefines the upper bounds of what is possible using silicon-germanium nanotechnology techniques."
Georgia Tech Professor John Cressler and Phd student Ram Krithivasan examine a silicon germanium chip inside a cryogenic test station at the Georgia Electronic Design Center at Georgia Tech in Atlanta. IBM...
Click here for more information.

The accomplishment will be reported in the July issue of the journal IEEE Electron Device Letters. The research has been supported by IBM, NASA, and the GEDC at Georgia Tech.

"This groundbreaking collaborative research by Georgia Tech and IBM redefines the performance limits of silicon-based semiconductors," said Bernie Meyerson, vice president and chief technologist at the IBM Systems and Technology Group. "IBM is committed to working closely with our academic and industry partners to deliver the insight and innovation that will enable a new generation of high-performance, energy efficient microprocessors."

The silicon-germanium heterojunction bipolar transistors built by the IBM-Georgia Tech team operated at frequencies above 500 GHz at 4.5 Kelvins – a temperature attained using liquid helium cooling. At room temperature, these devices operated at approximately 350 Ghz. Performance measurements were made using a specialized high-frequency test system in the Georgia Electronic Design Center.

The devices used in the research are from a prototype fourth-generation SiGe technology fabricated at IBM on a 200-millimeter wafer using an older un-optimized mask set. Simulations suggest that the technology could ultimately support much higher (near-Terahertz) operational frequencies at room temperature, Cressler said.
Georgia Tech Phd student Ram Krithivasan examines a silicon germanium chip inside a cryogenic test station at the Georgia Electronic Design Center at Georgia Tech in Atlanta. IBM and Georgia Tech have announced...
Click here for more information.

"Having a silicon-based technology that is compatible with low-cost IC manufacturing – while still providing these extreme levels of performance – allows us to envision integrating these devices into systems that would be affordable for emerging commercial markets as well as defense applications," Cressler said.

The next step in this research will be to understand the physics behind the silicon-germanium devices, which display some unusual properties at these extremely low temperatures.

"We observe effects in these devices at cryogenic temperatures which potentially make them faster than simple theory would suggest, and may allow us to ultimately make the devices even faster," said Cressler, who heads the world's largest university-based silicon-germanium research team at Georgia Tech. "Understanding the basic physics of these advanced transistors arms us with knowledge that could make the next generation of silicon-based integrated circuits even better."

http://www.eurekalert.or...6-06/giot-gtt061706.php

Score: 0

By rijp

posted Jun 22, 2006 - 11:35 AM

A transistor is a component. Sometimes known as an IC, or integrated chip. The chip includes the circuit board, but the transistor is part of the chip.

so technical, maybe they did just create a transistor, but that transistor becomes the key component in a "chip". To the avg user its just easier to call it a "chip"

Score: 0

By Gerwin

posted Jun 23, 2006 - 3:40 AM

OMG, another one. Listen guys, just because 'the average guy' understands the word 'chip' better than 'transistor', that doesn't make it true.
It was about a transistor that can switch itself at 500GHZ. This has nothing, really nothing to do with a chip that runs at 500 GHZ.

Score: 0

By minus_seven_fold

posted Jun 22, 2006 - 11:37 PM

oh god some of us were told that in HS, but you think your better than most so...

Score: 0

By Babylon2x

edited Jun 22, 2006 - 9:59 AM

Transistor? That certainly is getting more interesting, seems quite a few have misrepresented this including the BBC news article.

I can't help but feel most of the rest of the comments here (but not all) have missed the point. Yea, PCs have other bottlenecks for now, and yea they super cooled it. But it ran extremely fast at room temperature also. The other things will improve to catch up no doubt. But without a super fast CPU, why would they even bother trying to catch up?

Now a super fast CPU might be reality, I'm sure the other potential bottlenecks will become a more serious issue to research and improve.

Everything in PCs have come a long way since the humble beginnings, developments like this are going to be extremely important.

I certainly hope some day I'll have a PC running at 350Ghz some day, so I'll dare to dream which is probably how the rest of our PC improvements came along.

If they're talking about using these in wireless devices, cellphones, the price obviously isn't going to be too far out of this world.

Score: 0

By Straspey

posted Jun 22, 2006 - 5:52 AM

My first PC was a 486sx which ran on a 25Hz chip with 4 Megs of RAM. Back then (the mid-80's) HD's sold for the equivalent of about $1 per MEG and the concept of the Ghz chip was considered futuristic.

Would you also like to hear about the time when color TV was new, there was no such thing as cable, touch-tone phone service was a costly luxury and there were no answering machines...which meant after ten rings you just realized that nobody was home and hung up the phone & went outside to interact with other people in "real time" ?

Score: 0

By seier

posted Jun 22, 2006 - 6:41 AM

Hi Straspey,

You might want to know that it was 25Megahertz not Hertz. Your first processor was a million times faster than you thought it was, and ironically, yes it was still slow :).

Cheers,
Christian Blackburn

Score: 0

By Straspey

posted Jun 22, 2006 - 8:27 AM

Yes...thank you for that.

I may be old, but not *that* old. :)

Score: 0

By ingram091

edited Jun 22, 2006 - 3:51 AM

yea you'll need it for the new Windows over the next several years... Remember Vista is a huge resource hog, and is merely a stepping stone for the Next Gen OS that MS is working on for 2009/2010.

Even their reps say the only reason Vista is coming out is to fund its further research and development in what MS once code named Blackcomb, but is now called Windows Vienna... Trust me. THIS improvement in chip performance WILL be needed...

Score: 0

By Muggs

edited Jun 22, 2006 - 11:35 PM

Have you actually ever used MS Vista? I thought not. I installed it (beta) on the same machine that formerly ran Windows XP. It is at least as quick and responsive as as XP.

Score: 0

By hotfoot13

posted Jun 22, 2006 - 12:59 PM

It's more than needed. This along with a minimum of 256GB of RAM will be needed to run Vista with all the bells and whistles.

Score: 0

By Galway

posted Jun 22, 2006 - 3:00 AM

("Understanding the basic physics of these advanced transistors arms us with knowledge that could make the next generation of silicon-based integrated circuits even better," said John D. Cressler, Byers Professor in Georgia Tech's School of Electrical and Computer Engineering.)

LOLage at your post, how the hell do you thing things improve and get better ?

Score: 0

By crazeebob2000

posted Jun 22, 2006 - 2:16 AM

500GHz? Psshht! Is that all?

Score: 0

By rijp

posted Jun 22, 2006 - 10:34 AM

hahaha.. You just can't make some people happy.... : )

Score: 0

By mshulman

posted Jun 21, 2006 - 8:15 PM

So, maybe if your computer is running slow, just set your AC to the lowest setting - maybe that will speed it up ;)

Score: 0

By Kramy

posted Jun 22, 2006 - 1:20 AM

:P

Actually, more ghz almost always sucks more power, so that would likely slow it down. You could stick it in your freezer though... ;)

Score: 0

By BDU33

edited Jun 21, 2006 - 8:02 PM

This is actually a DARPA MTO Program that was reported at the August 2005 Symposium.
www.darpa.mil/darpatech2...tations/mto/pappert.pdf

You won't see any benefit from this until InP or GaN based processors that do not utilize a bus.

The really neat stuff that you'll hear about in the future will be RF Photonic integration into semiconductors. There will no longer be a dividing line between silicon and photonics.

Score: 0

By paulm

posted Jun 21, 2006 - 6:51 PM

Sounds good in theory. But they have to overcome all the other bottlenecks in a conventional computer system before these CPU speeds become useable. That will be the biggest challenge!

Score: 0

By Kramy

posted Jun 22, 2006 - 1:17 AM

Regardless, at 500ghz you will have 100% of 10ghz available to 50 or more applications, which should mean no slowdown ever. If these processors end up cheap, I'm sure nVidia will strap 80 memory slots on a mobo and connect them all to the CPU. :P

Score: 0

By jbaltz69

posted Jun 21, 2006 - 6:24 PM

I've got this chip in my cell phone right now, I can stream digital porn SUPER FAST! Thanks IBM.

Score: 0

By bigsexy022870

posted Jun 21, 2006 - 6:06 PM

It really severs no purpose to have such a experiment. It's not like this chip will ever be sold. It's just braging rights and to be honest not much of that. Since it's not gonna sell chips cause it's just a toy. They should focus there efforts on making faster commerical chips instead of having fun.

Score: 0

By rijp

posted Jun 22, 2006 - 10:39 AM

Tell me you aren't this stupid. Ever heard of AS/400 there, mr living in the dark ages? This chip may not see the light of day on YOUR computer, because you are obviously too stupid to use one, but IBM has a major line of hardware that pretty much runs like 80% of all the manufacturing and retail companies like walmart and home depot.. Their chips are probably close to this speed, and since they can manufacture a chip this fast and test it, they will be putting it in FUTURE AS/400 machines.

So when they do experiments, it presents the possibility, obviously it can be done, now they will set forth and manufacture them on a large scale.

You must be a complete idiot, testing is ALWAYS done first to see what problems you will run into, how hot things will get, what happens when you run software that fast. They have to test rocket ships into space, which could lead to other advancements, but just because your car may never have a rocket, doesn't mean you can't benefit from TESTING an alternate product, you stupid fool.

Score: 0

By M.Sweazey

edited Jun 26, 2006 - 4:56 PM

Score: 0

By Marticus

posted Jun 22, 2006 - 1:37 AM

And 20 years ago they were working on those silly silly 2Ghz chips too.... what a stupid idea, I mean by testing the CPUs early they can only increase the speed in which they can start production.

And furthermore, stop trolling.

Score: 0

By NULLedge

posted Jun 22, 2006 - 12:51 AM

ibm is a super computer company. ever think they might be using it in those?

Score: 0

By The-One

edited Jun 21, 2006 - 7:15 PM

Wow, I hate to say this, but this comment is not showing much intellect. How do you think we advance science or technology? "Oh, lets just increment this a bit here and there?"

Today's chips were once 'toys' and fantasy as well.

Score: 0

By mfarmilo

posted Jun 21, 2006 - 6:39 PM

I'd hardly call this 'having fun'. The only way commercial chips progress is through research like this. I fully expect research like this will eventually produce real-world chips that can blow the socks off anything we have today.

Score: 0

By altern8energy

posted Jun 21, 2006 - 4:39 PM

you know I don't care if they come out with a 100Ghz AMD or Intel CPU for consumer use. Most PCs have a bottleneck with the storage medium. When I power up my AMD Athlon64 X2 4400 machine with an ATI 1900XT card and 4GB of RAM, it still takes too long for the damn harddrive to bootup, even if you have a RAID. Now if they want to comeout with a SSSD (Solid State Storage Device) with DDR3 memory that is equal to current large harddrives and inexpensive to own, now that would interest me. With today's machines, one should beable to hit the power button and within a few seconds have it completely booted up from a cold state and ready to use...

Score: 0

By Muggs

posted Jun 22, 2006 - 11:48 PM

Spend a little bit of time reading about Vista and you will see that Microsoft has a hardware certification planned in the works that includes a hybrid hard drive/nonvolatile RAM hard drive specification that will permit superfast boot up's and shutdowns. Sheesh, some people who post about technology seem to believe that nothing is ever going to change. It's so weird.

Score: 0

By Marticus

edited Jun 22, 2006 - 1:44 AM

I don't care if my PC takes 20 30, or even 60 seconds to boot.

What i really care about to overall power and design. Now increasing bandwidth between storage and CPU is important, but that is always being worked on.

My 2 cents

Also, I think you should check this out.
http://video.google.com/...3709&q=fast+boot+pc

See how fast that is? It's called a I-RAM, and you can even purchase it.
http://nvnews.pricegrabb...d.php/masterid=11596387/

So, have fun with this wonderfull technology that cost money, I however will pay less and wait that whole 60 seconds for my PC to boot.

Score: 0

By swattz101

posted Jun 22, 2006 - 1:35 AM

I could be wrong, but it is my understanding that eventially EFI (replacement to bios) will eventually be set up to load the kernal and real mode drivers and most of what is needed to run the os, and with programs loaded into RAM, your hd will just be for storage.

On a side note, I remember having a computer years ago (386 or 486 age) with lots of ram. I set up a ram-drive, and attempted to set up a batch program to dump the entire contents of my huge 30MB HD into the ramdrive at startup to run the computer quicker. I never was succesful, but I didn't have the time to put to much work into it.

Score: 0

By M.Sweazey

edited Jun 26, 2006 - 4:56 PM

Score: 0

By NULLedge

posted Jun 22, 2006 - 12:49 AM

hows the weather way out there in left field?

Score: 0

By wav

posted Jun 21, 2006 - 6:00 PM

'Hibernation' does VERY well for few seconds from cold to ready to use.. Try it. I use it for my laptop and it takes only about 6 seconds to load to the desktop. I never reboot it unless something for installs (such as drivers and/or security updates) call for it, so there shouldnt be any worries about it slowing down over long periods of not rebooting.

Score: 0

By Kramy

posted Jun 22, 2006 - 1:11 AM

My Win2k install with XP Home ntldr and NTDETECT.COM files can unhibernate in 1-2 seconds, plus another second for the screen to change resolutions.

With a RAM-drive or Hybrid-HD, I'd be in bliss!

Score: 0

By rijp

edited Jun 21, 2006 - 4:49 PM

First of all, RAID is dependent on not only speed of the drives, but the NUMBER of drives.

I am guessing you don't have SCSI, so therefore you are not running the fastest RAID. Secondly, the more drives you add, the speed INCREASES by a fair amount with each 18K RPM drive.

Hard drives are slow, but they are quick enough to keep up with demands of software.

Secondly, CPU runs programs, not the hard drive. Once the program is loaded into memory, given the right amount of RAM, it almost NEVER needs the drive again.. unless you save stuff, so NO its not going to depend on the hard drive.

Also, they have FLASH media they are working on to store the OS on, so it does have instant boots.. this was developed over 10 years ago.. but it was too expensive.

Score: 0

By bigsexy022870

posted Jun 21, 2006 - 6:02 PM

Hard drives are still one of the major bottleneck. It just can't handle the speeds. This is true regardless of how many drives you strip together.

Score: 0

By M.Sweazey

edited Jun 26, 2006 - 4:56 PM

Score: 0

By rijp

edited Jun 22, 2006 - 10:47 AM

Dude any comment after that "Since it's not gonna sell chips cause it's just a toy" lack of intellect post, I am not going to acknowledge ANYTHING you have to say.

I have said some really dumb things on here, but to say that this chip will never come about ESPECIALLY when they are TESTING it RIGHT now, is totally barbaric and idiotic.

you are a moron, and will dismiss EVERY one of your comments with conviction.

You are probably the dumbest person I have seen on here to date. And that's saying a lot.

Score: 0

By RootWebGod

edited Jun 21, 2006 - 8:11 PM

You apparently did not read or understand the point; once the program is loaded into memory the actual processing can be done with the CPU and RAM, independent of the HDD.

For starters, you can get a significant speed boost by upgrading to a relatively affordable Western Digital 10K RPM SATA drive with 16 MB cache. (There are also 74 GB and 36 GB editions for those on a more limited budget.)

See: http://www.wdc.com/en/pr...roducts.asp?DriveID=189

Use PriceGrabber and search for the drive's model number to find a great price.
http://www.pricegrabber.com/

Score: 0

By Muggs

posted Jun 22, 2006 - 11:54 PM

I have one of those 10,000 RPM Raptors-generally makes a difference. I bought mine about two years ago. I have read, but not confirmed, that the new Barracuda approaches the Raptor performance.

Score: 0

By Kramy

posted Jun 22, 2006 - 1:13 AM

And these help me get into windows 2 seconds after I hit the power button, how?

Score: 0

By Banquo

posted Jun 21, 2006 - 4:38 PM

"To accomplish this feat, the researchers froze the chip to near absolute zero -- minus 451 degrees Fahrenheit."

Oh yeah, what a great accomplishment. I'm sure Dell will be selling 500GHz systems any day now because we all have cryogenic cooling systems in our computers anyway. No big deal...

Score: 0

By Paradise-FH-

posted Jun 21, 2006 - 4:52 PM

in other news ... cold fusion was accidently achieved while attempting to run a 1THz processor.

Score: 0

By rijp

posted Jun 21, 2006 - 4:43 PM

You people never cease to amaze me! READ!

*At room temperature, the chip operated at 350 GHz, slightly slower than the record but still over 87 times faster than the fastest commercially available chips*

Score: 0

By minus_seven_fold

posted Jun 22, 2006 - 11:38 PM

OMG!!!

Score: 0

By Noremacam

posted Jun 21, 2006 - 6:45 PM

Ditto!

Score: 0

By Babylon2x

posted Jun 21, 2006 - 5:31 PM

I agree with rijp, amazingly. Okay, so it needed to be really cold to run at 500Ghz, but the most important part to think about is how fast it can run at room temperature. I also read on the BBC website they hope to run it up to 1Thz at room temperature eventually.

http://news.bbc.co.uk/1/hi/technology/5099584.stm

"The team believes it is possible to make chips run at 1,000 Ghz, or one Terahertz, at room temperature."

Score: 0

By Paradise-FH-

posted Jun 21, 2006 - 4:54 PM

yes but to post that extra 150GHz that put it way over the record (WHICH IS WHAT THE ARTICLE IS ABOUT) they froze it colder than a witch's ...

Score: 0

By Tenoq

posted Jun 21, 2006 - 7:20 PM

And you don't think the previous record was set with the same sort of cooling? :P

Score: 0

By rijp

edited Jun 21, 2006 - 5:00 PM

*. . . than a witch's ...*

axe? man? CPU?

Score: 0

By Vegitable

edited Jun 21, 2006 - 7:34 PM

What's with all the HardDrive crying? Did we forget about the lame-o bus speeds?

Score: 0

By bane73

edited Jun 22, 2006 - 9:04 AM

Man, I just don't get some of you people and the way you guys think (or don't!) -- "lame-o bus speeds"???? Stop and just think for a minute... lame-o bus speeds are merely due to bottlenecks inherent in the design and due to the bus PROCESSOR speeds... the first part of the problem can be solved similarly to adding more lanes to a freeway or adding a whole new additional freeway to a city's infrastructure... fairly easy to do... the second part of the problem, the bus's processor speed, is practically automatically solved by the newly developed faster CPU processor so that's hardly even a problem to "solve" anymore, just one to be implemented...

Lame-o bus speeds??? Hardly. Now that the CPU has been developed, bus speeds will be one of the things that will benefit from the new development and will naturally become better...

... some of you guys amaze me... why do you even participate in this forum anyways, you either don't appreciate and understand the subject matter well enough, or you are too lazy to read and think...

Score: 0

By Vegitable

edited Jun 22, 2006 - 9:30 AM

....yes, some of YOU are amazing.

No where did I say that increasing bus speed was a "technical hurdle yet to be accomplished", however, I did say that bus speed is "lame-o", which it most certainly is.

Take for example the bus speed for DDRx vs PCI or eve PCI-E. Based on your response, it would seem increasing bus speed accross all channels is easier than increasing specific channels while leaving the rest alone.

And yes, Lame-o bus speeds when you concider a processor operating at 3Ghz with data feed at a whopping 33Mhz or 66Mhz.

Anyhow, flame on little one.

Score: 0

By bane73

posted Jun 22, 2006 - 1:15 PM

.... ok, I apologize for the flame... probably a little overboard, I admit... that being said, throwing acronyms around doesn't impress either... it's "the other side of the coin" of flaming...

Your reply is a bit outdated and assumes much. My systems FSB is 800Mhz right now, hardly lame-o considering my processor is only 2Ghz... and yes, while PCI is only 66Mhz, the assumption is that currently there is a huge *NEED* for anything much faster... if there were, it would be developed by now since other bus architectures exist that go beyond 66Mhz... take for example, current FSB's available... those are still buses, and if they can achieve rates greater than 800Mhz, there is really no reason that a redesign of PCI couldn't do the same thing... it's all based on market needs and pricing...

Score: 0

Nov 21 - 6:51 PM ET Coalition urges better laws for e-mail and cell phone privacy

Nov 21 - 6:42 PM ET How is Internet advertising faring in this bad economy?

Nov 21 - 6:15 PM ET From out of the Amazon Cloud, an AWS winner emerges

Nov 21 - 5:20 PM ET Mufin music finder enters public beta, adds widgets

Nov 21 - 5:01 PM ET Google launches its SearchWiki semantics plug-in

Nov 21 - 4:57 PM ET Asus previews a slimmer, less costly Eee PC

Nov 21 - 4:32 PM ET New Adobe Media Player ushers in AIR 1.5

Nov 21 - 2:11 PM ET AT&T to add Pantech's low-cost C630 phone to 3G lineup

Nov 21 - 1:11 PM ET iPhone gets firmware 2.2 update

Nov 21 - 11:51 AM ET The Dell surprise: Higher earnings on lower revenue