http://www.msnbc.msn.com/id/13434023/

SAN FRANCISCO - IBM has built a transistor that runs about 100 times faster than current chips, a development that could pave the way for ultra-fast computers and wireless networks, the computing giant said Monday.

IBM achieved the record speeds by building one from silicon laced with exotic chemical element germanium.


The transistor achieved a speed of 500 gigahertz, which is more than 100 times speedier than the fastest PC chips sold today.

That speed was hit only when IBM researchers, working with counterparts from the Georgia Institute of Technology, cooled the transistor to near absolute zero, but Meyerson said the device still ran at 300 gigahertz at room temperature.

http://www.msnbc.msn.com/id/13434023/

SAN FRANCISCO - IBM has built a transistor that runs about 100 times faster than current chips, a development that could pave the way for ultra-fast computers and wireless networks, the computing giant said Monday.

IBM achieved the record speeds by building one from silicon laced with exotic chemical element germanium.


The transistor achieved a speed of 500 gigahertz, which is more than 100 times speedier than the fastest PC chips sold today.

That speed was hit only when IBM researchers, working with counterparts from the Georgia Institute of Technology, cooled the transistor to near absolute zero, but Meyerson said the device still ran at 300 gigahertz at room temperature.

Hello!

Very nice!

But I don´t want to know the price of this... :-)


Greetings!

:-)

---

Germanium exotic???
The *first* transistors were made of germanium!
Even before that, the original "crystal" radios used germanium crystals!
There's nothing exotic about germanium. Maybe in the way they use it, but sure not in the material itself! Talk about going back to the roots! Haha.

---

Jim

Some people plan their life out and look back at the wealth they've had.
Others live life day by day and look back at the wealth of experiences and enjoyment they've had.

The transistor achieved a speed of 500 gigahertz, which is more than 100 times speedier than the fastest PC chips sold today.

Humm... if it was a single transistor, chances are they are quoting an f-sub-t (sorry, I don't know how to display it properly using BBCode). That is a unity gain cutoff frequency of sorts, commonly used to characterize RF transistors. I never even knew that parameter for my transistors when I was designing microprocessors at Intel.

If true, I rather suspect current microprocessor transistors have the correspondig parameter considerably more than one one-hundredth of that, still more so if one compared them at the same temperature.

Given current trends in microprocessor design, one of the very first questions some one interested in utility would ask would be about leakage. (hint--at the time of the germanium to silicon transition, leakage was one key silicon advantage).

---

Germanium exotic???
The *first* transistors were made of germanium!
Even before that, the original "crystal" radios used germanium crystals!
There's nothing exotic about germanium. Maybe in the way they use it, but sure not in the material itself! Talk about going back to the roots! Haha.

Hehe, well if you compare it to... I don't know, copper, germanium is way more exotic and harder to come by. ;)

---

/Mav

We have lingered long enough on the shores of the cosmic ocean.
We are ready at last to set sail for the stars.
(Carl Sagan)

If I remember correctly, germanium transistors melt down at a much cooler temperature than silicon ones. If we use those we'll really have to watch the cooling.

---

IBM dind't disclose what kind of chip they tested, but I think it was a rather small one.
Anything silicon based that runs at 500GHz will have a horrendous power consumption per square mm...

Regards Hans

---

Actually, according to this report it's much more impressive.

They're running a whole multi-junction chip at 500GHz, not just a single transistor.

Admittedly it's liquid helium cooled, but they suggest 350GHz is possible at normal temperatures.

And just for fun a 1,0000 GHz (1 Tera-hertz) chip is on the cards !

Looking forward to see the WU crunch times ;)

---

As to the heat question in a germanium based circuit, the chip would probably generate less heat than a silicone based one. Silicon junctions have a forward voltage drop of 0.6 to 0.7v while a germanium junction has a drop of ~0.2v.

Here is a chart of various junction materials

Semiconductor   Junction forward  Electron     Hole         Max Junction
material 	voltage           mobility     mobility     temperature
                V @ 25 °C         m/s @ 25 °C  m/s @ 25 °C   °C
Ge 	        0.27              0.39         0.19         70 to 100
Si              0.71              0.14         0.05         150 to 200
GaAs            1.03              0.85         0.05         150 to 200
Al-Si junction 	0.3                —            —           150 to 200

This would mean that an identical junction made from germanium would create less heat due to resistance losses than a silicon junction (<1/2)
However as Clyde mentions the max temp of a pure ge junction is much less than silicon which is probably why they use liq he cooling! Not to mention that the resistance of the junctions will be less at the lower temps.

But another thing appears in the chart above, the speed of a germanium junction is much slower than silicon. So evidently the researchers have developed a junction that combines the low voltage drop of a germanium junction with the speed of silicon! Very impressive!

---

Jim

Some people plan their life out and look back at the wealth they've had.
Others live life day by day and look back at the wealth of experiences and enjoyment they've had.

now if we could only get Ge like low voltage and GaAs like speeds out of a material... Zzooommmm!

~20 years from now, folks will be sitting around moaning that multi-THz clock rates are the limit of IC technology...

...and hopefully folks like IBM will be close to getting a 1QHz IC running at that point.

---

there was something like this on the bcc news site but it was talking about a whole chip not a transistor and it was supercooled with liquid helium. Don't know if its any relevance. I've been trying to find it again but's been pushed off by football related rubish!

When they make a processor are the transistors all in one layer? I understand that there might be two or three per square micron in the current technology. Maybe they could start stacking the layers to get more transistors on a single chip. But they might just have to use that liquid helium to keep that cool. With the germanium or silicon-aluminum junction maybe power dissipation might drop as the square of the voltage, but I don't know. It depends on the resistance, too.

---

now if we could only get Ge like low voltage and GaAs like speeds out of a material... Zzooommmm!

~20 years from now, folks will be sitting around moaning that multi-THz clock rates are the limit of IC technology...

...and hopefully folks like IBM will be close to getting a 1QHz IC running at that point.

Yep, I remember not too long ago an issue of a popular computing magazine had a picture of a 486 chip on the cover with the big headlines "fastest computer ever - 50 MHZ!" and in the article it was saying that it would be hard to beat this speed! Ha! It's not the limits of the chip but the limits of our knowledge *at that time*. As technology (meaning our knowledge and application of that knowledge) improves we will see more amazing things than we can even imagine right now! When the article I mentioned was written, who could imagine that we would have 3+ Gigahertz processor speeds and all of the other improvements that have come into being since then?

---

Jim

Some people plan their life out and look back at the wealth they've had.
Others live life day by day and look back at the wealth of experiences and enjoyment they've had.

When they make a processor are the transistors all in one layer? I understand that there might be two or three per square micron in the current technology. Maybe they could start stacking the layers to get more transistors on a single chip. But they might just have to use that liquid helium to keep that cool. With the germanium or silicon-aluminum junction maybe power dissipation might drop as the square of the voltage, but I don't know. It depends on the resistance, too.

Ah, the old dream of 3D IC's.

You noted the cooling problem right away, and there is a way around it. Just not using the manufacturing techniques we currently have.

When people talk about the semi-conductor industry "hitting limits", the biggest limit we've really hit is in =manufacturing=.
For 40+ years we've basically been etching IC into substrates using photographic like techniques. Photo-lithography has been a wonderful workhorse, but it has just about run its course.

At some point in the next ~20 years we are going to have to transition from =etching= IC's to =building= IC one molecule or atom at a time using nanotechnogy assembly methods.

When that happens, 3D ICs will become much easier to make.

...and how will we cool them? IMHO, part of the answer will be in the physical structure we give them. Instead of circuits "pancaked" on top of each other, I envision a thin layer of IC surrounded by space and connected to other thin layers like the walls of bubbles in foam.
Picture one of the shuttle heat tiles shrunk down to the size of a small pill made mostly of space and whose connective "tissue" is not the Si of the shuttles heat tiles, but rather filaments of IC.
Such a "IC foam die" will allow us to finally build 3D circuits that can be cooled in a reasonable manner.

...but the enabling technology is nano-assembly rather than photo-lithography and we are aways from having it.

But it =must= come.
2006 is the year of 65nm. 2008 45nm. 2010 32nm. 2012 22nm. 2014 16nm. 2016 11nm. Even if PL can do all that, sometime in that time period nano-assembly will probably become more economically viable than PL.

Another factor is that all this multi-core stuff we are doing is going to make it hard to keep dies small and that creates other pressures on the economics of ICs in general and PL in specific.
Sometime ~2010 the semi-conductor industry will go from using 300mm wafers to 450mm wafers. Usually it takes ~10years for wafer size to increase in this way.
Unfortunately, multi-core CPUs get big fast. And that means either we start increasing wafer size faster than 1/10 years, say to ~1/5 years or so, or IC costs start going through the roof, or that we invent a new way to make ICs.

The other implication of all this is that the present love affair w/ high IPC multi-core CPU design is just another fad. Just as the previous love affair with clock rate was. The reality is that at some point we are going to =have= to start increasing clock rates at a brisk pace again. On this subject, my take is that when 16 cores fit on a die and each of those cores is doing 16 IPC, we will "hit another wall". If not sooner. At that point we had better have solved our present heat problems because we are going to need serious clock rate increases to keep increasing performance at the 2x every 18-24month pace we've gotten used to and call "Moore's Law".

---

there was something like this on the bcc news site but it was talking about a whole chip not a transistor and it was supercooled with liquid helium. Don't know if its any relevance. I've been trying to find it again but's been pushed off by football related rubish!

this?
...as referenced below ;)

---

@MikeSW17 - Thats the one. Thanks

@EricVonDaniken

Theres also carbon nano tubes that can be made using methods today already but not easily controllable and you'd have to figure out how exactly to use them.
All be much easyer when we finaly find/develop a cheap material that is superconductive at close to or at room temp. Also be good for the electricity bills ;)
or we could go down the route of organic components.
Then theres always the biggy- quontum computing :)

Then theres always the biggy- quontum computing :)

Moore's law of quantum computing: one more bit every 2 years :o)

Regards Hans

---

Then theres always the biggy- quontum computing :)

Moore's law of quantum computing: one more bit every 2 years :o)

Regards Hans

Isn't that "Intel's Theory of Marketing" or "M$'s Theory of SW Bloat"?

---

@MikeSW17 - Thats the one. Thanks

@EricVonDaniken

Theres also carbon nano tubes that can be made using methods today already but not easily controllable and you'd have to figure out how exactly to use them.
All be much easyer when we finaly find/develop a cheap material that is superconductive at close to or at room temp. Also be good for the electricity bills ;)
or we could go down the route of organic components.
Then theres always the biggy- quontum computing :)

All that stuff is far easier to do if you assemble your IC than etch it.

As for where you use C-NT's: as replacements for wires.

---

Maybe the principles of electron microscopy could be applied to CPU manufacture. I think they coat the specimen then somehow get the electrons to discriminate features. Years ago the resolution of electron microscopes was about 10 to 20 angstroms (one to two nanometers). If it gets much better it'll be possible to see individual atoms. Maybe the electrons could be made to carve features and make the CPU.

---