Frozen Chip from IBM hits 500 GHz

sideshow2004 writes "EETimes is reporting this morning that IBM and Georiga Tech have demonstrated a 500 GHz Silicon-germanium (SiGe) chip, operating at 4.5 Kelvins. The 'frozen chip' was fabricated by IBM on 200mm wafers, and, at room temperature, the circuits operated at approximately 350 GHz."

Comments

Re:I RTFA..

[peragrin]

Several cell phones just run at real time. So they really do run at 2.4 ghz for the signal processor, while the system itself is on another chip at a different speed.

REmember even though it's running at 2.4 ghz it's extremely dedicated and doesn't produce a lot of heat.

Why?

[reset_button]

It's interesting, but wouldn't it be better to just use two of these chips at room temperature, rather than spend time/money/space on cooling the chip to 4.5 Kelvins?

Can these these chips do any calculations?

[the_humeister]

Or have they just been fabricated to demonstrate that they can attain high GHz rates?

Re:Can these these chips do any calculations?

[EvanED]

at light speed, an electron wouldn't have enough time to make it through the long circuit paths before the next clock cycle.

It doesn't need to go through the long circuit path...

In fact, signals haven't gone through a whole path since (at the latest!) the 286. The processing is already divided into stages, and it only passes through one stage in each clock cycle. (Look up pipelining.)

It would be theoretically possible to design a chip that operated at a lot higher clock speed just by making the stages shorter.

Think of an old fire fighting bucket brigade. If you have one person carry the bucket from the source to the fire, you're gonna have a hard time getting control. If you add people, at some point you can add people until everyone just passes the buckets down the line without moving. If you continue to add people, the buckets will probably not move a lot faster, but you'll have more buckets "in flight" at any given time. Note that the time it takes any given bucket to get to the fire has actually INCREASED because there's overhead in the handoffs and everyone isn't synchronized, but you're gonna get a lot more water on the fire than if you just had one person. You'll also see a bucket being thrown onto the fire much more frequently.

In some sense, the buckets are like instructions, the firefighters are like pipeline stages, and the frequency with which any given person changes buckets is like the clock speed. In a processor, you can add more stages to your pipeline and make it so that each stage has each instruction for less time. There is a limit to the minimum time, just as there is for the firemen (you at least have to grab the bucket from the person before you and let go when the person in front has it, and adding more people to the line won't help at all with that overhead), but the limit isn't the time from fire hydrant to fire.

Re:I RTFA..

[Tim C]

I think that's the point. Reading between the lines, this isn't about general-purpose CPU chips, this is about specialised signal processors. In other words, don't expect to be buying an Intel or AMD chip running at 30+GHz anytime soon.

Re:I RTFA..

[not already in use]

I didn't read the article, but people don't seem to be making a big deal out of the fact that they are comparing the frequency at which a cellphone transmits data to the clock speed of a processor.

Re:THAT WASN'T THE POINT

[nonlnear]

Exactly.

By finding the last point on the temp/speed curve, they are able to much more accurately determine the entire curve. i.e. It's a lot easier to interpolate to more realistic cooling levels. And it makes for a cool headline too.

Uberistor?

[Lord of Hyphens]

Hrm... a batch of transistors that'll relay at clock speeds of 350Ghz. Then they tossed on their P4 cooler and watched it superconduct. Why am I not surprised at 500Ghz? At 4.5K, it's clearly superconducting. And the phone comparison... I like EE Times, but that writer needs to be shot. The editor deserves a slap on the wrists for letting it in (unless they're referring to some strange property of phones). "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,[and absurd cooling that allows us to leverage the properties of superconductivity]" (fixed). IBM: Design it Today, Figure out what the hell we're going to do with it 7 years from Tomorrow. (And yes, I'd get a microprocessor designed with these ubersistors).

Joke/Your Head

[DragonHawk]

You do know Moore's Law relates to the number of transistors on a chip, and doesn't have anything to do with clock speed, right?

You do know that jokes are meant to be funny, and don't have to be factually accurate, right?

Re:Obsolete Units

[Dance_Dance_Karnov]

I'm sure that is the author writing down to his audience. I would have thought the cellphone comparison made that clear.

Re:computers in space

[Skinny Rav]

Since these temperatures only occurs naturally in space, why not build a super, big cluster of these things, hook them up to a satallite and launch it into orbit.

Maybe because heat dissipation in space is poor? I know you can do magic with water evaporation under such low pressure to dissipate heat, but how much water would you need to send up there to provide cooling for reasonable time?

Cheers

Raf

Radiation, most likely

[Opportunist]

Radiation is a big issue for computers in space. Shielding equipment is heavy (=expensive to get up there), and the smaller (and faster) CPU's ICs become, the more susceptible to radiation they become.

There's a reason why NASA is trying their best to get their fingers on ancient CPUs.