2 GHz clock from 2, 1 GHz clocks

You would need to multiply them together with them being 90' out of phase with an xor gate or something, you might as well generate the second clock with a 90' phase shifter. The other option is to use a frequency multiplier.

Colin =^.^=

Yes, by XORing them together. You don't need 2 separate clocks, just one is enough. You delay it by 90 degrees (0.25 ns) and feed the original & delayed signal to the XOR. But at such high frequencies there may be practical issues to deal with...

If it is possible then, what is stopping us from having THz/TFlops microprocessors?! Please, don't say "speed of light".

Speed of dark, then?

martin

Maybe that Intel/AMD/... haven't hired you yet.

Engineers waste too much time being distracted by idiots posting on usenet from google groups slowing down the design of such microprocessors.

Do the computing world a favour and shut up - ok?

The world would be a better place if people could actually tell the difference between a design question and basic questions, which belong in sci.electronics.basics

You're asking the wrong question.

If you'd asked "How can I get a 2GHz signal from a 1GHz signal" then the way to do it is multiply the signal with a frequency multiplier. THere is nothing new about them. And then you'd actually get a 2GHz signal that was traceable to the 1GHz signal, rather than some bastard version from two different signals that may not even be on the same frequency.

MIchael

It's that the transistors inside the cpu that limits how fast you can clock it, not how high a clock rate you can put on it.

No, you can't do it because you can't guarantee the phase relationship between the two clocks. Clocks speed up and slow down over time and environment. Two clocks won't react exactly the same, therfore you can't form a hard and fast relationship needed to generate a pure 2 GHz clock.

---Matthew Hicks

phase

clock

multiplier.

well the same limitations that apply to all electronic circuits, such as unwanted capacitance resistance inductancce, charge carrier transit time etc this limits the max frequency. its probably easier to make a 2ghz oscillator anyway.

maaybe one day we will have thz cpus who knows ? or even the long awaited quantum computer.

Colin =^.^=

GHz microprocessors do not receive GHz clocks. They already have internal clock multipliers right on the chip, and actually are driven from much lower frequency clocks. There's no reason why the clock multiplier can't multiply up to 3 GHz, or 12 GHz, or 999 GHz, except that the logic of the chips can't work that fast.

In other words, the clock ain't the problem.

John

Well, OK, what IS the speed of light? ;-) And "please, don't say "3e9 meters/second."

Go take a course in semiconductor physics and come back and post the answer to your question. It isn't light that's running around inside the semiconductor devices, for the most part.

At some level, "resistance, capacitance, inductance and electron drift velocities" might be a decent answer, but from your question, I suppose you'll need some time to think about that to fully understand it. But if you base your processor on some totally different technology, perhaps you can get away from those limiting factors. Many folk are working on doing that in various ways.

Cheers, Tom

erm ok 3e8m/s

whatever you use to transmit the bits it cant travel faster than c in transmision lines its in the order of 0.7c I assume silicon is similar, but semiconductor physics does my head in, this means clock speeds are limited by the dimensions of the device as much as anything else, especially when synchronous clocks are spread out over a large chip.

If asynchronous logic was used it wouldnt matter about clock skews from one part of the chip to another, the data would be transfered with handshaking signals precisly when its available, rather than waiting till the next clock edge.

I gues this is unpopular becuse so much store is placed on clock frequencies for marketing and for some reason asynchronous logic always seems to be frowned upon even when it makes for a simpler soultion, its suposedly more deterministic, at least untill you overclock it anyway.

I wonder, will phemts find there way into logic chips ?

Colin =^.^=

other replies

Sure, I was lost my mind, but there must be an easy to for you guys to point out that. If it is annoying there are some other ways too, I mean killfile. As you would have guessed, there could not be any possible useful contributions from me. I could only contribute through (dumb?!) questions like that.

I keep on posting here, if I think it is SED. Soon, I will learn to distinguish between SED/SEB.

This thread ends here. :)

Why over-engineer it with 2 clocks?. Any clock has synchronous 'twice frequency' features already built in.

Multiplying 1GHz to 2Ghz with any regular gate, even with the fastest one, will don't work. But the warm water was already discovered... :)

best, Vasile