# Maximum clock frequency ?

S

#### Skybuck Flying

Hi,

I am beginning to understand how CPU's and chip like that work.

I think the clock frequency is very important.

It's almost like the hearbeat of the computer.

It determines how fast the signals change from zero to one and back.

And these signals are what drives the computer.

So I think a very important question would be:

What is the maximum frequency that these kind of "cpu clocks" can provide ?

Other components should ofcourse also be able to handle these high
frequencies.

Like t_flipflops etc

So what's the best of the best at the moment ?

Bye,
Skybuck.

A

#### ALuPin

Hi Skybuck,

this is a VHDL discussion board ;o)
And not a general information board ;o)

Rgds
André

J

#### jon

The fastest transistor I've heard of operates at 604 GHz.

Cheers,
Jon

N

#### Neo

The fastest transistor I've heard of operates at 604 GHz.

Cheers,
Jon

I thought the Illinois university guys had obtained transistors
operating at 1 Tera Hz .

R

#### Roger Hamlett

Skybuck Flying said:
Hi,

I am beginning to understand how CPU's and chip like that work.

I think the clock frequency is very important.

It's almost like the hearbeat of the computer.

It determines how fast the signals change from zero to one and back.

And these signals are what drives the computer.

So I think a very important question would be:

What is the maximum frequency that these kind of "cpu clocks" can
provide ?

Other components should ofcourse also be able to handle these high
frequencies.

Like t_flipflops etc

So what's the best of the best at the moment ?

Bye,
Skybuck.
Remember also, that the clock frequency is only one 'part' of the equation
giving what the system can do. It is effectively multiplied by the 'width'
of the registers, when dealing with numerical values, and also may be
multiplied internally using either a PLL, or short phase delays, so
multiple operations are generated in each clock. you should really think
about the clock, like the speedlimit on a road, with the width of the
registers, being perhaps like the difference between a motorbike, and a
bus, while the internal phase seqencing, acts like having multiple lanes
of traffic running in parallel. In terms of the actual number of people
the road can carry, you have to multiply all these differences together,
but the clocking limits the minimum time a single person can take to
transit part of the road. An internal PLL, acts like a local 'expressway',
reducing this time.

Best Wishes

H

#### Helmut Sennewald

Skybuck Flying said:
Hi,

I am beginning to understand how CPU's and chip like that work.

I think the clock frequency is very important.

It's almost like the hearbeat of the computer.

It determines how fast the signals change from zero to one and back.

And these signals are what drives the computer.

So I think a very important question would be:

What is the maximum frequency that these kind of "cpu clocks" can provide
?

Other components should ofcourse also be able to handle these high
frequencies.

Like t_flipflops etc

Hello Skybuck,

forget T-fliflops. This is the least important part of an
IC-design. I know you read about it in a VHDL intro book,
but they made this just as an example and not because it's
important.

You should really look first for a FPGA design flow,
e.g. www.altera.com , ..., www.xilinx.com.
After you understand that you can look for full costom ICs.

You can buy divider flipflops with 20GHz.

Best regards,
Helmut

S

#### Skybuck Flying

Roger Hamlett said:
Remember also, that the clock frequency is only one 'part' of the equation
giving what the system can do. It is effectively multiplied by the 'width'
of the registers, when dealing with numerical values, and also may be
multiplied internally using either a PLL, or short phase delays, so
multiple operations are generated in each clock. you should really think
about the clock, like the speedlimit on a road, with the width of the
registers, being perhaps like the difference between a motorbike, and a
bus, while the internal phase seqencing, acts like having multiple lanes
of traffic running in parallel. In terms of the actual number of people
the road can carry, you have to multiply all these differences together,
but the clocking limits the minimum time a single person can take to
transit part of the road. An internal PLL, acts like a local 'expressway',
reducing this time.

Ok so if this is my clock rate:

111111111111111000000000000000

Can the PLL divide it into:

1111111100000000111111100000000 ?

or maybe even further into

11110000111100001111000011110000 ?

I think that's what it could do/does ?

So the PLL creates a faster clock rate inside the chip ?

Would this mean the "normal" clock is outside of the chip ?

How many real clocks or oscilators(=clock?) are there ?

Can the main memory chips and the cpu share the same clock ?

However a transistor is not a clock so my original question is not yet

So what's the fastest clock on todays computers and say 5 years from now ?

For my one bit streaming cpu a PLL might not be interesting...

Since I was more thinking of having all the bits stream
across/through/into/out of the cpu directly from main memory

The question is if multiple bits can be transfered in serial over a single
wire with a slower clock frequency and then still be processed by a cpu with
a faster clock frequency... the cpu would have to buffer some bits I guess.
Or something like that... maybe not possible or maybe it would be
possible with bursts.

However... since I don't like all this complexity... (and I dont like bursts
too ) I am wondering if my simple cpu will need a clock at all

So just for the kick of it

Can VHDL also be used to design/test circuitry without clocks in them ?

Bye,
Skybuck.

T

#### Tim Wescott

Skybuck said:
Hi,

I am beginning to understand how CPU's and chip like that work.

I think the clock frequency is very important.

It's almost like the hearbeat of the computer.

It determines how fast the signals change from zero to one and back.

And these signals are what drives the computer.

So I think a very important question would be:

What is the maximum frequency that these kind of "cpu clocks" can provide ?

Other components should ofcourse also be able to handle these high
frequencies.

Like t_flipflops etc

So what's the best of the best at the moment ?

Bye,
Skybuck.
bookstore or on the web for information on computer architecture?
Trying to get 50,000 words worth of information from random newsgroup
queries is like trying to build a house out of cut-off ends of 2x4
lumber -- and it makes us work at answering questions bit by bit that
are better answered all at once (or at least in much larger chunks).

M

#### Mac

Hi Skybuck,

this is a VHDL discussion board ;o)
And not a general information board ;o)

Rgds
André

Actually, he is cross-trolling to comp.arch, comp.lang.vhdl, and
sci.electronics.design, which is where I'm reading it.

--Mac

B

#### Bernd Paysan

Neo said:
I thought the Illinois university guys had obtained transistors
operating at 1 Tera Hz .

The fastest transistors seem to be carbon nanotubes, and IIRC, they go well
above 1 THz. The main problem to get real figures is probably the
unsuitable equipment to measure these things. In theory, a carbon nanotube
oscillator should be able to drive a 300nm carbon nanotube dipole, emitting
blue light (now that would be something that's not that hard to measure).

J

#### JeffM

I think the clock frequency is very important.
bookstore or on the web for information on computer architecture?
Tim Wescott

In addition, learning to use a search engine
and finding out how to narrow a search
has an instructive quality all its own.

Here's the one for the sci.electronics.* part of the Usenet Archives:

S

#### Skybuck Flying

JeffM said:
In addition, learning to use a search engine
and finding out how to narrow a search
has an instructive quality all its own.

Here's the one for the sci.electronics.* part of the Usenet Archives:

I am getting a error:

"
Server Error
The server encountered an error and could not complete your request.
If the problem persists, please report your problem and mention this error
message and the query that caused it.
"

Maybe only a temporarely problem.

I still haven't seen any answer to my question...

What's the maximum frequency nowadays ?

Bye,
Skybuck.

D

#### Dan Koren

While one should not in general feed the trolls as a
matter of principle, I will make an exception just to
point out that your question suggests you have not
really thought about the problem in any depth.

What matters most for real circuit design is not the
highest frequency that the clock can generate -- but
rather how far can the clock signal be propagated
reliably and without excessive distortion (distance
and fan-out).

dk

X

#### xray

Bye,
Skybuck.

Is something we are all beginning to wish you would take more seriously.

H

#### Helmut Sennewald

Skybuck Flying said:
I am getting a error:

"
Server Error
The server encountered an error and could not complete your request.
If the problem persists, please report your problem and mention this error
message and the query that caused it.
"

Maybe only a temporarely problem.

I still haven't seen any answer to my question...

What's the maximum frequency nowadays ?

Bye,
Skybuck.

Hello Skybuck,

you should see a dozen answers to your question if not then

Again,
You can achieve 20GHz(20e9Hz) with a T-fliflop.

Other insist they can get 100Tflop.
Now I know why you wanted a design with T(flip)flop.
Maybe this "flip" doubles the Tflop.
http://www.hoise.com/primeur/03/articles/live/AE-PL-06-03-5.html
Sorry, I couldn't resist.

A simple processor alone could achieve some GHz, e.g. 10GHz,
but it becomes very expensive and anotrher problem is feeding
such a beast with enough data from memory.
This is then something for an IEEE paper, but nothing which
will make you rich.

Best regards,
Helmut

S

#### Skybuck Flying

Dan Koren said:
While one should not in general feed the trolls as a
matter of principle, I will make an exception just to
point out that your question suggests you have not
really thought about the problem in any depth.

What matters most for real circuit design is not the
highest frequency that the clock can generate -- but
rather how far can the clock signal be propagated
reliably and without excessive distortion (distance
and fan-out).

Yesssss but maybe the propagation length doesn't matter for a small circuit
like my 1 bit cpu =D

So I would still like to know what the maximum frequency is

And in case the circuit is bigger how do circuits nowadays propagate the
clock signal reliably

Bye,
Skybuck

S

#### Sander Vesik

In comp.arch Skybuck Flying said:
Yesssss but maybe the propagation length doesn't matter for a small circuit
like my 1 bit cpu =D

Or maybe it does? But its up to you to find out.
So I would still like to know what the maximum frequency is

There is no such thing. Or very close to no such thing, except possibly
somewhere in UV/r?ntgen border area.
And in case the circuit is bigger how do circuits nowadays propagate the
clock signal reliably

You should first use Google and then ask. Or will you only be sensible when
skybucks^H^H^H^H^H^H^H^Hpigs fly?

S

#### Skybuck Flying

Helmut Sennewald said:
Hello Skybuck,

you should see a dozen answers to your question if not then

Again,
You can achieve 20GHz(20e9Hz) with a T-fliflop.

??

But a T-flipflop is not a clock ???

D

#### Dan Koren

Skybuck Flying said:
Yesssss but maybe the propagation length doesn't matter for a small
circuit
like my 1 bit cpu =D

So I would still like to know what the maximum frequency is

And in case the circuit is bigger how do circuits nowadays propagate the
clock signal reliably

For a single bit you can use your great-grandfather's
pocket watch. It shouldn't make any difference.

dk

M

#### Mike

and it makes us work at answering questions bit by bit that
are better answered all at once (or at least in much larger chunks).

.... we're back that 1-bit Turing machine again ...

Maybe this thread is a real live Turing test?