So, there is a lot of talk about clock speed, how it is the thing to
consider when buying a PC. Does this mean that the emphasis is not on
how many calculation you can do in a second because of shear processor
power, but rather how many you can do because the thing is just running
electrons around faster?? I liken it ot he old horse power/torque
debate. a 200 hp car with 120 lbs torque is much different than a 120 hp
truck with 200 lbs torque. Yeah the car is faster, but put 3 adult in it
and you lose a lot of your power.

So, what about a hypothetical PII 450 and a PIII 450.... I would think
(assuming that they all have the same set up....) the PIII is
preferable..... right??(as far as performance is concerned.)

I think most of it is advertising B.S. myself.....

As long as I can produce a perfect sphere on an infinite checker board,
I am cool


-paul

--
-------------------------------------------#
Paul Daniel Jones
120 Chandlee Laboratory
Penn State University
814-865-2090
pdj### [at] psuedu

http://research.chem.psu.edu/glassgrp/paul
--------------------------------------------#

Post a reply to this message

Paul Daniel Jones <pdj### [at] psuedu> wrote:
: but rather how many you can do because the thing is just running
: electrons around faster??

  I think that this was just an allegory, but still it bothers me a bit.

  Electrons run pretty slowly on conductors. With constant current they move
something like a couple of centimeters per hour. With alternating current
they practically don't move at all (well, yes, they move a bit back and
forward with the current, but the average movement is about 0).
  The phenomenon that we call 'current', however, moves through a conductor
at about 0.7c (in copper at least, I think; it may vary from material to
material). This means that if you send a signal through a conductor, it will
move at about 210000 km/s.

  However, when we are talking about processors and CPU clocks, the speed
of electrons and electrical current is irrelevant.
  Processors consist nowadays of transistors. You can make almost everything
with transistors (logical ports, arithmetic operations on bits and so on).
I'm not going into the details about how a transistor works, but it's
basically a "switch" with 2 inputs and one output which allows the current
to pass through depending on the state of the inputs.
  A transistor is not infinitely fast (nothing physical can be), so if the
inputs change, it will take some time before the output changes (we are
talking about nanoseconds or even smaller amounts of time, but still it
takes some time). As the current change travels through several (hundreds,
thousands...) transistors, it takes its time. Some paths are shorter and
others are longer, so a signal takes different time to travel through
different paths.
  This is the reason why we need some way of synchronizing all the
transistors. We need all the results from several paths at the same time
and thus we need to synchronize to ensure that this happens.
  The CPU clock is this synchronizing element.
  Of course the faster the clock "ticks", the faster things will be done.
However, since transistors are not infinitely fast as I said before, raising
the clock frequency too high will cause the problem that some signals will
not travel through its path fast enough to get in time to the end, and thus
that signal is lost (with usually fatal results, ie. the processor
malfunctioning and crashing).
  The speed of a transistor depends at least on its materials, its quality
and its temperature (note that temperature is not the only element; you
can't raise the clock frequency of a 8088 to 1 GHz no matter how low your
temperature is; the material and quality of the transistors affect the result
as well).
  So the speed of electrical current is irrelevant. It's the speed of the
CPU clock which affects how fast each step is done.

: So, what about a hypothetical PII 450 and a PIII 450.... I would think
: (assuming that they all have the same set up....) the PIII is
: preferable..... right??(as far as performance is concerned.)

  CPU clock speed would be the only affecting thing only if each assembler
instruction took exactly the same amount of time in those two processors.
  However, there's still a lot to be optimized. Many assembler instructions
take more than one clock to execute. Some instructions take more or less
clocks depending on the status of the system (eg. the data on the cache
memory). With multiple pipelines some instructions can be done at the same
time with other instructions, while other instructions can't (and there are
even instructions that must be done all alone, without any other instructions).
  If the PIII has more L1 cache, more optimized pipelines and more optimized
assembler instructions, it will be faster than the PII at the same clock
speed.

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):_;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message