Subject: Variation of CPU speed explained!
Date: Mon, 2 Dec 1996 08:31:01 -0700
From: Matthew Mastracci
Organization: The University of Calgary
Newsgroups: comp.sys.hp48
Owing to the large amount of discussion over the constantly varying
CPU speeds, I would like to offer the results of my investigation. It
seems that the HP48 suffers from its own form of relativity. As many
physics students may know, if you head towards a spaceship travelling at
0.5c while travelling at that speed yourself, the ship will not be heading
towards you at 1.0c, but at a number significantly less than that, owing
to the Lorentz transformation:
v - u
v' = ------------ (Eq. 1)
1 - uv / c^2
where: v' is the relative speed
v is your speed
u is his speed
c is the speed of light (3.0 x 10^8 m/s)
There is also a parallel between this relativity and relativity
dealing with CPU speeds. Note the similar units:
Velocity: m/s
Processor speed: Hz, or 1/s
(speed is analagous to velocity when travelling in the positive direction
and as the HP48 PC register always travels in the positive direction
[except when using the "undo" function], we will refer to this as
"absolute processor velocity")
Also, there is a parallel between the speed of light (maximum
attainable velocity) and something I would like to call the "terminal
processor velocity". The value of the speed of light is (as stated
before) approximately 3.0 x 10^8 m/s. The terminal processor velocity is
therefore 300MHz (note that there are no P300s!). We can therefore
rewrite Eq. 1, the Lorentz transformation, as follows:
The Alcuin Transformation:
v - u
v' = ------------ (Eq. 2)
1 - uv / c^2
where: v' is the relative processor velocity (apparant CPU speed,
measured with the self-test)
v is your brain's processor velocity
u is the calculator's absolute processor velocity
c is the terminal processor velocity (300 MHz)
The results clearly state this: the processor speed does not actually
change over a period of time, you are simply thinking faster or slower
(depending on the direction of change) and changing your processing frame
of reference. This explains why the speed goes down with heat: you are no
longer thinking about working, but instead about beaches and relaxation,
therefore concentrating less on the calculator.
To counter this phenomenon, there are a few things you can do to help
you increase the apparant CPU speed of your calculator:
1) Stare at your calculator while it performs intense calculations
2) Get a friend to stare at your calculator too
3) Turn down the temperature in the room so that you concentrate less on
beaches and more on the calculator
4) Meditate 20-30 minutes before using the calculator to focus your mind
on it
This "law of processor relativity" can also be extended to explain
the processing speed difference between the HP48S-series and the
HP48G-series. As the HP48G calculators have fancy pick-lists, interactive
screens and large amounts of visually pleasing greyscale games, you think
about them twice as much and the processor speed is therefore twice as
much.
An interesting result of this law is that no computer may actually
calculate at a speed of 300MHz. As with the theory of relativity for
light, however, it is possible to lie on either side of this value, with
values higher than this yielding a travel in time backwards, or in the
case of CPUs, a negative increment in the instruction pointer register.
Supercomputers, like the Cray, that process faster than 300MHz must
therefore be programmed backwards to achieve the desired results.
It is unlikely that home computers will reach this speed however, as
all existing software would have to be run through a filter to reverse the
order of all the bytes.
/\/\att /\/\astracci mmastrac@acs.ucalgary.ca
---------------------------------------------------------------------
Subject: Re: Variation of CPU speed explained!
Date: Mon, 02 Dec 1996 11:00:59 -0800
From: Dave Arnett
Organization: Hewlett-Packard Mobile Computing Division
To: Matthew Mastracci
Newsgroups: comp.sys.hp48
This is an interesting idea, but I fear it has a small flaw in that it
applies quantum principles with a classical mindset.
The concept which is missing is that the bits are quantized, and so the
observed effects can be more profound than those predicted. There is
not a continuum of speeds at which the bits can move, because they are
trapped inside a box. The bits must propagate through the CPU at
certain allowed energies related to the CMOS Cell parameters and the
dopant levels used in fabricating the CPU.
The best battery efficiency in such a system is achieved by placing the
energy of the bits at the desired quantum level, so the production of
stray fluxons is reduced.
We placed the default energy just above the selcted quantum level to
allow for certain variances. However, we forgot to account for the fact
that any point on the surface of the earth is moving faster through
space at midnight than at noon. This is where the Lorentz/Alcuin
transformations come into play, in that the difference in v can cause
large quantized jumps in bit speed as the Gibbs Free Energy of the
system is changed during planetary rotation.
This phenomenon was first observed during the Gemini missions. The
real reason that HP calculators have flown on so many NASA flights
is that they use the variations in the reported CPU speed as a means
of measuring their orbital velocity. This was quite important on
Apollo, since they had no earth-based references for navigation while
on the far side of the moon.
Omitting these factors in the Yorke IC design was a serious error on our
part.
Good day!
Dave.
------
I don't speak for Gibbs or Gerber when I post here.