Jeks ... I just killed a real long description of what
could be the formula .... again...
> >Well, I'd rather go for Zuse. Not only because he is
> >German(*), but rather because there is still no basic
> >or derivated SI unit starting with Z, so Z would give
> It's sufficiently obscure and I like it. Lessee a 400milliZuse
> system... Yep rolls on the tongue nicely and sounds oh
> so technical.
We need some Hollywood Germans to pronounce it ...
> >If this has to be successful we have to fit it into
> >the SI system :)
> Go with it and we also need a "standard".
Ok, I read suggestions for
a) size
b) weight
c) power consumption
d) memory
e) processing speed
a) Size
of course size - we all agree her size does matter,
and bigger is better - isn't it ? Now, size is a
measured in meters, we just have to decide if we go
for area (m^2) or volume (m^3). I'd rather like to
see area area included. Tree reasons: first (most
important to me), area numbers are usualy larger
than volume numbers for for similar machines.
Second, volume would give tall systems and advantage
over systems with a large footprint. Third by using
just the 2d footprint we can cover odd shaped machines
very easy. And a PET should get a higher value of Z
than any tower-PC :)
b) weight - no problem her - more is better.
BTW, the kg is the only unit where the base unit
inkludes a prefix ...
c) power consuption - Yea, raise my power bill and
switch of the neighborhood... Unit is Watts (w).
I would suggest to use the amount of input power,
so including all losses in Power supplies atc, which
belong directly to the system - and if you happen
to have an integrated steam engine to drive your
mechanical computer, the wats have to be calculated
from the amount of coal needed to run the engine ...
d) Memory. Most would go for bytes, or kilobytes, but
first, only units without a prefix are to be used, and
second, bytes are not realy applicable to all machines.
for example the early Zuse machines had only a storage
for floating point numbers ... I think bit may be an
apropriate number. The backside of this is that the
numbers get pretty soon very high - and the Z number
quite low. To be counted is all external Memory. I'd
exclude registers etc. Only real main memory. If there
is no memory at all, the number 1 is to be used (well,
sounds odd, I can't come up with a machine totaly without
memory). High numbers are bad.
e) processing speed - lower is better
The only measurement I know for that would be IPS
(instructions per second). Defined as whole instructions
to be executed within a second. Since this number is somewhat
fuzzy on most machines, we may have different Z numbers. One
can use a peek IPS rate, a average IPS, a median IPS or a
maximum IPS rate
Maximum: If the machine is only executing its fastest operation,
Peek: something like from a linpak
Average: like taken from an analyzer which traces the number of
instructions for some time over typical aplication or calculated
by an average instruction mix.
median: just the average over all instruction timeings.
Taking the above arguments we can come up with a formula
like:
m^2 * kg * w
------------
bits * IPS
So if we have a machine, lets say like a 4k PET, we get the
following numbers (rough estimations, I don't have one right
here
Area: 45x50 cm = 0,225 m^3
Weight: 17,5 kg
Power: 105W
Bits: 32768
IPS: 250 kips (assumption)
Result: 50,47 nZ (Nanozuse)
or a Kim 1
Area: 20x40 cm = 0,08 m^3
Weight: 1,5 kg (including power supply
Power: 15W
Bits: 9120
IPS: 250 kips (as above)
Result: 0,79 nZ
Well, the machine is a lightwight baby with the same processing
power than the PET .. only the large footprint helps
I think machines like a Northstar (or most S100 boxes)
have a big advantage ... lage footprint, superheavy
powersups and not to much processorpower ...
I like it.
Gruss
H.
--
VCF Europa 3.0 am 27./28. April 2002 in Muenchen
http://www.vcfe.org/
Received on Thu Apr 18 2002 - 13:16:50 BST