computers....

A thread of 2 messages, 5 March 1992 to 6 March 1992, in the order they were sent.

#3831 computers....

Scott Kellogg:

>I have run into the problem of constructing ships larger than allowed by
the
>rules as reguards to computers.
>
>What I did was to allow multiple computer set-ups.  Triple installation of
>computers would monitor all ship functions that they are capable of
>monitoring.
>Thus a ship that had twice as many CPs as allowed for the computer she was
>carrying would have to have at least 6 computers rather than the normal 3.
>On top of this, I added three more computers to integrate the data of the
>first two.  Thus the above example would have 9 computers rather than the
>normal 3.
>
>I figure that three computers should be able to handle just about all the
>integration of many many multiple computer installations.
>
>Thus if a computer can only handle X CP, and a ship has Y CP in it's make
>up, the formula for the number of computers becomes Computers=Y/X+3.
>
>I found this useful when designing a 50,000 ton ship with a spinal mount
at
>TL 8.  :-)

By a curious coincidence, Cynthia and I were discussing that very subject
just last night.  Our solution was somewhat more complex than Scott's, but
possibly more consistent with MT's (admittedly maimed) rules regarding
computers and controls:

1)  A computer can control as many control points as it's (Max Input) times
it's CP Multiplier.  It can accept an input equal to it's Max Input.

2)  Computers can operate in series by connecting the output of a computer
into the inputs of other computers, to the limit of the Max Input of the
subordinate computers.  Thus, a computer can drive a number of subordinate
machines equal in number to it's CP multiplier.

3)  For control purposes, any size machine may be built up from smaller
machines using the above guidelines, and so any size ship may be built at
any TL, with a penalty due to the increased size/cost of the computers.

For instance, a model-10 computer can be built up from 50,000,002 model-0
computers, at a cost of MCr3,000,000.12, taking up 25,000,001 Kl (1,851,852
displacement tons).  Anyone need a TL5 supercomputer?  Codename Merlin,
perhaps?  A simpler example is a TL7 substitute for a model-5:

     Model-5 output = 100,000 x 35 = 3,500,000
     model-2(TL7) output = 10,000 x 15 = 150,000.
     so, 24 model-2s would provide sufficient computing power, with two
     more to control those twenty-four (since the model-2 has a CP
Multiplier
     of 15, two are required to control more than 15 subordinate machines),
     and a single machine to control the two machines and link the whole
     setup into a single computer.  The final machine consists of 27
model-2
     computers, at a cost of MCr51.3 (compared to MCr9.7 for a M-5), and
     taking up 94.5 Kl (compared to 6.5Kl for a M-5).

In general, the computers produced are larger than their high-tech
analogues,
but they do sometimes work out cheaper, for those who like to really
squeeze every last credit before letting go.

Later...

- -- Via DLG Pro v0.985b

[email removed]

I have learned to use the word "impossible" with the greatest caution.
                                    --W. von Braun

#3833 Re: computers...

hi,

Steve writes,

>By a curious coincidence, Cynthia and I were discussing that very subject
>just last night.  Our solution was somewhat more complex than Scott's, but
>possibly more consistent with MT's (admittedly maimed) rules regarding
>computers and controls:

>1)  A computer can control as many control points as it's (Max Input) times
>it's CP Multiplier.  It can accept an input equal to it's Max Input.

>2)  Computers can operate in series by connecting the output of a computer
>into the inputs of other computers, to the limit of the Max Input of the
>subordinate computers.  Thus, a computer can drive a number of subordinate
>machines equal in number to it's CP multiplier.

Both this solution and Scotts don't really mesh very well with a tiny detail
from the class Book 8 where massively parallel machines are allocated TL=8.

Once they become available, adding extra machines isn't going to improve
performance all that much (because you already have thousands of them).

The limits could be imposed by the software methodologies present at the
given TL rather than the hardware being pushed to the limit.  i.e. If building
large parallel machines were simply a matter of throwing lots of $$$$s and
silicon at the problem then we'd have them already (and my PhD would not be
worth finishing :-).





        						Pauli

Paul Dale               | Internet/CSnet:            [email removed]
Dept of Computer Science| Bitnet:       [email removed]
Uni of Queensland       | JANET:           [email removed]
Australia, 4072         | EAN:                          [email removed]
                        | UUCP:           [email removed]
f4e6g4Qh4++             | JUNET:                     [email removed]