Re: Zip GS questions

[pubpc1@library.ucla.edu]

dempson@actrix.gen.nz (David Empson) wrote:
>An increase in the size of the cache does not generally produce as much
>of a noticeable speed improvement as an increase in the CPU speed.  It
>depends on what the software is doing.
>
>I have had personal experience of 8/16 (my current card), 8/64 (my
>original card, now dead), and I've also used a 9/64 card.  I didn't
>notice a significant drop in the apparent speed between the 8/64 and
>8/16 cards.  The 9/64 card was noticeably faster than either of the
>others.
>

That is very true.  I have a Zip running at 13.75MHz.  The actual chips
in the cache make up 64K (it used to be set at 64K).  The problem was,
it would not run stable.  When I reset the DIP switch configuration to
16K of cache (even though the chips were still 64K w/all 4 sockets filled),
the board became very stable.

I have not noticed much of a performance hit compared to 13.75/64 and the
extra megahertz definitely make it faster than when it was a 9/64.

I also did not have to replace the 7400.

>
>Let us consider an example of three ZIP cards: a 7/16, 7/32 and 8/16.
>
>The peak execution speed is given by the CPU speed in each case, thus
>the 8 MHz card will be able to execute fully cached code faster than
>either of the other cards.  This means that tight loops will be
>noticeably faster (about 14%) compared to the 7 MHz cards.  (The average
>speed of the computer will not increase as much.)
>
>An increase in the cache size doesn't help with tight loops, but allows
>the average speed of more complicated (or widely distributed) code to
>increase, because more of it can be stored in the cache.
>
>Considering that operations like moving memory and graphics operations
>involve tight loops, the higher clock speed produces a more visible
>increase in speed.
>
>I think you would find that an 8/16 card will be faster on average than
>a 7/32 card.  If you can't afford both higher speed and larger cache,
>then go for the speed.
>
>The same argument applies for higher clock speeds, but you get a
>steadily smaller relative increase in peak and average speed for each 1
>MHz increase in CPU speed.  At some point, you would have more to gain
>by increasing the cache size than by adding another 1 MHz.
>
>(In addition, there is a limit to how much you can increase the clock
>speed, which depends on the particular machine and accelerator.)

An additional factor to consider is that it has been stated that the
newer Zip 7/32 boards come with an 8MHz processor anyways and that they
are just 7MHz because of a slower oscillator.  They could easily be
turned back to speed with the right oscillator.

Zip Tech (MCTA Technologies) is just propagating the rumor of difference
to make more money on the 8/16 model.

:)

-Scott G.