Re: Zip GS questions

[spec@vax2.concordia.ca (Mitchell Spector)]

In article pubpc1@library.ucla.edu writes...

>Barry, yes it will work.  I'm in the dark about how the Zip maps memory,
>but from experimenting on the Zip, it seems that any time you have four
>32k x 8 SRAMs installed, you can set the DIP switches to any of the
>configurations from 8k all the way to 64k.

    That's odd, are you sure you have 32K x 8 SRAMs installed? I do on
mine, but if I configure the DIP switches for anything _but_ 32K or 64K, 
the board will not function (it crashes and does strange things before 
even the Finder desktop can appear). I'm using a mix of 300 and 600 mil 
chips, 45ns Data, 15ns Tag, though I don't think that makes a difference.

>It is my belief that 16k is the "magic" number in terms of cache size.
>It appears to be the cache size at which the caching becomes effective
>and beyond that, increasing the cache size will not improve performance
>that much more.  The speed for P8 and GS/OS programs like AppleWorks GS
>also showed massive degradation in performance at 8k, but no noticeable
>difference between the other (16k, 32k, and 64k) configurations.

    I noticed this too. Before I received a 15/64, I ran with an 8/16
board for two years. If I slow my newer board down to an 8/64, there is
a difference, however it's hardly noticable. I also noticed, or rather
I didn't, the difference between running my board as a 15/32 and 15/64.
I also have a TransWarp GS 6/8, and this board is _much_ slower than my
old 8/16 ZipGS (though this is the fault of both cache and clock-speed).

    Performance may level off at 16K cache, but keep in mind there is
also a point where a faster clock-speed does too. I found the once you
hit about 10-12 MHz, there is almost no noticable difference with any
speed increases. This is mostly due to the 1 MHz bus of the GS.

>The 55MHz and 60MHz oscillators are available from Digi-Key.  They
>are the "half-size" kind, but the Zip can use both types.

    I have a full-sized 60 MHz oscillator, so they do exist. If I can
find one, I may try a 64 MHz oscillator (16 MHz) although there would
be little point in doing so.

Mitchell Spector
spec@vax2.concordia.ca