Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!zaphod.mps.ohio-state.edu!mips!apple!netcom!mcmahan
From: mcmahan@netcom.UUCP (Dave Mc Mahan)
Newsgroups: comp.sys.amiga
Subject: Re: Constructive Question
Message-ID: <10814@netcom.UUCP>
Date: 11 Apr 90 08:24:40 GMT
References: <16450@snow-white.udel.EDU>
Organization: NetCom- The Bay Area's Public Access Unix System {408 249-0290 guest}
Lines: 55

In article <16450@snow-white.udel.EDU> BARRETT%FOREST.ECIL.IASTATE.EDU@cunyvm.cuny.edu (Marc Barrett) writes:
>
>   does anyone know whether or not the A2360 ("Lowell") video card
>has the ability to accomplish such things as palette-switching
>(on-the-fly between scanlines) or HAM modes?
>
>   If this card can do this, this may be Commodore's way out of the
>fix that they've found themselves in with the now-far-superior video
>capabilities of the Apple 8/24GC video card.  If the A2360 can
>accomplish a normal 320x200 6-bitplane HAM mode, then it should
>also be able to accomplish a 1024x768 8-bitplane "Extended HAM" mode
>(262,144 colors at once!).  If the card can switch palettes between
>scanlines (A LA Dynamic HiRes) then it could display approx. 200,000
>colors at once (256 colors per line * 768 lines).  If it can do
>both, Apple would be in deep trouble!
>
>                                  -MB-


Well, let's look at your numbers.  If we assume that we get 30 screen
refreshes per second, and that the processor changes all 256 colors EVERY
scan line, and that it takes 1/2 microsecond for each color on the 2360 or
whatever processor your using (which I doubt it can do, but maybe), we find
that :

    Time_per_scan = 1/30 seconds = 33,333 uSecs
    Time_taken_to_switch_1_line = 256 colors * 768 * .5 uSec = 98,304 uSecs

    Total_%_of_CPU_Time = 98,304 uSec/33,333 uSec * 100% = 29.5%

Are you willing to sacrifice a conservative 29.5% of your available CPU
resources to obtain this type of resolution?  I'm not.  Please not that
this estimate is VERY conservative, and is probably somewhere closer to
50% or 60% by the time the CPU gets finished figuring out which palette
it should use, where to get it from, etc.  Now you have the power of the
68030, running at a little more than the throughput found in the 68000
you have in the standard A500, A1000, and A2000.  Seems like kind of a
shame, don't you think?  And we haven't even taken into account all the
extra memory required to store all those palettes.  This figure comes
out to around 576 Kbytes, just to store the palettes for each line.
That's approximately 64 Kbytes more than are in my current amiga, but
just barely doable in a machine with a fat agnus.  The data for this
wonderful picture you are creating will take up about 768 Kbytes.  Add
that to what we have already consumed for our color palette, and we
come up with just about 1.3 Mbytes to store one picture.  Slightly more
than 30% bigger than the current fat agnus can deal with.  And we haven't
even taken into account contention on the chip bus or the extra memory
required to store links to the color palette so we can switch at such
high rates.  As we can see, sometimes C= DOES know what they are doing
by including custom chips in the amiga to do this kind of processing.

Now we know why they invented custom chips, don't we?


  -dave