home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!usc!chaph.usc.edu!phakt.usc.edu!not-for-mail
- From: baffoni@phakt.usc.edu (Juxtaposer)
- Newsgroups: comp.sys.atari.st
- Subject: Re: The Great Falcon Debate
- Date: 16 Dec 1992 12:03:50 -0800
- Organization: University of Southern California, Los Angeles, CA
- Lines: 42
- Message-ID: <1go236INNdkn@phakt.usc.edu>
- References: <1992Dec9.145828.2736@aston.ac.uk> <1992Dec10.221041.28151@dcs.warwick.ac.uk> <29415@castle.ed.ac.uk>
- NNTP-Posting-Host: phakt.usc.edu
-
- In article <29415@castle.ed.ac.uk> marks@castle.ed.ac.uk (Mark Steyn) writes:
- >leo@dcs.warwick.ac.uk (Leo Hendry) writes:
- >>In article <1992Dec9.145828.2736@aston.ac.uk> wintermi@cs.aston.ac.uk writes:
- >>Ahh, but the standard memory is not 32-bit. If 3rd party 32-bit memory
- >>upgrades using the 68882 socket can be made then you are right, but I have
- >>not been convinced than this is possible.
-
- >No, the standard memory is 32 bit. Only the 68030 had 16 bit access to
- >memory, other parts of the Falcon have full 32 bit access.
-
- > Mark
-
- If the only thing skimped on is the 16 datalines from the CPU to the
- "standard" memory, wouldn't it be possible to simply make an adapter from the
- 68882 socket (which to my knowledge _must_ have all data lines to the CPU)
- and run it to either a) a nice adapter that plugs inbetween the memory card and
- the memory socket or b) a nice hack that directly connects the 16 data lines
- to the appropriate 16 data lines on the memory card? This alone should give
- a roughly 2x speed increase on CPU throughput (memory expansion aside - I
- personally do not see that the first-time/casual/hobby-type person who buys
- the Falcon will be likely to need more than 14MB, but lets not get into _that_
- flammage :). Of course, the speed will still be somewhat slower than that of
- the TT as (especially in higher rez/color modes) there will still be a
- significant amount of bandwith being used for video, but nearly all CPU
- accesses will be twice as fast (I say nearly because of factors such as cache,
- instruction pipelining, etc. already speed up the '030 to reduce the effects
- of such bottlenecks).
-
- Or is the '030's ability to use 8/16/32 bit data buses not "auto-
- sensing" (in which case you would have to alter the logic that decides whether
- it is using a 16/32 bit bus)? Would the software conflict (ie it expects
- 16 bits/transfer) or is the bus size transparent to the software?
-
- I apologize if this has already been asked/answered but I am in the
- process of trying to catchup to some 400 messages :(.
-
- >--
- >"A little nonsense now and then, \ Mark Steyn : \
- > Is relished by the wisest men" \ marks@castle.ed.ac.uk \
- > Willy Wonka \ mpjs@dcs.ed.ac.uk \
-
-
-