home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.atari.st.tech
- Path: sparky!uunet!mcsun!sun4nl!wn1.sci.kun.nl!sci.kun.nl!sanders
- From: sanders@sci.kun.nl (Sander Stoks)
- Subject: Re: Falcon BUS..
- Message-ID: <BxG859.2Cu@sci.kun.nl>
- Sender: news@sci.kun.nl (NUnet News Owner)
- Organization: University of Nijmegen, The Netherlands
- References: <mpnolan.720881057@unix1.tcd.ie> <27701@castle.ed.ac.uk> <jonal.95.720952501@dhhalden.no> <1992Nov5.114927.14711@dcs.warwick.ac.uk>
- Date: Mon, 9 Nov 1992 12:31:56 GMT
- Lines: 40
-
- In <1992Nov5.114927.14711@dcs.warwick.ac.uk> leo@dcs.warwick.ac.uk (Leo Hendry) writes:
-
- >I asked questions similar to these a month or two ago, but nobody answered.
- >Perhaps somebody knows the answers now.
-
- >1) What is the speed of the Falcon's memory (Somebody said it had 0 wait
- >states which seems to imply 16MHz, but they may have been wrong)
- >2) How does the DSP access the main memory (Somebody said it is not connected
- >to the data bus, in which case I suppose it must be some kind of serial link)
- >3) Does the TOS V4 use the DSP to speed up things (for that matter does it
- >make intellegent use of the Blitter?)
- >4) Is the data bus between the memory and the 68030 16 or 32 bits wide (There
- >seems to be an argument going on about this).
- >5) Why does the benchmarks that I have seen posted here only indicate an
- >approx 3x speed increase and real-world tests (excluding any disk access)
- >only twice as fast? If the data bus is twice as wide, the memory twice as
- >fast, the processor at least twice as efficent and running at twice the clock
- >speed it should be more like 10 times the speed.
-
- Most of the things you ask I cannot answer, but I can see your last
- point is not quite right. If every wheel of a car runs twice as fast,
- the whole car runs twice as fast, not eight times. The double clock of
- the 68030 in the falcon as opposed to the ST gives a factor 2 ONLY if
- the memory is twice as fast too. Now the 68030 is indeed more
- efficient (check out those bit shifts in QIndex...!) and it is said to
- run at 3.84 MIPS. The ST is considered approx 0.8 MIPS, so there
- should be a factor of about four. Add the fast shifter (but more
- complex graphics, so not too big a difference), and the DSP which can
- cut short some very CPU-intensive tasks. This should give a real-life
- average speed factor of about 4-5. And include disk access; if there
- really is a throughput of 3M/s (!!!) and an average access time of 12
- ms (!!!) we could see some humongous speed increases here. (And
- virtual memory swapping to disk would be really smooth...)
-
-
- >Thanks,
- >Leo
-
- Sander SToks
- sanders@sci.kun.nl
-