At 23:54 Uhr +0200 04.07.1997, Bill Studenmund wrote:
>>
>> On Fri, 4 Jul 1997 17:59:29 -0400 (EDT)
>>  "Steven N. Hirsch" <shirsch@ibm.net> wrote:
>>
>>  > So I've learned!  Urk.  I had it in my head that Apple were smarter than
>>  > that by the time the SE/30 was designed.  Presumably, later vintage M68K
>>  > Macs had more sophisticated hardware?
>
>I think the idea of doing it all in software, which dated to the late
>70's and the Apple II, still held firm. Unfortunately.

To somewhat defend Apple here: My SE/30 can move data from/to a decent SCSI
drive (Quantum Empire, IBM DORS) with about 1.8 MBytes/sec (according to
SilverLining's TimeDrive, using 4MB blocks) which is about as much as the
53C80 chip can do. The NetBSD/mac68k 53C80 sbc driver still defaults to a
slower, not interrupt-driven mode because otherwise there are strange
interactions between some drives and the scsi engine. (Allen, Scott? ;)

Actually, I happen to like the non-DMA design of the Macintosh. There are
terrible examples of DMA stuff around: The i386 ISArchitecture uses onboard
DMA only for floppy access and memory refresh since an 8 MHz 80286 happened
to be faster than the 8237 DMA (not to mention the 24Bit address space);
and the Atari STs DMA/Blitter stuff was similarly braindead.

>The IIfx had special I/O Processors, which either did DMA, or did
>everything, then called the CPU to just do a block transfer.

But the IIfx also has a 53C80.

	hauke



--
"It's never straight up and down"     (DEVO)