5 Jun
2005
5 Jun
'05
5:45 p.m.
From: "Patrick Finnegan" <pat at computer-refuge.org>
Sent: Sunday, June 05, 2005 7:25 PM
Randy McLaughlin declared on Sunday 05 June 2005 06:12
pm:
I type slower than I think so I often have typo's, sorry.
They took a 5mhz brain dead version of a 16 bit
chip and ran it on an 8
bit bus @ 4.77mhz and put a RAM limit of 448mb of RAM.
Err, I think you have the memory limit wrong (I'm assuming you meant
448kB), 640kB (or 64kB on the original motherboard if you don't use a
memory expansion card) is more accurate. Also, using an 8-bit bus let them keep the cost lower,
as they required
1/2 as many latches for data, and it was easier to get 8-bit interface
parts (like 8255s and other 82xx series chips) than 16-bit parts cheaply
in 1981..
You'll note that when the AT came out, they did have a full 16-bit data
bus (and 24-bit address bus) available on the expansion bus.
The 8 bit parts used at the time are still used today and could have been
used then with an 8086. The latches cost < $0.25, it would have required
twice as many RAM chips but they usually came with four banks so a 2x2 would
have worked.
<pet-peeve>BTW, m means milli, M means mega, b
means bit, and B means
byte.</pet-peeve>
The 5% + 10% used to keep the RAM refreshed is huge. When you look at the
upper end processors you will find that 15% means lots of $. More to the
point to run a machine at less than 100% speed with zero reason makes no
sense. For less than $1.00 the machine could have been tremendously more
powerfull just by using the lessons of the time:
4Mhz Z80 is better than 2Mhz 8080, just as 5Mhz 8086 is better than 4.77Mhz
8088.
Access to full memory map of 1MB is better than 448KB.
MMU's allowing shadow ROM / separate bank for video is good.
Randy
www.s100-manuals.com
The excuse of running
4.77mhz was for color burst but I never saw a video card that didn't
use its own crystal negating the need for the 4.77mhz. In any case
Ok, I don't understand that... Colorburst is (on NTSC), about 3.58 MHz,
which isn't easy to derive from 4.77MHz. I highly doubt IBM's reason
for using that speed had anything to do with (at least NTSC) video.
By the way, the difference in speed is less than 5% between what IBM used
and what the processor was rated at. IMO, 5% is a meaningless
difference.
Pat
--
Purdue University ITAP/RCAC --- http://www.rcac.purdue.edu/
The Computer Refuge --- http://computer-refuge.org