Aucbvax.4596
fa.works
utzoo!decvax!ucbvax!works
Wed Oct 21 23:43:06 1981
WORKS Digest V1 #21
>From JSol@RUTGERS Wed Oct 21 23:02:58 1981
WorkS Digest Thursday, 22 Sep 1981 Volume 1 : Issue 21
Today's Topics:
Other Considerations Of 32 Bit Processors
Videodisks As Memory
----------------------------------------------------------------------
Date: 21 Oct 1981 1918-PDT
From: SCHIFFMAN at SRI-KL
Subject: Addition to processor taxonomy
In my consideration of the "32-bitness" of CPUs, I now realize that I
left out two other things that have a width:
4) The number of physical address bits presented to the
memory bus.
Obviously physical address bits cost more IC pins and
backplane wires, contributing to cost. Not so obviously,
address bits can contribute greatly to the cost of memory
management hardware (and the time cost of using it). The
point is that you don't really want to pay much more for
physical address bits than you need to address the memory
your systems will be able to afford.
5) The internal "name-space" of the architecture, i.e. log2 of
the number of unique objects that can be referenced.
The reason why we can't more simply define this
(as say, the number of bits in a `pointer') is that
some architectures can have arbitrarily complex ways
of encoding internal names, say by using the data-type
field of an op-code to provide extra "logical-address
bits".
Again, a large name space can be a mixed blessing. One of
the most important (for instruction encoding efficiency)
areas of CPU architecture is how to compress memory
requirements for expressing local names. This can be done
badly for machines with even a relatively small name space
(i.e. the PDP-11, where all offsets take 16 bits).
Under these additional criteria, the updated table...
WIDTH (Bits)
CPU Registers ALU Memory Phys. Log.
Data Addr Addr
------------------------------------------------------------
Z8000 64,32,16,8 16 16 24 24
MC68000 32 32 16 24 24
I8086 16,8 16 16 20 20
I8088 16,8 16 8 20 20
NS16000 32 32 16 24 24
iAPX-432 to 80 80? 16 24 40
____________________________________________________________
I reread the 43203 data sheet and decided that the memory data bus
was, by the definition that I gave, not "user-supplied", but 16 bits.
The reason for this confusion is that the data bus operates in burst
mode (only one address output for the transaction) for operations on
words longer than 16 bits; this is real optimization since addresses
usually take as much as half a bus cycle.
-Allan
------------------------------
Date: 21 October 1981 2333-EDT (Wednesday)
From: Stephen.Hancock at CMU-10B
Re: Rotating Memories ("video disks")
The October issue of Mini-Micro-Systems has an artice entitled
"Rotating- memory devices move to optical reading". The article
informs us that european systems are being devoloped by Philips NV
and Thomson-CSF. The device is supposed to be a 10G bit drive with a
twelve inch platter. There are plans to have a system with a 64
platter magazine (thats 640G bits). The system is expected to be
released in about 2 years the projected OEM cost is between $2000 and
$10000 each. The article also points out the sales dollar volume and
number of optical units sold will be only a small fraction of the
total rotating memory market.
Steve
------------------------------
End of WorkS Digest
*******************
-------
-----------------------------------------------------------------
gopher://quux.org/ conversion by John Goerzen <
[email protected]>
of
http://communication.ucsd.edu/A-News/
This Usenet Oldnews Archive
article may be copied and distributed freely, provided:
1. There is no money collected for the text(s) of the articles.
2. The following notice remains appended to each copy:
The Usenet Oldnews Archive: Compilation Copyright (C) 1981, 1996
Bruce Jones, Henry Spencer, David Wiseman.