27 May
2005
27 May
'05
2:39 p.m.
Subject: Re: Remembering RAMAC
From: "Dwight K. Elvey" <dwight.elvey at amd.com>
Date: Fri, 27 May 2005 09:44:38 -0700 (PDT)
To: cctalk at classiccmp.org
The big advantage and values of drums was high inerta plus parallel
heads. Some designs were 16bits wide (maybe there were wider)
plus timing tracks. So while small ( I remember the 128kw
swapping drum on the KA10s) they were extremely fast.
Of course late in the game was the 32kW RS08 disk for the PDP8
systems. Small but very fast word wide storage.
Every so often I get the itch to find a small drum and get it
working.
Allison
From: "Vintage Computer Festival"
<vcf at siconic.com>
---snip---
I was thinking more in terms of space. A disk is much more efficient in
this regard. You can stack N times the number of disks in the same space
that a drum takes up.
The problem with disk, of course, is the variable BPI.
Today, with the electronics so cheap, the CD-ROM is made
with a spiral track and it runs at a constant BPI by
changing the rotation speed as it reads the disk.
Dwight
27 May
27 May
2:59 p.m.
>>>> "Allison" == Allison
<ajp166 at bellatlantic.net> writes:
Allison> The big advantage and values of drums was high inerta plus
Allison> parallel heads. Some designs were 16bits wide (maybe there
Allison> were wider) plus timing tracks. So while small ( I remember
Allison> the 128kw swapping drum on the KA10s) they were extremely
Allison> fast.
Some disks had parallel heads too -- the CDC 6638 is 12 bits parallel.
Allison> Of course late in the game was the 32kW RS08 disk for the
Allison> PDP8 systems. Small but very fast word wide storage.
Not fast at all, because that one is NOT parallel -- the RS08 (and its
PDP11 successor RS64, as well as the RF11/RS11) are all serial fixed
head drives. The doc says the RS08 takes 32 microseconds per (12 bit)
word; the RF11 was slightly faster at 15 microseconds per (16 bit)
word.
paul
3:01 p.m.
Some disks had parallel heads too -- the CDC 6638 is
12 bits
parallel.
I'd even say that _most_ disks have parallel heads, though they
generally throw away that parallelism. (I don't know why; the most
plausible reason that comes to mind is a "buh-but that's not how disks
work" mindset.)
/~\ The ASCII der Mouse
\ / Ribbon Campaign
X Against HTML mouse at rodents.montreal.qc.ca
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B
4:47 p.m.
>>>> "der" == der Mouse <mouse
at rodents.montreal.qc.ca> writes:
> Some disks had parallel heads too -- the CDC 6638
is 12 bits
> parallel.
der> I'd even say that _most_ disks have parallel heads, though they
der> generally throw away that parallelism. (I don't know why; the
der> most plausible reason that comes to mind is a "buh-but that's
der> not how disks work" mindset.)
Maybe, but probably not. More likely the answer is "it cost a lot (in
1970) to build N read/write channels rather than just one". With
serial data, you just have one read/write channel, and someplace very
close to the heads you run that through a switch to connect to one of
N heads. For parallel operation, you have to duplicate a lot of
logic. Today that's just one chip, but not so in 1970...
paul
5 p.m.
der Mouse wrote:
I'd even say that _most_ disks have parallel
heads, though they
generally throw away that parallelism. (I don't know why;
Because at modern track pitches, there is not enough dimensional
stability to guarantee that the positioner is on-track on all
surfaces simultaneously. In fact, if it's on-track on one surface
it can be off by many tracks on another. This is one reason why
they no longer have a servo surface, and instead use embedded
servo on every data surface.
Eric
7001
days inactive
7001
days old
4 comments
4 participants
participants (4)
-
ajp166＠bellatlantic.net -
eric＠brouhaha.com -
mouse＠Rodents.Montreal.QC.CA -
pkoning＠equallogic.com