[TriLUG] Comparing a Univac 1105 with today's PC
neilson at windstream.net
Mon Feb 8 15:28:30 EST 2010
Judy Hallman wrote:
> The UNC Computation Center was dedicated in March 1960. We're working on
> some materials celebrating 50 years of computing on campus. The
> description of UNC's Univac contained:
> Memory Capacity
> Core (fast: access time .000008 second) 8,192 words
> Drum (slow: access time .017 second) 32,000 words
> With help from Google, I have:
> The Univac 1105 at UNC-Chapel Hill had 8,192 36-bit "words," each of
> which could be used to store an integer, floating point number, or 6
> characters. In today's measure, its memory capacity was less than 50
> It took .000008 seconds between initiation of a call for data from the
> Univac's core memory and delivery of the data (access time). Access
> speeds today are about 40 nanoseconds (.000000040 seconds).
> Auxiliary storage was a magnetic drum which stored 32,000 words with
> access time of .017 second. The average access time of today's typical
> hard drive is roughly 5.6 milliseconds (.0056 seconds).
> Does this look correct? Can you improve it?
> Judy Hallman (UNC Class of '59)
I would suggest giving timing specs in microseconds.
There is a lot more information available online.
See http://en.wikipedia.org/wiki/UNIVAC_1105 for a picture of the
installation at the Bureau of the Census.
http://ed-thelen.org/comp-hist/BRL61-u.html contains (rather far down on
the page) a whopping amount of info including specifically the UNC
installation. This was a vacuum-tube computer that used some
transistors. The electrical power requirements were tremendous, 110 kw
for the computer and 25 kw for the air conditioner.
It is amusing to note that an additional 4k of core memory could be
>>rented<< for $4500 a month!
The magnetic drum almost certainly used head-per-track technology,
rather than the moving head that's used on disks. Drums were expensive
and tended to wear out in just a few years. A major use of drum memory
was for "overlays" in which portions of the executable code would reside
on drum and be called into core for execution as needed. This was prior
to swapping and virtual memory, although around this time (1961)
Burroughs already had virtual memory available in the B5000. The
Burroughs sales staff saw no way to sell VM until after Univac and IBM
had it five years later.
More information about the TriLUG