Matt,

On Fri, Apr 19, 2013 at 8:05 AM, Matt Mahoney <[email protected]>wrote:

> Perhaps I could interest you in ENIAC. Specs:
>
> Computation rate:
> 5000 10-digit additions or subtractions per second.
>

WOW, that is even faster than the IBM-650 that was my first vacuum tube
computer, after a couple of relay-based computers, including an IBM-407 and
one of my own construction that was at the time of its construction the
fastest computer on the planet for computing game theory solutions. USAF
General Curtis LeMay even gave me an award for it.

The IBM-650 used a rotating magnetic drum for its 2,000 word main memory,
that rotated at 200 revolutions per second. Careful placement of
instructions could often execute 5 per revolution, but any sort of random
access was slowed down to one per revolution. The net effect was that
instructions took from 1-5ms each to execute.

After the IBM-650 came an IBM-709, another vacuum tube computer that
computed at about ENIAC's speed.

The following ENIAC speeds were from software:

357 multiplications per second.
> 40 divisions per second.
> 3 square roots per second.
>
> Memory: 20 10-digit registers when built in 1946, later increased to
> 100 words of magnetic core memory in 1953.
>

Yea, there was an option for 60 words of core memory for the IBM-650, but
they didn't have it on the machine that I used.

>
> Software: none. Programs were written by plugging wires into a
> switchboard.


Yea, that is how the 407 worked. With this, relays of that era could often
run as fast as vacuum tubes that followed, by doing pretty much everything
in parallel. The IBM-407 would typically process 150 cards per minute,
including all computations!!!


> A stored program capability was added in 1948.
>
> MTBF: initially a few hours, increased to 2 days in 1948 using
> high-reliability vacuum tubes (17,458 of them).
>

Don't believe that 2 days. That was only temporarily with nearly new
tubes.

Then, as computers got bigger and cheaper, like the IBM-709, MTBF
deteriorated to ~90 minutes. Every large program had checkpoints to be able
to restart after the computer crashed.

>
> I/O: Card reader and punch.
>
> Size: 8 x 3 x 100 feet.
>
> Weight: 30 tons.
>
> Power: 150 KW.
>
> Cost: $500,000 in 1946 ($5.9 million today).
>
> http://en.wikipedia.org/wiki/ENIAC
>
> I knew a fellow Herb Berke who obtained two IBM-709s, and used one for
parts to keep the other running. I once spent an afternoon listening to his
repair stories, some of which were transferable to modern day
microcomputers!!!

For example, if anything went wrong with the CPU, the computer came down
hard. However, whenever the I/O controller stopped working, he would
typically find ~5 bad tubes in it!!! He explained that before dying, that
the I/O controller would usually slow the computer down, probably by making
errors and recovering from them.

The same situation now exists in modern-day "glue chips" that interface
CPUs to the rest on a modern-day laptop. Many of them operate with ongoing
malfunctions that impair but don't stop operation.

Brings back memories...

Steve



-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to