Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hardware of the Past

Author: David Blackman

Date: 03:15:32 05/16/01

Go up one level in this thread


For comparison purposes, the early Pentiums ran at 60 or 66 MHZ and got 40 to
100MIPS for chess purposes. For chess, all of these antiques are slower than a
Pentium 66. For some other purposes involving floating point, or heavy IO, a few
of the larger systems might beat a Pentium 66.

On May 16, 2001 at 00:55:22, Robert Hyatt wrote:

>On May 15, 2001 at 15:48:21, Joshua Lee wrote:
>
>>I was looking at one of my books on computer chess which covers the WCCC of 1974
>>and previous events and i wondered about that old hardware and how it compares
>>to lets say the fist pentiums. The list is :
>>Burroughs B5500
>
>very slow

About 1 MIPS was claimed, not sure if it was true. Very weird machine with a 47
bit word, stack based. The instruction set was optimised for small code size
rather than speed. The operating systems were multi-user multi-tasking, but
there was no memory protection in the hardware. The compilers were supposed to
generate code that never mangled other programs memory, and you weren't supposed
to use assembler. I never got to use one (i'm not sure if any were even sold in
my country) so i can't vouch for how well it all worked.

I didn't know anyone wrote a chess program for these.

>
>>Data General Nova 2, 800
>
>16 bit minicomputer.  < 1 mip.

Very cheap by 1970s standards and fairly popular. I know of one small business
that still used one of these for everything circa 1990! The cpu was pretty quick
for the money, getting close to 1 mips, the rest of the box was cheap and nasty.

>>Control Data 6400 , 6600
>
>6600 was very fast.  For scientific apps it will still likely zap today's
>micros.

No way. It peaked at about 3 MIPS and 3 marketing Megaflops. I don't have
Linpacks handy but probably about 1 Megaflops. Slightly slower than a Pentium-66
on most scientific stuff, and much slower for chess.

6400 was compatible with 6600 but cheaper and quite a bit slower. These were 60
bit machines targetting the scientific floating point market. Both models went
on sale in 1964 or 65 and were considered obsolete by 1974.

>>CDC Cyber 73, 74

I'm a bit hazy about these. There were two different series of Cybers. First the
6600 compatible machines, then the 64 bit cybers. If the NorthWestern program
ran on them, they must have been 6600 compatible. They ranged from about as fast
as a 6400 (and cheaper) up to several times faster than a 6600. Some had add-on
vector units that were comparible in speed to the Cray 1 and Cray XMP for long
vectors. I'm not sure exactly what the 73 and 74 were, but it's a fair bet they
were slower than a Pentium 66 for chess. (Possibly faster in some other
applications though).

>>Hewlett-Packard HP 3000

16 bit minicomputer. Horrible to use and program. Over the years there were many
versions of different speeds. The one i used was about 3 mips, but back in 1974
they were probably significantly slower than that.

>>HP 2100
Another 16 bit mini, i think. Slower than most of the 3000s?

>>Honeywell 635, 6060

Mainframes of up to 3 MIPS, but there were slower cheaper models also. I think
it was a modified 635 that ran Multics, quite a nice old OS. Otherwise they ran
horrible Honeywell systems. I think they had 36 bit words.

>>IBM 704, 7090, 360/65, 360/91, 370/145,370/155, 370/195

704 and 7090 were late 1950s 36 bit machines of about 1 MIPS or less.

The 360 and 370 series had many models and varied from very slow to a few MIPS
but were all mostly compatible with each other. IBM still makes systems
compatible with these! These were 32 bit machines. The 360 launched in 1964, the
370 maybe about 1970? These machines were incredibly successful commercially.
Most western governments and big businesses would use nothing else. Even in the
USSR they cloned the things.

>The /360 model 91 was very fast... same for the 195 /370 machine.
>They were mainframes with 100x the I/O thruput of today's micros, but
>not as fast cpu-wise.

The IO on these monsters was quite good, but not that good. A few MB/sec per
channel with potentially lots of channels. A well equiped modern micro has about
the same total IO throughput, not counting video. IBM claimed the CPUs were
faster than a CDC6600, but most people who used both disagree.

In certain real world applications these machines did amazingly well, running
whole government departments database systems for a large country on one
computer for instance. But this was mostly due to careful software design that
did exactly what was needed and nothing more. The actual hardware was nothing
special.

>>DEC PDP-6, PDP-10, PDP 11/45
>
>11/45 was dec's 16 bit mini...  nice box but very slow by today's
>standards.

PDP-6 was about 0.3 Mips 36 bit small mainframe. About 30 were made in the
mid-1960s but allegedly the one at MIT was the only one that actually worked,
thanks to fixes done by the students there.

PDP-10 was mostly compatible with the PDP-6, but much more reliable. Various
models ranged from about 0.5 MIPS to about 3 MIPS. Very nice machines, but
horribly slow by modern standards. These were very popular especially at
universities, and huge numbers of programmers who started in the 1970s learned
on a PDP-10. Greenblatt's MacHack was installed on many of the university
machines, usually playable only at night or on weekends. I played it a few
times.

11/45 was a 16 bit mini of about 0.8 mips. Quite nice except the small address
space. This was the classic UNIX machine for most of the 1970s, but DEC also
sold at least three of their own operating systems for it. On a modern pc of
about 200MHz or more, there are public domain simulators that will emulate a
11/45 faster than the real thing.

I once saw one of Ken Thompsons early chess programs running emulated like this.
Fascinating, but a very weak program. He obviously made a lot of software
improvements as well as getting special hardware for Belle.

>>GCS Alpha 16
?

>>MANIAC I

Truly ancient. Probably a few thousand instructions per second.

>>UNIVAC 1106,1110, 1108, 494, 418III
>
>the 1106, 08 and 10 were fast machines.  36 bit words, good FP, but not
>up to today's standards by a long way.
>
>>Varian 620i

>>Xerox Data Systems 940

Didn't Blitz run on one of these once? I don't know anything about it, but i
would have expected Bob to comment.

>>ICL 4/70, 1909/5
>>M-20
>
>The ICL was a piece of trash.

Yes. The uni i went to had one before they upgraded to the DEC PDP-10. Actually
the ICL was reasonably fast, maybe 2 MIPS or so, but it was so unreliable that
most of the scientific programs at our uni didn't have enough time to run
between crashes. Even ignoring that, it was very user unfriendly. Everyone was
very happy when they got the PDP-10, even though it was more expensive and not
necessarily faster.

>>
>>Any Info would be appreciated. Thankyou
>
>
>None of those machines approach today's speeds overall, although most would
>bury a micro in terms of I/O throughput today.  Many would bury a micro in
>terms of memory bandwidth also.

This was more or less true up to about 1990, but not any more. And it only
applies to the supers and some mainframes, not the minis.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.