Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
Re: No Glory for the PDP-15
78 views
Skip to first unread message
Patrick Scheible
Jun 11, 2012, 4:54:56 PM6/11/12
to
"Carl Appellof" <doct...@yahoo.com> writes:
> "Quadibloc" <jsa...@ecn.ab.ca> wrote in message
> news:f8dba46c-9133-4e7e...@c4g2000hsg.googlegroups.com...
>> There is a *reason* that the Soviets chose to copy the IBM 360 instead
>> of the CDC 6600!
>>
>
> Speaking of the CDC 6600. If you thought 36 bits was weird, how about 60
> bits? With 12-bit peripheral processors. Now THAT was a number cruncher!
> In my mind, one of the original RISC machines.
60 bits is not so weird. People who buy CDC machines want to crunch
lots of floating point numbers with reasonable precision. 32 bit
floating point is not enough precision for anything but games, and 36
bits is only beginning to be acceptable. A dedicated floating-point
number cruncher wants 60 bits. Don't waste address lines enabling the
programmer to address data smaller than the typical use.
> Speaking of Soviets, didn't DEC end up selling its PDP-11 business to a
> Russian PDP-11 clone maker? I swear I heard that when it happened, but
> can't track down the references now, except many acknowledgements that the
> Russians and other eastern block countries produced PDP-11 clones with wild
> abandon.
I remember eastern block clones of the PDP-11, but I didn't realize they
got any permission or assistance from DEC to make them.
-- Patrick
> "Quadibloc" <jsa...@ecn.ab.ca> wrote in message
> news:f8dba46c-9133-4e7e...@c4g2000hsg.googlegroups.com...
>> There is a *reason* that the Soviets chose to copy the IBM 360 instead
>> of the CDC 6600!
>>
>
> Speaking of the CDC 6600. If you thought 36 bits was weird, how about 60
> bits? With 12-bit peripheral processors. Now THAT was a number cruncher!
> In my mind, one of the original RISC machines.
60 bits is not so weird. People who buy CDC machines want to crunch
lots of floating point numbers with reasonable precision. 32 bit
floating point is not enough precision for anything but games, and 36
bits is only beginning to be acceptable. A dedicated floating-point
number cruncher wants 60 bits. Don't waste address lines enabling the
programmer to address data smaller than the typical use.
> Speaking of Soviets, didn't DEC end up selling its PDP-11 business to a
> Russian PDP-11 clone maker? I swear I heard that when it happened, but
> can't track down the references now, except many acknowledgements that the
> Russians and other eastern block countries produced PDP-11 clones with wild
> abandon.
I remember eastern block clones of the PDP-11, but I didn't realize they
got any permission or assistance from DEC to make them.
-- Patrick
Johnny Billquist
Jun 11, 2012, 6:11:14 PM6/11/12
to
On 2012-06-11 22:54, Patrick Scheible wrote:
The PDP-11 business was sold by DEC to Mentec. An Irish company.
Johnny
> "Carl Appellof"<doct...@yahoo.com> writes:
>
>> Speaking of Soviets, didn't DEC end up selling its PDP-11 business to a
>> Russian PDP-11 clone maker? I swear I heard that when it happened, but
>> can't track down the references now, except many acknowledgements that the
>> Russians and other eastern block countries produced PDP-11 clones with wild
>> abandon.
>
> I remember eastern block clones of the PDP-11, but I didn't realize they
> got any permission or assistance from DEC to make them.
Rather off-topic, but anyway, no...
>
>> Speaking of Soviets, didn't DEC end up selling its PDP-11 business to a
>> Russian PDP-11 clone maker? I swear I heard that when it happened, but
>> can't track down the references now, except many acknowledgements that the
>> Russians and other eastern block countries produced PDP-11 clones with wild
>> abandon.
>
> I remember eastern block clones of the PDP-11, but I didn't realize they
> got any permission or assistance from DEC to make them.
The PDP-11 business was sold by DEC to Mentec. An Irish company.
Johnny
Quadibloc
Jun 14, 2012, 2:59:07 PM6/14/12
to
On Jun 11, 2:54 pm, Patrick Scheible <k...@zipcon.net> wrote:
> "Carl Appellof" <doctor...@yahoo.com> writes:
> > "Quadibloc" <jsav...@ecn.ab.ca> wrote in message
computers with 30 bit words. The reason they had 30 bit words instead
of 32 bit words was because of fan-out limitations of the kind of
transistor circuit they were using - so going to 32 bits would have
required an extra stage of buffer circuitry, leading to a slower cycle
time.
A lot of number crunchers had a 48-bit word length, because that
allowed you 10-digit numeric precision, comparable to what was used on
scientific pocket calculators.
I've noted the howls of dismay that went up when the shift from the
7090 to the 360 led to single-precision floating-point going down from
36 bits to 32 bits, so I'll agree with the statement about 36 bits
being something of the low threshhold.
On the other hand, 64 bit double precision did tend to be thought of
as overkill, and so 60 bits would be pretty much just as good for the
cases where 48 bits is not enough - at least back in the classic
scientific computing environment.
However, the question really is whether you want a 60 bit word
combined with a 15-bit unit for instructions and a 6-bit character
size - instead of a 16-bit unit for instructions and an 8-bit
character size. 60 bits can be characterized as "weird" on the basis
that while it's a good length for floating-point numbers, it doesn't
mesh well with requirements for other data types.
John Savard
> "Carl Appellof" <doctor...@yahoo.com> writes:
> > "Quadibloc" <jsav...@ecn.ab.ca> wrote in message
> >news:f8dba46c-9133-4e7e...@c4g2000hsg.googlegroups.com...
> >> There is a *reason* that the Soviets chose to copy the IBM 360 instead
> >> of the CDC 6600!
> > Speaking of the CDC 6600. If you thought 36 bits was weird, how about 60
> > bits? With 12-bit peripheral processors. Now THAT was a number cruncher!
> > In my mind, one of the original RISC machines.
> 60 bits is not so weird. People who buy CDC machines want to crunch
> lots of floating point numbers with reasonable precision. 32 bit
> floating point is not enough precision for anything but games, and 36
> bits is only beginning to be acceptable. A dedicated floating-point
> number cruncher wants 60 bits. Don't waste address lines enabling the
> programmer to address data smaller than the typical use.
Back at Univac, Seymour Cray was involved in the design of some
> >> There is a *reason* that the Soviets chose to copy the IBM 360 instead
> >> of the CDC 6600!
> > Speaking of the CDC 6600. If you thought 36 bits was weird, how about 60
> > bits? With 12-bit peripheral processors. Now THAT was a number cruncher!
> > In my mind, one of the original RISC machines.
> 60 bits is not so weird. People who buy CDC machines want to crunch
> lots of floating point numbers with reasonable precision. 32 bit
> floating point is not enough precision for anything but games, and 36
> bits is only beginning to be acceptable. A dedicated floating-point
> number cruncher wants 60 bits. Don't waste address lines enabling the
> programmer to address data smaller than the typical use.
computers with 30 bit words. The reason they had 30 bit words instead
of 32 bit words was because of fan-out limitations of the kind of
transistor circuit they were using - so going to 32 bits would have
required an extra stage of buffer circuitry, leading to a slower cycle
time.
A lot of number crunchers had a 48-bit word length, because that
allowed you 10-digit numeric precision, comparable to what was used on
scientific pocket calculators.
I've noted the howls of dismay that went up when the shift from the
7090 to the 360 led to single-precision floating-point going down from
36 bits to 32 bits, so I'll agree with the statement about 36 bits
being something of the low threshhold.
On the other hand, 64 bit double precision did tend to be thought of
as overkill, and so 60 bits would be pretty much just as good for the
cases where 48 bits is not enough - at least back in the classic
scientific computing environment.
However, the question really is whether you want a 60 bit word
combined with a 15-bit unit for instructions and a 6-bit character
size - instead of a 16-bit unit for instructions and an 8-bit
character size. 60 bits can be characterized as "weird" on the basis
that while it's a good length for floating-point numbers, it doesn't
mesh well with requirements for other data types.
John Savard
Quadibloc
Jun 14, 2012, 3:39:26 PM6/14/12
to
Incidentally, I see that in the original thread, I gave a link to
http://www.quadibloc.com/arch/ar0506.htm
which didn't work because I mistakenly put .html on the end of the URL
instead of .htm .
John Savard
http://www.quadibloc.com/arch/ar0506.htm
which didn't work because I mistakenly put .html on the end of the URL
instead of .htm .
John Savard
glen herrmannsfeldt
Jun 14, 2012, 5:37:17 PM6/14/12
to
In alt.sys.pdp10 Quadibloc <jsa...@ecn.ab.ca> wrote:
(snip discussing 60 bit words and such)
> Back at Univac, Seymour Cray was involved in the design of some
> computers with 30 bit words. The reason they had 30 bit words instead
> of 32 bit words was because of fan-out limitations of the kind of
> transistor circuit they were using - so going to 32 bits would have
> required an extra stage of buffer circuitry, leading to a slower
> cycle time.
Missed it by that much, again...
I went to a talk some years ago (seemed like a regular seminar
talk, but actually a recruiting talk for the Cray-3.)
The talk was about the Cray-2, designed for a (then fast) 4ns clock.
To make it easier to distribute the 4ns clock throughout the
machine, they built it in a box resonant at that frequency.
When they started to put the machine together, it turned out
that the logic would only run at 4.2ns, but not 4.0ns.
But a 4ns resonant box isn't at all close to 4.2ns!
Set them back about a year fixing that one.
> A lot of number crunchers had a 48-bit word length,
> because that allowed you 10-digit numeric precision,
> comparable to what was used on scientific pocket calculators.
Except that most of that was before scientific pocket calculators.
The HP 9100 scientific desk calculator was around sometime
earlier than the HP 35, maybe with significant overlap
with 48 bit machines.
> I've noted the howls of dismay that went up when the shift from the
> 7090 to the 360 led to single-precision floating-point going down from
> 36 bits to 32 bits, so I'll agree with the statement about 36 bits
> being something of the low threshhold.
I believe 27 bit significand for the 7090, down to 24 (or 21
if you aren't lucky), yes.
But more generally, as data sizes get larger you often need
more bits to keep the same result precision. The shift to 64 bits
was at about the right time.
> On the other hand, 64 bit double precision did tend to be thought of
> as overkill, and so 60 bits would be pretty much just as good for the
> cases where 48 bits is not enough - at least back in the classic
> scientific computing environment.
Again, problems were getting bigger and more bits were needed.
A feature of Stretch allowed one to select whether 0's or 1's
are shifted in on floating point post-normalization. In other
words, round down or up. The idea was that one would do a
calculation both ways and compare the results. Interesting
idea, but it didn't continue.
> However, the question really is whether you want a 60 bit word
> combined with a 15-bit unit for instructions and a 6-bit character
> size - instead of a 16-bit unit for instructions and an 8-bit
> character size. 60 bits can be characterized as "weird" on the basis
> that while it's a good length for floating-point numbers, it doesn't
> mesh well with requirements for other data types.
Well, many machines stored smaller integers in a larger word.
And Fortran still requires the default INTEGER and REAL to
be the same size. (Makes EQUIVALENCE work right.)
The 704 through 7090 stored 16 bit signed integers in 36 bits.
(The original Fortran allowed statement numbers 1 to 32767,
which is where the five digits came from. Later up to 99999.)
Some of the larger word machines would add and subtract the
full word, but only multiply and divide for smaller values.
Well, the whole idea behind S/360 was scientific and business
in the same machine. Not to optimize for one or the other.
CDC and Cray went for the scientific market.
-- glen
(snip discussing 60 bit words and such)
> Back at Univac, Seymour Cray was involved in the design of some
> computers with 30 bit words. The reason they had 30 bit words instead
> of 32 bit words was because of fan-out limitations of the kind of
> transistor circuit they were using - so going to 32 bits would have
> required an extra stage of buffer circuitry, leading to a slower
> cycle time.
I went to a talk some years ago (seemed like a regular seminar
talk, but actually a recruiting talk for the Cray-3.)
The talk was about the Cray-2, designed for a (then fast) 4ns clock.
To make it easier to distribute the 4ns clock throughout the
machine, they built it in a box resonant at that frequency.
When they started to put the machine together, it turned out
that the logic would only run at 4.2ns, but not 4.0ns.
But a 4ns resonant box isn't at all close to 4.2ns!
Set them back about a year fixing that one.
> A lot of number crunchers had a 48-bit word length,
> because that allowed you 10-digit numeric precision,
> comparable to what was used on scientific pocket calculators.
The HP 9100 scientific desk calculator was around sometime
earlier than the HP 35, maybe with significant overlap
with 48 bit machines.
> I've noted the howls of dismay that went up when the shift from the
> 7090 to the 360 led to single-precision floating-point going down from
> 36 bits to 32 bits, so I'll agree with the statement about 36 bits
> being something of the low threshhold.
if you aren't lucky), yes.
But more generally, as data sizes get larger you often need
more bits to keep the same result precision. The shift to 64 bits
was at about the right time.
> On the other hand, 64 bit double precision did tend to be thought of
> as overkill, and so 60 bits would be pretty much just as good for the
> cases where 48 bits is not enough - at least back in the classic
> scientific computing environment.
A feature of Stretch allowed one to select whether 0's or 1's
are shifted in on floating point post-normalization. In other
words, round down or up. The idea was that one would do a
calculation both ways and compare the results. Interesting
idea, but it didn't continue.
> However, the question really is whether you want a 60 bit word
> combined with a 15-bit unit for instructions and a 6-bit character
> size - instead of a 16-bit unit for instructions and an 8-bit
> character size. 60 bits can be characterized as "weird" on the basis
> that while it's a good length for floating-point numbers, it doesn't
> mesh well with requirements for other data types.
And Fortran still requires the default INTEGER and REAL to
be the same size. (Makes EQUIVALENCE work right.)
The 704 through 7090 stored 16 bit signed integers in 36 bits.
(The original Fortran allowed statement numbers 1 to 32767,
which is where the five digits came from. Later up to 99999.)
Some of the larger word machines would add and subtract the
full word, but only multiply and divide for smaller values.
Well, the whole idea behind S/360 was scientific and business
in the same machine. Not to optimize for one or the other.
CDC and Cray went for the scientific market.
-- glen
Quadibloc
Jun 14, 2012, 5:44:36 PM6/14/12
to
On Jun 14, 3:37 pm, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:
rather that both numeric precisions derived from a common
computational requirement. I could, for example, have noted that
logarithm tables and the like, while they started at four or five
digits, and went up to seven digits in common use, sometimes went up
to ten digits (without getting too much longer, just using fancier
interpolation). Or noted the range of precisions of mechanical adding
machines.
John Savard
> In alt.sys.pdp10 Quadibloc <jsav...@ecn.ab.ca> wrote:
> > A lot of number crunchers had a 48-bit word length,
> > because that allowed you 10-digit numeric precision,
> > comparable to what was used on scientific pocket calculators.
>
> Except that most of that was before scientific pocket calculators.
Um, I wasn't trying to suggest a cause-and-effect relationship, but
> > A lot of number crunchers had a 48-bit word length,
> > because that allowed you 10-digit numeric precision,
> > comparable to what was used on scientific pocket calculators.
>
> Except that most of that was before scientific pocket calculators.
rather that both numeric precisions derived from a common
computational requirement. I could, for example, have noted that
logarithm tables and the like, while they started at four or five
digits, and went up to seven digits in common use, sometimes went up
to ten digits (without getting too much longer, just using fancier
interpolation). Or noted the range of precisions of mechanical adding
machines.
John Savard
glen herrmannsfeldt
Jun 14, 2012, 7:12:22 PM6/14/12
to
(snip)
so more digits might have been needed than for floating point.
It is an interesting question, why about 10 digits for the
usual calculator. Though 8 might have been usual, too.
The Marchant I remember had 8 digit entry, and 8 positions
of shifting for multiply/divide, so about 15 digit accumulator.
But much calculation was done to three or so digits on a slide
rule, so going to 10 for a calculator or computer doesn't seem
so obvious. I think 10 is about right for the HP 9100 though.
-- glen
>> > A lot of number crunchers had a 48-bit word length,
>> > because that allowed you 10-digit numeric precision,
>> > comparable to what was used on scientific pocket calculators.
(then I wrote)
>> > because that allowed you 10-digit numeric precision,
>> > comparable to what was used on scientific pocket calculators.
>> Except that most of that was before scientific pocket calculators.
> Um, I wasn't trying to suggest a cause-and-effect relationship,
> but rather that both numeric precisions derived from a common
> computational requirement. I could, for example, have noted that
> logarithm tables and the like, while they started at four or five
> digits, and went up to seven digits in common use, sometimes went up
> to ten digits (without getting too much longer, just using fancier
> interpolation). Or noted the range of precisions of mechanical adding
> machines.
The usual mechanical calculator, such as Marchant, was fixed point
> Um, I wasn't trying to suggest a cause-and-effect relationship,
> but rather that both numeric precisions derived from a common
> computational requirement. I could, for example, have noted that
> logarithm tables and the like, while they started at four or five
> digits, and went up to seven digits in common use, sometimes went up
> to ten digits (without getting too much longer, just using fancier
> interpolation). Or noted the range of precisions of mechanical adding
> machines.
so more digits might have been needed than for floating point.
It is an interesting question, why about 10 digits for the
usual calculator. Though 8 might have been usual, too.
The Marchant I remember had 8 digit entry, and 8 positions
of shifting for multiply/divide, so about 15 digit accumulator.
But much calculation was done to three or so digits on a slide
rule, so going to 10 for a calculator or computer doesn't seem
so obvious. I think 10 is about right for the HP 9100 though.
-- glen
John Levine
Jun 15, 2012, 12:23:56 AM6/15/12
to
>I've noted the howls of dismay that went up when the shift from the
>7090 to the 360 led to single-precision floating-point going down from
>36 bits to 32 bits, so I'll agree with the statement about 36 bits
>being something of the low threshhold.
Partly that was the shorter words, but equally much it was the botched
>7090 to the 360 led to single-precision floating-point going down from
>36 bits to 32 bits, so I'll agree with the statement about 36 bits
>being something of the low threshhold.
design of the 360 floating point which made the precision effectively
three bits shorter than it should have been.
>On the other hand, 64 bit double precision did tend to be thought of
>as overkill, and so 60 bits would be pretty much just as good for the
>cases where 48 bits is not enough - at least back in the classic
>scientific computing environment.
bit registers in the x86 float units.
R's,
John
PS: So what was wrong with the -15? It always looked like a nice
machine to me, although by the time it came out, the march to
8 bit bytes was well under way.
--
Regards,
John Levine, jo...@iecc.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. http://jl.ly
Patrick Scheible
Jun 15, 2012, 1:06:14 AM6/15/12
to
The computer crunched the numbers. The output was readable but not
pretty. If you needed them typed up in a nice looking report, you had a
typist do that afterwards, probably on a Selectric or on a Wang word
processor. Typists' time was cheaper than the computer's time.
-- Patrick
Harry Vaderchi
Jun 15, 2012, 5:47:43 AM6/15/12
to
On Fri, 15 Jun 2012 00:12:22 +0100, glen herrmannsfeldt
<g...@ugcs.caltech.edu> wrote:
> In alt.sys.pdp10 Quadibloc <jsa...@ecn.ab.ca> wrote:
>
> (snip)
>>> > A lot of number crunchers had a 48-bit word length,
>>> > because that allowed you 10-digit numeric precision,
>>> > comparable to what was used on scientific pocket calculators.
>
> (then I wrote)
>>> Except that most of that was before scientific pocket calculators.
>
>> Um, I wasn't trying to suggest a cause-and-effect relationship,
>> but rather that both numeric precisions derived from a common
>> computational requirement. I could, for example, have noted that
>> logarithm tables and the like, while they started at four or five
>> digits, and went up to seven digits in common use, sometimes went up
>> to ten digits (without getting too much longer, just using fancier
>> interpolation). Or noted the range of precisions of mechanical adding
>> machines.
>
> The usual mechanical calculator, such as Marchant, was fixed point
> so more digits might have been needed than for floating point.
>
> It is an interesting question, why about 10 digits for the
> usual calculator. Though 8 might have been usual, too.
>
12 would be better!
<g...@ugcs.caltech.edu> wrote:
> In alt.sys.pdp10 Quadibloc <jsa...@ecn.ab.ca> wrote:
>
> (snip)
>>> > A lot of number crunchers had a 48-bit word length,
>>> > because that allowed you 10-digit numeric precision,
>>> > comparable to what was used on scientific pocket calculators.
>
> (then I wrote)
>>> Except that most of that was before scientific pocket calculators.
>
>> Um, I wasn't trying to suggest a cause-and-effect relationship,
>> but rather that both numeric precisions derived from a common
>> computational requirement. I could, for example, have noted that
>> logarithm tables and the like, while they started at four or five
>> digits, and went up to seven digits in common use, sometimes went up
>> to ten digits (without getting too much longer, just using fancier
>> interpolation). Or noted the range of precisions of mechanical adding
>> machines.
>
> The usual mechanical calculator, such as Marchant, was fixed point
> so more digits might have been needed than for floating point.
>
> It is an interesting question, why about 10 digits for the
> usual calculator. Though 8 might have been usual, too.
>
3.14159265359
v.
3.1415926535897932384626433832795
> The Marchant I remember had 8 digit entry, and 8 positions
> of shifting for multiply/divide, so about 15 digit accumulator.
>
> But much calculation was done to three or so digits on a slide
> rule, so going to 10 for a calculator or computer doesn't seem
> so obvious. I think 10 is about right for the HP 9100 though.
>
> -- glen
[dash dash space newline 4line sig]
Albi CNU
Quadibloc
Jun 15, 2012, 12:44:25 PM6/15/12
to
It had its good points.DEC never really promoted it, though, and so
the amount of software for that architecture was limited. The price
was much higher than for a PDP-8, but the difference in power wasn't
as great.
Couple that with the fact that it had an "old-fashioned" architecture,
like the PDP-8, the Honeywell DDP 316/516, or the HP 211x, and it's
not surprising that there would be a lack of interest.
Compare that to the PDP-11, which had a modern and elegant
architecture (even if being little-endian was annoying to some people
like myself) and for which DEC was committed to making new peripherals
and new software.
Also, the PDP-15 was a single machine, while the PDP-11 was a range of
compatible machines, so you could get a small PDP-11 or a big PDP-11.
That meant the PDP-11 was a viable choice for more people, making for
a larger user community.
John Savard
the amount of software for that architecture was limited. The price
was much higher than for a PDP-8, but the difference in power wasn't
as great.
Couple that with the fact that it had an "old-fashioned" architecture,
like the PDP-8, the Honeywell DDP 316/516, or the HP 211x, and it's
not surprising that there would be a lack of interest.
Compare that to the PDP-11, which had a modern and elegant
architecture (even if being little-endian was annoying to some people
like myself) and for which DEC was committed to making new peripherals
and new software.
Also, the PDP-15 was a single machine, while the PDP-11 was a range of
compatible machines, so you could get a small PDP-11 or a big PDP-11.
That meant the PDP-11 was a viable choice for more people, making for
a larger user community.
John Savard
Johnny Billquist
Jun 15, 2012, 1:15:47 PM6/15/12
to
didn't it have a Unibus?
However, with those things said, I agree that DEC didn't push much for
it, nor did it get much attention compared to the PDP-11, either from
hardware or software, from DEC.
It was basically just an upsale to existing PDP-7/9 customers I guess.
Johnny
Rod Speed
Jun 15, 2012, 3:05:41 PM6/15/12
to
Quadibloc <jsa...@ecn.ab.ca> wrote
Yeah, that last is the real reason it never flew. The 11 left it for dead.
Even the 8 was left for dead by the 11.
Even the 8 was left for dead by the 11.
Rod Speed
Jun 15, 2012, 3:23:19 PM6/15/12
to
"Johnny Billquist" <b...@softjar.se> wrote in message
news:jrfqk3$nvf$1...@Iltempo.Update.UU.SE...
> On 2012-06-15 18:44, Quadibloc wrote:
>> On Jun 14, 10:23 pm, John Levine<jo...@iecc.com> wrote:
>>
>>> PS: So what was wrong with the -15?
>>
>> It had its good points.DEC never really promoted it, though, and so
>> the amount of software for that architecture was limited. The price
>> was much higher than for a PDP-8, but the difference in power wasn't
>> as great.
>>
>> Couple that with the fact that it had an "old-fashioned" architecture,
>> like the PDP-8, the Honeywell DDP 316/516, or the HP 211x, and it's
>> not surprising that there would be a lack of interest.
>>
>> Compare that to the PDP-11, which had a modern and elegant
>> architecture (even if being little-endian was annoying to some people
>> like myself) and for which DEC was committed to making new peripherals
>> and new software.
>>
>> Also, the PDP-15 was a single machine, while the PDP-11 was a range of
>> compatible machines, so you could get a small PDP-11 or a big PDP-11.
>> That meant the PDP-11 was a viable choice for more people, making for
>> a larger user community.
> Hmm. Wasn't there like two, or three models of the PDP-15 made?
Yes, but that was more just a variety of configs than
>> On Jun 14, 10:23 pm, John Levine<jo...@iecc.com> wrote:
>>
>>> PS: So what was wrong with the -15?
>>
>> It had its good points.DEC never really promoted it, though, and so
>> the amount of software for that architecture was limited. The price
>> was much higher than for a PDP-8, but the difference in power wasn't
>> as great.
>>
>> Couple that with the fact that it had an "old-fashioned" architecture,
>> like the PDP-8, the Honeywell DDP 316/516, or the HP 211x, and it's
>> not surprising that there would be a lack of interest.
>>
>> Compare that to the PDP-11, which had a modern and elegant
>> architecture (even if being little-endian was annoying to some people
>> like myself) and for which DEC was committed to making new peripherals
>> and new software.
>>
>> Also, the PDP-15 was a single machine, while the PDP-11 was a range of
>> compatible machines, so you could get a small PDP-11 or a big PDP-11.
>> That meant the PDP-11 was a viable choice for more people, making for
>> a larger user community.
> Hmm. Wasn't there like two, or three models of the PDP-15 made?
completely different machines like was seen with the 11.
> Also, didn't it have a Unibus?
http://www.pdpplanet.com/TemplateMain.aspx?contentId=52
> However, with those things said, I agree that DEC didn't push much for it,
> nor did it get much attention compared to the PDP-11, either from hardware
> or software, from DEC.
> It was basically just an upsale to existing PDP-7/9 customers I guess.
David Dyer-Bennet
Jun 15, 2012, 4:24:40 PM6/15/12
to
Quadibloc <jsa...@ecn.ab.ca> writes:
> A lot of number crunchers had a 48-bit word length, because that
> allowed you 10-digit numeric precision, comparable to what was used on
> scientific pocket calculators.
Check the dates on that. When 48-bit words were in vogue, there were no
> A lot of number crunchers had a 48-bit word length, because that
> allowed you 10-digit numeric precision, comparable to what was used on
> scientific pocket calculators.
pocket calculators yet, let alone scientific ones (which came out in the
early 1970s).
--
David Dyer-Bennet, dd...@dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info
Quadibloc
Jun 16, 2012, 2:48:59 PM6/16/12
to
On Jun 15, 2:24 pm, David Dyer-Bennet <d...@dd-b.net> wrote:
calculators, rather, I'm claiming that both got their numeric
precisions from the needs of ordinary computation in the sciences.
Thus, the vacuum-tube NORC computer had a numeric precision in this
general area.
John Savard
> Quadibloc <jsav...@ecn.ab.ca> writes:
> > A lot of number crunchers had a 48-bit word length, because that
> > allowed you 10-digit numeric precision, comparable to what was used on
> > scientific pocket calculators.
> Check the dates on that. When 48-bit words were in vogue, there were no
> pocket calculators yet, let alone scientific ones (which came out in the
> early 1970s).
Yes. I'm not claiming that 48-bit computers were influenced by pocket
> > A lot of number crunchers had a 48-bit word length, because that
> > allowed you 10-digit numeric precision, comparable to what was used on
> > scientific pocket calculators.
> Check the dates on that. When 48-bit words were in vogue, there were no
> pocket calculators yet, let alone scientific ones (which came out in the
> early 1970s).
calculators, rather, I'm claiming that both got their numeric
precisions from the needs of ordinary computation in the sciences.
Thus, the vacuum-tube NORC computer had a numeric precision in this
general area.
John Savard
Hans Vlems
Jun 17, 2012, 1:56:38 PM6/17/12
to
On 11 jun, 22:54, Patrick Scheible <k...@zipcon.net> wrote:
> "Carl Appellof" <doctor...@yahoo.com> writes:
> > "Quadibloc" <jsav...@ecn.ab.ca> wrote in message
> "Carl Appellof" <doctor...@yahoo.com> writes:
cloned the (micro)VAX.
Hans
Johnny Billquist
Jun 17, 2012, 7:52:01 PM6/17/12
to
Check the wikipedia article on the PDP-11 to get a fair list of clones
made in various east-european countries...
The VAX was probably also cloned several times, although it was a little
more complex, and so the clones were fewer. (Que the famous quote in
Russian on the CVAX-chip.)
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: b...@softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
0 new messages