Ok, what is the fastest pdp/11?
Bruce Oneel
bruce
--
| Bruce O'Neel | internet : on...@heasfs.gsfc.nasa.gov|
| Code 664/STX | span : lheavx::oneel |
| NASA/GSFC Bld 28/W281 |compuserve: 72737,1315 |
| Greenbelt MD 20771 | AT&Tnet : (301)-286-4585 |
Thats me in the corner, thats me in the spotlight, losin' my religion -- rem
Ted Sarbin
>Ok, what is the fastest pdp/11?
>
mor...@ramblr.enet.dec.com
>Ok, what is the fastest pdp/11?
>
Assuming one rules my previous answer (VAX 8650) invalid, either because it's
a VAX or it turns out to be not as fast in PDP emulation mode as a real PDP,
there are two contenders for fastest, depending on what you're doing.
One is the 11/70, the other the 11/9x machines. The 11/9x have the same
CPU and clock speed as the 11/8x, but its memory is basically 4mb cache, so
it is usually faster than the 11/8x.
The winner is: Neither!
There is a machine, the 11/74, which is approximately a 4 processor 11/70. Only
a few field test machines went out to the world before the project was
cancelled, however. At least one of them is still used within DEC (I even saw
it today).
-Mike
Steven M. Schultz
>One is the 11/70, the other the 11/9x machines. The 11/9x have the same
>CPU and clock speed as the 11/8x, but its memory is basically 4mb cache, so
>it is usually faster than the 11/8x.
I always thought the fastest 11 was the 11/55 - a 11/45 with
racing stripes :-)
Steven
Don Stokes
> Ok, what is the fastest pdp/11?
VAX 8650, at a guess.
I believe there have been a few experiments done with multiprocessor
11/70s.
The 11/94 is of course the fastest 11 which is actually marketed as
an 11.
Where's my chocolate fish?
Don Stokes, ZL2TNM / / d...@zl2tnm.gp.co.nz (home)
Systems Programmer /GP/ GP PRINT LIMITED Wellington, d...@gp.co.nz (work)
__________________/ / ---------------- New_Zealand__________________________
Don Stokes
> Still more interesting; what is the fastest PDP-8?
pdp11/60 with pdp8 microcode, used internally at DEC for DECmate
development (WPS-8, OS/278, COS-310 etc). That's *two* chocolate fish.
Ted Sarbin
That is quite correct. At one time all of the (remaining) PDP-8 programming
was done using RSTS/E on PDP-11/60s using microcode which simulated a PDP-8.
This microcode was written by an extremely talented programmer at DEC who
always approached a new machine by writing a PDP-8 simulator for it. He
similarly wrote one for the LSI-11 chipset. It worked quite well and did
get some people thinking about a product but nothing came of it.
At another time someone at DEC (Gordon Bell, I think) proposed a very high
speed PDP-8 using the same sort of arguments more recently used to support
RISC architectures. But, at that time PDP-11 mania had struck DEC and
noone was much interested in a new PDP-8.
John R Thurston
Indiana University under the design name of LD20,LD30,LD40...The PDP-8
is used as a basis for exploring new design and construction techniques,
so there is almost always a new PDP-8 implementation in the works.
Several years ago, I worked up designs for two PDP-8's (one cmos VLSI based,
the other TTL PAL based) that would blow the socks off anything DEC
made. I don't have my design notes here at work, but I could probably
find them again if I had to.
John Thurston jthu...@magnus.acs.ohio-state.edu
Andrew M. Boardman
>There is a machine, the 11/74, which is approximately a 4 processor
>11/70. Only a few field test machines went out to the world before the
>project was cancelled, however. At least one of them is still used
>within DEC (I even saw it today).
[This is coming out of mostly-dead brain cells; any correction or
clarification appreciated.] Wasn't there a kit that could be used, along
with four 11/70's, to create one of these beasts? I seem to recall
seeing such an item in someone's basement at some point, but I don't
remember who...
-a
Bill Pechter
>
>> Still more interesting; what is the fastest PDP-8?
>
>pdp11/60 with pdp8 microcode, used internally at DEC for DECmate
>development (WPS-8, OS/278, COS-310 etc). That's *two* chocolate fish.
>
>
I was told Reuters used the same "fast 8" in real production world wide
when DEC didn't make any faster 8's in the late '70s.
Bill
--
Bill Pechter | "The postmaster always pings twice."
Pyramid Technology | bi...@pyrite.nj.pyramid.com
10 Woodbridge Center Drive | rutgers!pyrnj!pyrite!bill
Woodbridge, NJ 07095 (908)602-6308 | pyramid!pyrnj!pyrite!bill
Nolan Hinshaw
Is that the notorious Ted Sarbin, formerly of DEC? Howdy from
someone who remembers the legend. I saw a pretty fast PDP-8 a
while ago. It was in the trunk of a Caddy on I5 down around
Firebaugh doing upwards of 95. Fastest one I ever saw.
--
Nolan Hinshaw Internet: no...@twg.com
The Wollongong Group Dingalingnet: (415)962-7197
Piobairi Uillean, San Francisco
Mise mo drumadoir eile fein!
Bill Pechter
>
> I always thought the fastest 11 was the 11/55 - a 11/45 with
> racing stripes :-)
>
> Steven
That's what my customers at Naval Air Propulsion said about their 11/55. It
beat the 11/70 at straight Fortran crunch.
Bill Pechter
>a few field test machines went out to the world before the project was
>cancelled, however. At least one of them is still used within DEC (I even saw
>it today).
>
I thought they decommissioned that machine in 1986. (I serviced an 11/71 -
well, what do you call a fourth of an 11/74 at DEC Princeton in 1985-6.)
Run the xxdp+ diags and see a KB11-CM processor type jump back out at you.
christophe m kalisiak
>In <1i121...@zl2tnm.gp.co.nz> d...@zl2tnm.gp.co.nz (Don Stokes) writes:
>>I believe there have been a few experiments done with multiprocessor
>>11/70s.
Yeah, the 11/74. I extracted something from here a little while ago:
In article <1991Apr14....@watdragon.waterloo.edu> tb...@watsol.waterloo.edu (Tim Bray) writes:
>And another PDP-11 operating system that no-one mentioned. TRAX-11 - a
>transaction processing monitor that DEC built and released, was a MASSIVE
>failure, unbearably slow and horribly buggy... it was withdrawn quickly.
>
>Further to the RSX-11's of this world. I was told that RSX-11M+ grew out of
>the PDP-11/74 project. Never heard of that? The PDP-11/74 was a 4x11/70
>multiprocessor, which, they said, was one helluva hot machine. It was
>cancelled, say the loyalists, out of a desire not to compete with this hot
>new 32-bit Virtual Address eXtension machine.
Yes and no. I actually heard TRAX just was too much for the 11/70 to handle
and that some of it's advances ended up in RSX11M.
The 11/74 was a great machine and it was used up until 1986 or so at DEC to
build all the RSX flavors in RSX development.
The problem with the 11/74 was maintenance. Diagnostic loading required that
the machine have part of the shared memory available for testing and the long
memory box cables were supposed to be a maintenance nightmare. DEC felt the
techs weren't up to it and dropped it as non-maintainable.
I tested and worked on an 11/74 in Princeton, well, uh 1/4 of an 11/74.
DEC split most of them up into pseudo 11/70's for use in house. The MK11
MOS memory later saw use on the 11/70 and the boards were used in the 11/750
and 11/730 as well.
There was an 11/74 follow up called the 11/68 which was multicpu, shared memory
with 22 bit addressing and software loadable microcode. Kind of the sun of
11/74 with 11/60 mixed in. This one never saw the light of day.
DEC used to have a notes file with the full history of the 11 series. I wish
it was public, it was fascinating.
Bill
ex-DEC Field Service
--
Bill Pechter | "The postmaster always pings twice."
Pyramid Technology | bi...@pyrite.nj.pyramid.com
10 Woodbridge Center Drive | rutgers!pyrnj!pyrite!bill
Woodbridge, NJ 07095 (908)602-6308 | pyramid!pyrnj!pyrite!bill
>>The 11/94 is of course the fastest 11 which is actually marketed as
>>an 11.
>
>That depends on what you are looking at. I'd say that an 11/70 would
>be faster on I/O, which means it would probably "feel" faster in
>a heavy load environment. If you are sitting alone, the 11/94 would
>be faster on numbercrunching.
I would find that hard to believe. The 11/84 is faster than the 11/70,
and the 11/94 is faster than the 11/84, so the 11/94 _must_ be faster
than the 11/70 in just about everything.
Also, being that the 11/94 has to go through a KTJ11 (?) Unibus to
Q-Bus adapter to talk to the peripherals, it is probable that the 11/93 is
marginally faster than the 11/94. The 11/93 can talk directly to the Q-Bus
memory and peripherals.
Chris Kalisiak
V076...@ubvms.cc.buffalo.edu
kali...@acsu.buffalo.edu
-- One of the few remaining PDP-11 buffs at UB...
Johnny Billquist
>on...@heawk1.rosserv.gsfc.nasa.gov ( Bruce Oneel ) writes:
>> Ok, what is the fastest pdp/11?
>VAX 8650, at a guess.
>I believe there have been a few experiments done with multiprocessor
>11/70s.
There appearantly was an in-house 11/72. A dual 11/70. Not marketed,
because it would be faster than an 11/780, which came at the same time.
(I think I read about it in "A DEC view on hardware design" or something
like it)
>The 11/94 is of course the fastest 11 which is actually marketed as
>an 11.
That depends on what you are looking at. I'd say that an 11/70 would
be faster on I/O, which means it would probably "feel" faster in
a heavy load environment. If you are sitting alone, the 11/94 would
be faster on numbercrunching.
Johnny
Tim Bray
stories. This would be about 1982, I was a DEC software service type,
spending most of my time trying to make overloaded 780's and 750's run
just a little faster, and persuading VMS to run (snicker) real-time.
Anyhow, I was up in Ottawa, where DEC had a big manufacturing plant that
was making backplanes, and I got a tour. About 3/4 of the floor in this
one bulding was full of immigrant women sitting in front of wire-wrap
machines - well, I guess it beat the textile mills. Off in a corner, they
had these 6 *huge* backplane wire-wrap machines. They looked like big
lathes or something - each had a pair of "hands" that grasped a backplane,
and one or two little wire-wrappers that actually did all the connections.
The "hands" and wire-wrap heads all were moving at once. The effect was
almost balletic and very impressive. Anyhow, they took us into a little
room to see the computers that were controlling these beasts and meet the
programmer. The programmer was a nice young fellow and I think glad of the
diversion. He led us to "the computers" which were - one (1) PDP-8. I
think he really enjoyed my look of dazed astonishment. How, I gasped, how
on earth could that little CPU possibly do that?
"Easy," he said. "No operating system to get in the way."
Don Stokes
> >>The 11/94 is of course the fastest 11 which is actually marketed as
> >>an 11.
Someone else wrote:
> >That depends on what you are looking at. I'd say that an 11/70 would
> >be faster on I/O, which means it would probably "feel" faster in
> >a heavy load environment. If you are sitting alone, the 11/94 would
> >be faster on numbercrunching.
kali...@acsu.buffalo.edu (christophe m kalisiak) writes:
> I would find that hard to believe. The 11/84 is faster than the 11/70,
> and the 11/94 is faster than the 11/84, so the 11/94 _must_ be faster
> than the 11/70 in just about everything.
Not necessarily.
CPU-wise, yes, the /94 is faster than the /70. I/O is another matter
entirely.
The 11/94, as you said, is actually a Qbus processor, which talks to
Unibus peripherals through an adapter.
The 11/70, on the other hand, has MASBUSS adapters, for talking to truly
wonderful devices like RM05s and RP06s (can you say "washing
machines"?). This combination makes for far faster I/O than any Unibus
or Qbus box out, in fact even CI equipped VAXes couldn't match it until
fairly recently (MASSBUSS equipped VAXes could, and also had things like
RP07s supported for even faster I/O).
christophe m kalisiak
>
[...]
>The 11/94, as you said, is actually a Qbus processor, which talks to
>Unibus peripherals through an adapter.
Right. With this in mind, I think the 11/93 would be faster.
>The 11/70, on the other hand, has MASBUSS adapters, for talking to truly
>wonderful devices like RM05s and RP06s (can you say "washing
>machines"?). This combination makes for far faster I/O than any Unibus
>or Qbus box out, in fact even CI equipped VAXes couldn't match it until
>fairly recently (MASSBUSS equipped VAXes could, and also had things like
>RP07s supported for even faster I/O).
Why was MASSBUS rendered obsolete, then? Was it purely because of
the size of the peripherals? The peripherals could have been replaced
with smaller, faster components, while still doing MASSBUS.
There must have been other reasons for getting rid of MASSBUS other
than the size of the MASSive disk/tape drives. What are they? Anything
to do with performance? I mean, you don't throw out a high speed
interface because the supported peripherals are huge...
Chris Kalisiak
V076...@ubvms.cc.buffalo.edu
kali...@acsu.buffalo.edu
Jan Brittenson
kali...@acsu.buffalo.edu (christophe m kalisiak) writes:
> Why was MASSBUS rendered obsolete, then?
Perhaps because it was replaced by designs better suited for
VAX-BI. You gotta be kidding if you're telling us there is *any*
comparison between a Q "toy" bus and an RH-70 equipped PDP-11/70
memory bus.
They are two entirely different beasts. Sure, a VAX SBI with MBAs
was a match, but *not* Q-bus!
> What are they? Anything to do with performance? I mean, you don't
> throw out a high speed interface because the supported peripherals are
> huge...
No, you you're right, you don't. But much has happened since the
RH-70. There are new designs that replace the SBI MBAs that early
VAXen were equipped with, better suited for upscale VAXen than MB ever
was. But MB was perfect for PDP-11, especially the RH-70 for the
PDP-11/70, which was the only PDP-11 with a 32-bit memory data bus.
The MB was the fastest device bus for the PDP-11, and the RH-70
controller was the fastest of them all. Almost twice as fast as the
ordinary RH-11, which by itself would run in circles around anything
you could sustain on a Q-bus machine.
In fact, the last time I asked, about two years ago, PDP-11/70s
with RH-70s were still worth real money. They are great for some
transaction-processing applications. A CPU that's not too slow, and
I/O that is capable of feeding it with data. Who on earth needs a
blitzing fast CPU that sits idle 75% of its time waiting for I/O to
complete?
-- Jan Brittenson
bs...@ai.mit.edu
mor...@ramblr.enet.dec.com
>was. But MB was perfect for PDP-11, especially the RH-70 for the
>PDP-11/70, which was the only PDP-11 with a 32-bit memory data bus.
Not counting the 32 bit wide memory bus the 11/83/84/93/94 have.
It's more than just a Q bus, folks.
-Mike
Mark Crispin
>Why was MASSBUS rendered obsolete, then? Was it purely because of
>the size of the peripherals? The peripherals could have been replaced
>with smaller, faster components, while still doing MASSBUS.
>There must have been other reasons for getting rid of MASSBUS other
>than the size of the MASSive disk/tape drives. What are they? Anything
>to do with performance? I mean, you don't throw out a high speed
>interface because the supported peripherals are huge...
Comparing MASSBUS with Unibus or Q-Bus is like comparing apples with
oranges. MASSBUS was a cabling and electrical protocol (called a
"bus") between a processor and peripherials, wherease Unibus or Q-Bus
were processor backplanes (also commonly called a "bus"). Think of
the difference between SCSI and NuBUS.
My two home DECSYSTEM-2020 systems have two UBAs (Unibus adaptor)
between the KS10 CPU backplane and two Unibuses (Unibi?). Each Unibus
has an RH11 channel which connects to a MASSBUS cable. One Unibus
contains only an RH11 and is used for disks; the other Unibus is a
complete BA11K assembly and has other peripherals on it (at least one
DZ11) as well as an RH11 used for tapes.
The PDP-11/70 had a different MASSBUS channel, the RH70, which was
faster than the RH11 and capable of supporting things such as RP07
disk drives and TU78 tape drives (interestingly, the TU78 is faster
than the RP07). Also, the KL10-based DECsystem-10 and DECSYSTEM-20
systems used RH20 channels which were also RP07/TU78 capable.
The main reason for retirement of the MASSBUS was that by modern
standards it has become a dinosaur. DEC's product strategy aimed
towards the CI. Additionally, in the outside world the focus moved
towards industry-standard interfaces such as SCSI. The MASSBUS was
highly proprietary to DEC and in fact DEC went to great efforts to
prevent other vendors from manufacturing MASSBUS equipment.
Finally, in defense of the poor RM05, I might say that it was a
helluva lot better than the RP06. An RM05 was 220V single-phase and
had 300MB capacity, compared to an RP06 which was three-phase and only
only held 200MB. The RM05 also used a good deal less power than the
RP06. For home use, the RM03 is more practical (110V single phase)
but with much less storage (67MB) at not that much of a power savings.
I use 1 (eventually to be 2) RM05 drive and 3 RM03 drives (2 of which
are normally not spinning) in my home configuration, which dual-ports
2 of the RM03 drives with the second CPU.
DEC never marketed the RM05 for the DECSYSTEM-20 or, I think, the
DECsystem-10. I guess they got a better deal from Memorex for RP06's
than they got from CDC for RM05's.
Warner Losh
>Sigh ... Lies, Damn lies and Benchmarks.....
Isn't that:
"Lies, Damn Lies, Statistics, Benchmarks and Release Dates"
I saw it in somebody's sig at bell labs about two years ago. I didn't
save the reference :-( A friend once commented that there was about an
order of magnitude difference between each step.
Warner
--
Warner Losh i...@Solbourne.COM
God is an Iron
Don Stokes
> In article <TPVg2...@zl2tnm.gp.co.nz> d...@zl2tnm.gp.co.nz (Don Stokes) wri
> >The 11/94, as you said, is actually a Qbus processor, which talks to
> >Unibus peripherals through an adapter.
>
> Right. With this in mind, I think the 11/93 would be faster.
That I'm not sure of -- the /9x processors use a private memory
interconnect (PMI); the adapter probably talks directly to that when
doing DMA, which would make the /94 faster than the /93. Whatever the
reason, the /94 is listed as being faster than the /93.
> Why was MASSBUS rendered obsolete, then? Was it purely because of
> the size of the peripherals? The peripherals could have been replaced
> with smaller, faster components, while still doing MASSBUS.
> There must have been other reasons for getting rid of MASSBUS other
> than the size of the MASSive disk/tape drives. What are they? Anything
> to do with performance? I mean, you don't throw out a high speed
> interface because the supported peripherals are huge...
To understand this, it helps if you know exactly what MASSBUS was, and
how it is integrated into the 11/70 system. A /70 block diagram looks
something like this:
+---------+ UNIBUS devices
| | | | |
| CPU |=======+===UNIBUS==+=======+=======+======+===+=+=+==>
| | | | | | |
+---------+ | | | | |
| UNIBUS I/O I/O I/O I/O
| +--------- MAP CTRLR CTRLR CTRLR CTRLR
| | | | | | | | | |
+-------+ | | | | | | | |
| CACHE |===32BIT=DMA=BUS===+=|=====+=|=====+=|====+ |
+-------+ | | | |
| | | | |
+-------+ | | | |
| MAIN | MASSBUS devices
| MEMORY|
+-------+
In short, MASSBUS is an adjunct to UNIBUS to provide fast DMA; it's not
an alternative to UNIBUS. As such, it was rather tied to the machines it
was developed for. Basically, it's a brute force UNIBUS speedup, using
a fast(er) 32bit path around the slower 16bit UNIBUS.
Early VAXes had MASSBUS available to them -- I'm not sure if it was
standard on early 11/780s; it was only an option on 11/750s. However,
the Way, the Truth and the Light of I/O, especially when VAXclusters
were invented, was the CI and HSC controllers with DSA disks. While
individual CI adapters didn't provide MASSBUS throughput, the CI/HSC
I/O system provided plenty of room to manoevre in terms of bandwidth
(MASSBUS couldn't get any/much faster) and provided flexibility that
MASSBUS didn't. The VAXBI bus came a little later and also provided
throughput that was nearly as good at a fraction of the price.
Don Stokes
> A CPU that's not too slow, and
> I/O that is capable of feeding it with data. Who on earth needs a
> blitzing fast CPU that sits idle 75% of its time waiting for I/O to
> complete?
Agreed there. I had a chuckle at a comparison between the VAX 9000 and
some new hot MIPS box in Digital Review a while ago. It claimed that
the two were the same speed, and implied that since the MIPS box was 10%
of the price of the 9000, it was far better value for money.
Of course when you looked at the configuration that the benchmarks were
run on, things looked a little odd:
VAX 9000 -- 256MB main memory, XMI bus etc etc
MIPS box, 32MB main memory, SCSI bus etc
While I'm not doubting that the MIPS box gives a lot of bang for buck, I
have to wonder which one is going to stay the distance when a *real*
load is applied, particuarly since the 9000 has been aimed squarely at
the TP market. I can see it now -- "MIPS box most popular TP machine of
the year"... yeah, right....
Sigh ... Lies, Damn lies and Benchmarks..... and from a "reputable"
publication too (although I think there are too many RISC/Unix bigots in
DR's staff).
Bill Pechter
>
>DEC never marketed the RM05 for the DECSYSTEM-20 or, I think, the
>DECsystem-10. I guess they got a better deal from Memorex for RP06's
>than they got from CDC for RM05's.
The official reason RM03/RM05's were never used on PDP's other than the 11/44 and
11/70 is that they were two fast for an RH11 on a loaded Unibus and had data
late problems in some configurations... so DEC wouldn't certify and support them.
We ran one on a PDP11/34 in Princeton's DEC office without problems.
I heard a rumor of a RM04 idea which was an RM05 slowed to 2400 RPM like the
RM02 was a downgraded RM03.
I guess they went against putting that out there. Many sites put third party
(Emulex or Systems Industries) controllers on with 9762/9766/Eagle drives.
I guess this didn't happen on KL sites.
Bill Pechter
>
>Early VAXes had MASSBUS available to them -- I'm not sure if it was
>standard on early 11/780s; it was only an option on 11/750s.
Even on 11/780's it was an option. The minimum 11/780 configuration was
an RK07 with an RK611 controller. Of course, the RL02 configuration existed
on the 11/730. The Unibus only Vax was crippled by poor I/O throughput
compared to the Massbus equipped machines.
Bill
Colin Dente
>[...] Who on earth needs a
>blitzing fast CPU that sits idle 75% of its time waiting for I/O to
>complete?
>
I've been saying this for ages - but all the salespeople ever seem to say is
'Err... <vacant smile> look how many MIPS this new box does...'
Wake up! ... throw off your chains! ...
God I hate SCSI... ;-)
Colin
--
Colin Dente | JANET: de...@uk.ac.man.ee.els
Manchester Computing Centre | ARPA: de...@els.ee.man.ac.uk
University of Manchester, UK | UUCP: ...!mcsun!ukc!manchester!dente
... I am the one you warned me of ...
David Charlap
>In article <15...@life.ai.mit.edu> bs...@rice-chex.ai.mit.edu (Jan Brittenson) writes:
>>[...] Who on earth needs a
>>blitzing fast CPU that sits idle 75% of its time waiting for I/O to
>>complete?
>>
You obviously do not multitask, do you? I've got a 10MHz 286, and when I
do work in Windows (most of the time now), I end up waiting for some process
that is printing or lost in deep thought (big spreadsheets do that). When
I use a similar setup at work on a 25MHz 386, this doesn't happen. There
is enough CPU available, that I can always launch a few more applications
while waiting for a process to complete.
--
David Charlap "Invention is the mother of necessity"
cd5...@mars.njit.edu "Necessity is a mother"
Operators are standing by "mother!" - Daffy Duck
Nick Felisiak
>>[...] Who on earth needs a
>>blitzing fast CPU that sits idle 75% of its time waiting for I/O to
>>complete?
>>
>Hear hear! (or is that here here!?)
>
>I've been saying this for ages - but all the salespeople ever seem to say is
>'Err... <vacant smile> look how many MIPS this new box does...'
Reminds me of a story. We were looking to buy some workstations
a couple of years ago. Various salescritters came through crying
'look how many MIPS this new box does ...'. One of them said something and
I wrote down '25Mhz 020'. Later, he says 'it's a 16.6 Mhz 020'.
'Hold on', says I, 'you said 4 mips...' (I may have the numbers wrong, but
the gist is right) ', that usually means 25Mhz'.
'Ah', he says, 'we have better compilers than everyone else'.
We didn't buy his box!
>
>God I hate SCSI... ;-)
SCSI will do around 3 Mb/sec (I don't have the exact number,
but it's 2 - 4!). Do you get anything like that from your applications?
SCSI-2 will go up to 20Mb/sec.
>
>Colin
>--
Nick
--
Nick Felisiak ni...@spider.co.uk
Spider Systems Ltd +44 31 554 9424
Jan Brittenson
cd5...@mars.njit.edu (David Charlap) writes:
>In article <24...@m1.cs.man.ac.uk> de...@els.ee.man.ac.uk (Colin Dente) writes:
>>In article <15...@life.ai.mit.edu> bs...@rice-chex.ai.mit.edu (Jan Brittenson) writes:
>>> [...] Who on earth needs a
>>> blitzing fast CPU that sits idle 75% of its time waiting for I/O to
>>> complete?
> You obviously do not multitask, do you? I've got a 10MHz 286, and ...
You bet your clone rear we multitask! :-)
Sure, the machine goes off to do something else while waiting for
the I/O to complete. But this doesn't help much in a multiuser system.
The time it takes for your process (the one you're sitting there
waiting for, not the printer spooler) to complete depends not only on
the time it takes to perform your program's I/O, but also the time it
takes to sit and wait for every *other* active user's I/O to get done!
The more users, the more demand paging (or swapping, if you prefer) -
even more I/O to sit and wait for.
Can you say "diskless SPARCstation"? That would be the joke of the
year.
> I use a similar setup at work on a 25MHz 386, this doesn't happen.
> There is enough CPU available, that I can always launch a few more
> applications while waiting for a process to complete.
Right... Try running three database, say Informix, applications
simultaneously on your 386 system, in different windows. Watch them
fight over the bus and disks while the CPU is idle.
-- Jan Brittenson
bs...@ai.mit.edu
Mark James
In article <1991May4.2...@watdragon.waterloo.edu> tb...@watsol.waterloo.edu (Tim Bray) writes:
>
>He led us to "the computers" which were - one (1) PDP-8. I
>think he really enjoyed my look of dazed astonishment. How, I gasped, how
>on earth could that little CPU possibly do that?
>
>"Easy," he said. "No operating system to get in the way."
Right on! I once wrote a PDP-8 interface program for a thingamajigger
that read electrocardiograms in over the phone. The program was sort
of like a TSR, in PC jargon; the first thing that it did when a call
came in was to swap OS/8 out. The system just got in the way.
--
=== T. Mark James ==== All opinions, errors etc are my own.
=== ma...@bdblues.altair.fr ==== "Hardware is that part of a computer
=== +33 (1) 39 63 53 93 ==== system that you can kick."
================================ -- Grace Hopper
Tom Neff
> >>> blitzing fast CPU that sits idle 75% of its time waiting for I/O to
> >>> complete?
In article <15...@life.ai.mit.edu> bs...@rice-chex.ai.mit.edu (Jan Brittenson) writes:
> Right... Try running three database, say Informix, applications
>simultaneously on your 386 system, in different windows. Watch them
>fight over the bus and disks while the CPU is idle.
Nevertheless, the answer is a faster bus, not a slower CPU.
--
I might have gone to West Point, but I was too / /\ Tom Neff
proud to speak to a Congressman. -- Will Rogers /__/ \ tn...@bfmny0.BFM.COM
Colin Dente
>In article <24...@m1.cs.man.ac.uk> de...@els.ee.man.ac.uk (Colin Dente) writes:
>>God I hate SCSI... ;-)
>
>SCSI will do around 3 Mb/sec (I don't have the exact number,
>but it's 2 - 4!). Do you get anything like that from your applications?
>
>SCSI-2 will go up to 20Mb/sec.
>
Ok - what I really should have said is `God I hate the way every time
my colleague needs to look at a cartridge tape, I suddenly can't
access the disk anymore because of the totally dumb-ass way this
machine is set up so the cartridge drive ties up the SCSI bus'. Does
it stand for `Starting Cartridge Stops Interaction'?... ;-)
Having said that - some of the applications for which I am responsible
(I just write the software - thank God) are entirely IO bound - We
distribute software to all universities and polytechnics in the UK
from Manchester (lots of it too).
James Thomas
bqt> This gets deeper...
bqt> The massbus was originally developed for the pdp-10, I think, and
bqt> the pdp-10 is 36 bits, so massbus can/is 36 bits too.
bqt> I'm not sure how the disks are connected to a DECsystem-10 or
bqt> a DECSYSTEM-20, but I think they don't pass through the front -11,
bqt> since the -11 has its own connection to the boot disk, and the highspeed
bqt> disks cannot be handled by an 11/40.
Digging back into over-hysteresised (sp?) cores....
I think I would doubt that MASSBUS started on the -10 . Weren't all the
device registers in the controllers 16 bits wide ?
Jim
Johnny Billquist
>To understand this, it helps if you know exactly what MASSBUS was, and
>how it is integrated into the 11/70 system. A /70 block diagram looks
>something like this:
[Diagram deleted]
Hokay, time for me to tread DEEP water. Forgive me for the many wrong
things I'm about to say...
I'm not sure the MASSBUS was default in the 11/70. The 11/70 had
a special unibus -> massbus adapter. All other pdp-11 used a slower
adapter, which couldn't take things like the RP07, but the 11/70
could.
>In short, MASSBUS is an adjunct to UNIBUS to provide fast DMA; it's not
>an alternative to UNIBUS. As such, it was rather tied to the machines it
>was developed for. Basically, it's a brute force UNIBUS speedup, using
>a fast(er) 32bit path around the slower 16bit UNIBUS.
This gets deeper...
The massbus was originally developed for the pdp-10, I think, and
the pdp-10 is 36 bits, so massbus can/is 36 bits too.
I'm not sure how the disks are connected to a DECsystem-10 or
a DECSYSTEM-20, but I think they don't pass through the front -11,
since the -11 has its own connection to the boot disk, and the highspeed
disks cannot be handled by an 11/40.
>Early VAXes had MASSBUS available to them -- I'm not sure if it was
>standard on early 11/780s; it was only an option on 11/750s.
I think it was optional, but almost mandatory, since all big disks
were massbus at that time. I don't know what the biggest unibus disk
was, but it could have been something like the RK07 (28Mb).
>Don Stokes, ZL2TNM / / d...@zl2tnm.gp.co.nz (home)
>Systems Programmer /GP/ GP PRINT LIMITED Wellington, d...@gp.co.nz (work)
>__________________/ / ---------------- New_Zealand__________________________
I just looove to talk old dec hardware, if you haven't noticed...
Johnny
Don Stokes
> I'm not sure the MASSBUS was default in the 11/70. The 11/70 had
> a special unibus -> massbus adapter. All other pdp-11 used a slower
> adapter, which couldn't take things like the RP07, but the 11/70
> could.
Well, yes and no. There were MASSBUS disk controllers that plugged into
UNIBUS machines, but the 11/70 had a dedicated I/O bus into which could
be plugged MASSBUS peripheral controllers. Thus, the 32 bit path to
memory only existed on the 11/70 (in the pdp11 world). I'm not aware of
an equivalent in the 10 world (-10s were never big in this country like
11s and VAXes), but the VAXes also had (optional) MASSBUS adapters that
attched directly to the CPU bus (SBI). The UNIBUS adapters didn;t support
the RM80 (fixed HDA) or the RM05 (256MB removable washing machine). Only
the VAX 11/780 MBA supported the RP07 fixed HDA 512MB drive.
I don't think it's really fair to say that UNIBUS machines with MASSBUS
disks attached really qualify as "MASSBUS equipped", as they were still
restricted by the UNIBUS.
> I think it was optional, but almost mandatory, since all big disks
> were massbus at that time. I don't know what the biggest unibus disk
> was, but it could have been something like the RK07 (28Mb).
How do you define "unibus disk"? I can plug any RA drive into a UDA50
(attached to Unibus) and it'll work, and I can have more space on a
single spindle than any MASSBUS device. 'course DSA wasn't out in the
days of early 11/780s and 11/70s. Mind you, I never saw an 11/70 without
MASSBUS disks.... (some also had DSA disks added more recently.
Don Stokes
> Did you know that RSTS/E knows how to talk to RP07's? (Note I said "talk
> to", since "supports" implies certain contractual obligations by DEC 8-).
No, I didn't know, but I'm not surprised. RSTS/E is chocka full of
wonderful things in the I/O department -- one version of BACKUP exposed
bugs in some model or other of tape controller, because no-one had ever
driven it that fast before. The fastest directory program I ever saw
ran on RSTS (it's called "DIR", as opposed to PIP which is/was what the
DCL DIR command ran).
> I have an 11/70 with 1 RP07 and 3 RM05's on it, running RSTS/E. You have
> to use the RP07 in interleaved mode (as do most VAXen), but if you install
> the PEP-70 and Hypercache, you can actually run it in 1/1 mode. Quite a
> screamer!
<envy> 8-)
Just having a squiz at the RP07 spex:
516MB:
9 platters
16 data surfaces + 1 servo
32 data heads + 1 servo
630 cylinders per head + 2 diagnostic
50 sectors
512 bytes per sector plus 128 bit header + 32 bit CRC
3633 rpm
1300 KB/s or 2200 KB/s with RP07-D option (allows 1/1 interleave?)
5ms track-to-track seek
23ms average seek
Nice drive, big power eater and noise maker I should think -- 8A at
240Vac, with a surge current of 70A -- do the lights in your neighbourhood
dim when you spin it up, Terry?
Terry Kennedy, Operations Mgr.
> Nice drive, big power eater and noise maker I should think -- 8A at
> 240Vac, with a surge current of 70A -- do the lights in your neighbourhood
> dim when you spin it up, Terry?
It plugs into a 15A 3-phase outlet (same as is used for the wimpy RP06).
We have 1500A 3-phase in our computer room, stepped down from the 22Kv main
feed by a large humming transformer behind the building. Thus, there is no
effect on the neighbors 8-).
It's also very quiet - a properly balanced RP07 is quieter than an RA81,
at least in the audible range. Remember, it's in a full acoustic enclosure.
The drive was originally designed (by ISS/Sperry) for the IBM 370-class
market (hence the unusual front panel design). If they didn't shield it well,
a roomful would be deafening (since an IBM shop might purchase 32 or so of
these).
Trivia comment: The drive logic is run by an 8085 microprocessor. The Mass-
bus interface is run by 2901's. Says something about Massbus timing require-
ments...
Terry Kennedy Operations Manager, Academic Computing
te...@spcvxa.bitnet St. Peter's College, US
te...@spcvxa.spc.edu (201) 915-9381
Anthony A. Datri
>(attached to Unibus) and it'll work, and I can have more space on a
>single spindle than any MASSBUS device. 'course DSA wasn't out in the
>days of early 11/780s and 11/70s. Mind you, I never saw an 11/70 without
>MASSBUS disks.... (some also had DSA disks added more recently.
Anyone out there remember the RP20?
Anyone ever have one go for a year without crashing?
--
Fly to the sky on GI-GI____________ and shout to
da...@convex.com
Walter C. Daugherity
The PDP-11/60 was specified to require 3-phase power, presumably to allow
3-phase disk drives to be attached, but the CPU itself did not require it.
So, IF (note the big IF) you had 1-phase disk drives like RM03's AND (note
the big AND) huge conductors, then you could tie the three phases together
at the computer and feed it with a single-phase hot wire!
Trivia question: What was the difference between RM02's and RM03's?
Hacker test: Did you ever modify the writable control store to invent new
instructions?
--
Walter C. Daugherity Internet, NeXTmail: dau...@cs.tamu.edu
Texas A & M University uucp: uunet!cs.tamu.edu!daugher
College Station, TX 77843-3112 BITNET: DAUGHER@TAMVENUS
---Not an official document of Texas A&M---
Terry Kennedy, Operations Mgr.
> The 11/70, on the other hand, has MASBUSS adapters, for talking to truly
> wonderful devices like RM05s and RP06s (can you say "washing
> machines"?). This combination makes for far faster I/O than any Unibus
> or Qbus box out, in fact even CI equipped VAXes couldn't match it until
> fairly recently (MASSBUSS equipped VAXes could, and also had things like
> RP07s supported for even faster I/O).
Did you know that RSTS/E knows how to talk to RP07's? (Note I said "talk
to", since "supports" implies certain contractual obligations by DEC 8-).
I have an 11/70 with 1 RP07 and 3 RM05's on it, running RSTS/E. You have
to use the RP07 in interleaved mode (as do most VAXen), but if you install
the PEP-70 and Hypercache, you can actually run it in 1/1 mode. Quite a
screamer!
Terry Kennedy Operations Manager, Academic Computing
Don Stokes
> It plugs into a 15A 3-phase outlet (same as is used for the wimpy RP06).
> We have 1500A 3-phase in our computer room, stepped down from the 22Kv main
> feed by a large humming transformer behind the building. Thus, there is no
> effect on the neighbors 8-).
Hmmm. Our 60kVA power conditioner gives up in disgust if I try to cold
start three VAXes and 18 RA81s & 82s (and HSCs, and other boat-anchors)....
> It's also very quiet - a properly balanced RP07 is quieter than an RA81,
That's *not* hard. RA81s are rattly things at the best of times, much
worse if your friendly Field Servoid fails to attach the motor properly
and the cabinet starts shaking, accompanied by much graunching and
unpleasantness until it gives up and shuts down.....
Joe Morris
>The PDP-11/60 was specified to require 3-phase power, presumably to allow
>3-phase disk drives to be attached, but the CPU itself did not require it.
>So, IF (note the big IF) you had 1-phase disk drives like RM03's AND (note
>the big AND) huge conductors, then you could tie the three phases together
>at the computer and feed it with a single-phase hot wire!
One of the problems I had with a KL-10 installation was that the DEC
CE's had no understanding of AC power systems. They complained repeatedly
about the (admittedly poor) quality of the AC power we supplied, but
couldn't understand why I threw a temper tantrum when I found that they
were dumping mucho amps down the *safety ground* wire (completely in violation
of the electrical code)...which elevated local ground to about +7 volts
above the rest of the campus, making for interesting communications
problems.
However stodgy its designs may be (and frequently are), IBM's mainframe
boxes are designed to use only 208V, single or three-phase power. This
means that you don't even have a neutral wire in the power cable (hot leads
plus safety ground only).
Since we're on the subject of AC power...we got an unpleasant surprise a
few years ago when we took delivery of a new Sun 3/280: somewhere along
the line the spec we made for 208V power was dropped and Sun delivered
a system which wanted 230 VAC single-phase power. Turns out that if
you say you may want to put in a second disk then Sun won't ship anything
*but* 230 VAC systems.
The problems was that this box was for use in a computer room which (like
most) has 208 VAC power...not 230. Sun's first reaction was to tell us
to try it on 208 VAC (but said that this would void the warranty); I
declined. We eventually resolved the problem by putting in a series
22 VAC boost transformer in a 208 VAC line.
Apparently Sun's designers were still assuming that every user of their
boxes was running in a storefront office which was fed by a single-phase
230 VAC line, center-tapped to give 115 VAC. (Just like your house.)
They apparently had no intention of configuring their boxes for 208 VAC
operation.
We were more than a little gratified to learn a few weeks later that
the local Sun sales office had ordered a 3/280 for its own use...and then
discovered that the office building couldn't supply 230 VAC. Just 208.
(A quick summary of AC Engineering 101 for readers: most commercial
buildings with three-phase power distribute all three phases to the
distribution panels, plus a neutral wire. You can get 208 VAC between
any two of the three phase leads, or 110 VAC between any phase lead
and the neutral wire. In a properly-balanced installation the loads on
the phase leads should be perfectly balanced, and the neutral wire
should have zero current. Of course, a properly-written operating system
has zero defects. These situations are equally probable.)
Joe Morris
Bill Pechter
>
>Trivia question: What was the difference between RM02's and RM03's?
They slowed down the RPM's of the drive from 3600 to 2500 (or was it 2600)
to avoid swamping the Unibus bandwidth on non-11/70 11/44 machines.
That's why they never ran RM05's on 11/34s. (Rumors of an RM04 were never
proven...)
Bill
ex-DEC Field Service
Bill Stewart 908-949-0705 erebus.att.com!wcs
]The fastest PDP-8's? Those are built by the computer science dept of
]Indiana University under the design name of LD20,LD30,LD40...The PDP-8
]is used as a basis for exploring new design and construction techniques,
]so there is almost always a new PDP-8 implementation in the works.
Those may have been the fastest HARDWARE PDP-8s, but there was a
PDP-8 simulation done in SOFTWARE on a PDP-10 by David Moon.
Forgot where I read about it, whether it was HAKMEM or just some of
the .unix-wizards discussion about programming languages.
Basically, since the PDP-8 instructions were 12 bits,
he just took an array of 4096 addresses and did a
goto array[instruction] /* however that's expressed in MACRO-11 */
which jumped to a hunk of code that did the job, then looped back to
the get-next-instruction code.
Apparently it was faster than a real PDP-8.
Any ideas how this compared to the speed of the newer hardware
that's been mentioned above?
--
Pray for peace; Bill
# Bill Stewart 908-949-0705 erebus.att.com!wcs AT&T Bell Labs 4M-312 Holmdel NJ
# I never wanted to be a hacker! I wanted to be --- a lumberjack!
sys...@iowasp.physics.uiowa.edu
>
> 516MB:
> 9 platters
> 16 data surfaces + 1 servo
> 32 data heads + 1 servo
> 630 cylinders per head + 2 diagnostic
> 50 sectors
> 512 bytes per sector plus 128 bit header + 32 bit CRC
> 3633 rpm
> 1300 KB/s or 2200 KB/s with RP07-D option (allows 1/1 interleave?)
> 5ms track-to-track seek
> 23ms average seek
>
to support the 2.2Mb rate (Early MS780's weren't interleaved so wouldn't
keep up with the faster rate...)
Willy
James Thomas
In article <bqt.67...@cia.docs.uu.se> b...@cia.docs.uu.se (Johnny Billquist) writes:
bqt> This gets deeper...
bqt> The massbus was originally developed for the pdp-10, I think, and
bqt> the pdp-10 is 36 bits, so massbus can/is 36 bits too.
bqt> I'm not sure how the disks are connected to a DECsystem-10 or
bqt> a DECSYSTEM-20, but I think they don't pass through the front -11,
bqt> since the -11 has its own connection to the boot disk, and the highspeed
bqt> disks cannot be handled by an 11/40.
jim> Digging back into over-hysteresised (sp?) cores....
jim> I think I would doubt that MASSBUS started on the -10 . Weren't all the
jim> device registers in the controllers 16 bits wide ?
bqt also wrote:
bqt> I'm not sure the MASSBUS was default in the 11/70. The 11/70 had
bqt> a special unibus -> massbus adapter. All other pdp-11 used a slower
bqt> adapter, which couldn't take things like the RP07, but the 11/70
bqt> could.
And don wrote:
don> Well, yes and no. There were MASSBUS disk controllers that plugged into
don> UNIBUS machines, but the 11/70 had a dedicated I/O bus into which could
don> be plugged MASSBUS peripheral controllers. Thus, the 32 bit path to
don> memory only existed on the 11/70 (in the pdp11 world).
and also:
don>In short, MASSBUS is an adjunct to UNIBUS to provide fast DMA; it's not
don>an alternative to UNIBUS. As such, it was rather tied to the machines it
don>was developed for. Basically, it's a brute force UNIBUS speedup, using
don>a fast(er) 32bit path around the slower 16bit UNIBUS.
Sorry about this, but if we go look back at page 278 ff of "Computer
Engineering" by Bell, et al., we see something a bit different.
The Massbus was developed to hook big fast (:-) disks to the RH11 (which
bacame the RH70 on the 11/70 and the RH10 on the 1070) for the 11/45. They
wanted to hook the RH11 to a private Unibus, but that didn't work. The
Massbus goes between the devices and the RH controller. Something else
connects the controller to the memory - a poor Unibus for the 11/45 and a
special 32-bit "backplane data path" on the 11/70 .
Jim
Mark Crispin
There's the alternate problem. DEC assumes that everyone has
three-phase power in a wye (not delta) configuation. Consequently,
everything that needs more than 110V is 208V or three-phase.
So, if you have single phase 115/230V (which can be from 110/220V up
to 125/250V in the US) you have to adjust the power supply on things
like RM05's or TU77's to make it handle the hotter voltage without
frying. It can be done, it's just something you don't think of doing
(especially when the device has a mendacious label that says it would
swallow 230V).
What fun!
R. Kym Horsell
[PDP-8 simulator on a DECsystem-10]
>Apparently it was faster than a real PDP-8.
>Any ideas how this compared to the speed of the newer hardware
>that's been mentioned above?
Well I think it'd be a pretty poor comparison. I think the KI that I
first used was somewhere around 1 MIPS (and _THAT_ may be streatching it
a bit). I guess a KL would've been a bit faster.
As has been pointed out the little PDP8 class projects have been based on
almost everything from hand-strung VLSI to n-bit slices to gates. The
examples I've personally seen (that _almost_ worked :-) did around 20 MIPS.
(If your department is willing to pay for _new_ parts you might get
50 MIPS in ECL or somesuch).
-kym
John Simutis
Both were re-packaged CDC (9762?) 80 mb/67.5 mb DEC formatted, but
the RM03 was a MASSBUS device!
I keep remembering a 67.5 mb SMD drive + controller at $19,000; I think
that's about what one pays for an RA90, 3366612 512-byte blocks today...
Nostalgia isn't what it used to be.
--
--------
< standard disclaimer >
John Simutis, simutis @ingres.com Alms! Alms for the vi-impaired!
P E Smee
>PDP-8 simulation done in SOFTWARE on a PDP-10 by David Moon.
>he just took an array of 4096 addresses and did a
> goto array[instruction] /* however that's expressed in MACRO-11 */
>which jumped to a hunk of code that did the job, then looped back to
>the get-next-instruction code.
Hmm, reminds me of the PDP-8 simulator I once wrote, to aid with a
hairy class of debugging problems. Except that my jump table was based
on the fact that the opcode portion of an instruction used only 3
bits. (That's right, 8 opcodes -- 6 of which had address fields, the
other 2 of which used the address field to microcode various operations
on the register(s) and I/O devices.)
What made mine fun was that it was written in Macro 8, to simulate a
PDP-8 on a PDP-8. Sloooow? You bet.
The object of this exercise was to allow (by program patching of the
simulator) me to set breakpoints based on ideas like 'stop when you
catch the d****d instruction which is overwriting loc XXX with value
YYY'. A couple of years later, someone came out with a bus
debugging-card which allowed the same style of checking to be done
using switches on the card, and so my simulator passed into history.
Where it probably should have stayed all along.
--
Paul Smee, Computing Service, University of Bristol, Bristol BS8 1UD, UK
P.S...@bristol.ac.uk - ..!uunet!ukc!bsmail!p.smee - Tel +44 272 303132
Terry Kennedy, Operations Mgr.
> The PDP-11/60 was specified to require 3-phase power, presumably to allow
> 3-phase disk drives to be attached, but the CPU itself did not require it.
> So, IF (note the big IF) you had 1-phase disk drives like RM03's AND (note
> the big AND) huge conductors, then you could tie the three phases together
> at the computer and feed it with a single-phase hot wire!
Of course, the neutral was designed to carry only the current resulting from
any phase imbalances, not the full load. Thus, doing this may result in exces-
sive current in the neutral lead. I'd suggest going to at least #10, probably
#8, before doing this.
> Trivia question: What was the difference between RM02's and RM03's?
Would you settle for 2400/3600ths of an answer 8-)
Don Stokes
> In article <16...@helios.TAMU.EDU> dau...@cs.tamu.edu (Walter C. Daugherity)
> >Trivia question: What was the difference between RM02's and RM03's?
>
> They slowed down the RPM's of the drive from 3600 to 2500 (or was it 2600)
> to avoid swamping the Unibus bandwidth on non-11/70 11/44 machines.
> That's why they never ran RM05's on 11/34s. (Rumors of an RM04 were never
> proven...)
2400rpm, like the RL & RK drives.
Joe Morris
>In article <16...@helios.TAMU.EDU>, dau...@cs.tamu.edu (Walter C. Daugherity) writes:
>> The PDP-11/60 was specified to require 3-phase power, presumably to allow
>> 3-phase disk drives to be attached, but the CPU itself did not require it.
>> So, IF (note the big IF) you had 1-phase disk drives like RM03's AND (note
>> the big AND) huge conductors, then you could tie the three phases together
>> at the computer and feed it with a single-phase hot wire!
> Of course, the neutral was designed to carry only the current resulting from
>any phase imbalances, not the full load. Thus, doing this may result in exces-
>sive current in the neutral lead. I'd suggest going to at least #10, probably
>#8, before doing this.
For years the NEC permitted buildings to be wired using a neutral conductor
size of 50% of the size of the phase conductors on the assumption that the
phase loads would average out and the neutral wouldn't need to handle much
current.
The code was updated recently to require that the neutral be sized at 200% of
phase conductor size. The change is required because of all of the very
dirty (electrically) loads which have become common in the past decades.
There have been documented situations where the three phase conductors are
carrying equal current loads ... and the neutral is carrying a higher
current. With sufficient harmonic content, you can have the three phase
legs carrying 100A apiece and still have > 100A in the neutral.
Odd but true: a major contributor of dirty harmonics in a building is
the standard static UPS. Most of them have switching power supplies and
don't have any filters on the AC input to prevent the harmonics from
getting back into the building power lines.
Joe Morris
Joe Morris
>There's the alternate problem. DEC assumes that everyone has
>three-phase power in a wye (not delta) configuation. Consequently,
>everything that needs more than 110V is 208V or three-phase.
Uh? The standard building transformer *does* use a wye secondary. Your
standard three-phase transformer has a delta primary and a wye secondary;
the center tap of the wye is grounded at the transformer and becomes
the common point for the ground and neutral conductors.
X ----------------| .--------------- A
o 8
8 8 . . . . . x=============== G and N
Y ------------8oooooo8 o o
| . .------------ B
Z -------------------- |
.------------------- C
(The center tap of the secondary is grounded but it's a little hard to show
a ground symbol with text characters! Two insulated lines, G (safety ground),
and N (current-carrying neutral) come from the center.)
Typically the primary (XYZ) is three-phase, 480 VAC, meaning that you have
480 VAC between any two of the three lines. The secondary generates three-
phase, 208 VAC, but because it's in a wye configuration you can get 115 VAC
between N and any of the three phase legs A, B, or C. If you've got a
power line delivering 208VAC, it's a good bet that it's coming from a
three-phase transformer.
230 VAC, on the other hand, is almost exclusively an artifact of a single-
phase line. That's why you have home appliances (air conditioners,
hot water heaters, etc) rated for use with 230 VAC: the line from the power
pole outside your house is 230 VAC, center tapped to provide 115 VAC for
"normal" use. It's still a single-phase supply, however. (Meaning that
you can't buy that surplus 3090 and run it in your basement!)
The point I'm making is that the 208 VAC doesn't really have to come from
a wye secondary, but it usually does. And computer equipment destined for
centralized facilities in modern buildings is usually designed to be fed
from a 208 VAC source (single or three phase), or 115 VAC which can be
derived from a 208 VAC wye secondary. There are installations with a
delta secondary, but they're relatively rare.
Of course, you've got the vendors who don't seem to believe that there
is anyone using their equipment in an environment which doesn't have
230 VAC. I posted a comment earlier in this thread about Sun's 3/280
and the fun we had installing it in a 208VAC environment...and the
same fun the local Sun office had with their own machine.
Quiz next Monday.
Joe Morris
Rich Alderson
>The massbus was originally developed for the pdp-10, I think, and
>the pdp-10 is 36 bits, so massbus can/is 36 bits too.
>I'm not sure how the disks are connected to a DECsystem-10 or
>a DECSYSTEM-20, but I think they don't pass through the front -11,
>since the -11 has its own connection to the boot disk, and the highspeed
>disks cannot be handled by an 11/40.
The MASSBUS is/was an 18-bit datapath. It directly connected peripherals to
RH20s in the I/O bay. It was also used with the -11 to hook up the boot device
(an RP06 on most systems by the time DEC killed the 36-bit systems), via an
RH11.
(Stanford even used the MASSBUS for hooking 20's to the Ethernet. The Massbus-
Ethernet Interface Subsystem, or MEIS, is still, to my mind, a better design
than the NI20.)
Everyone agrees that the plural of MASSBUS is MASSBI, right?
--
Rich Alderson 'I wish life was not so short,' he thought. 'Languages take
Tops-20 Mgr. such a time, and so do all the things one wants to know about.'
AIR, Stanford --J. R. R. Tolkien,
alde...@alderson.stanford.edu _The Lost Road_
Paul Hardy
> The fastest directory program I ever saw
> ran on RSTS (it's called "DIR", as opposed to PIP which is/was what the
> DCL DIR command ran).
When I used RSTS/E (V06*, mid-70's) the directory command was "catalogue"
or "catalog" or simply "cat"; however, there was also a system call
SYS(magic,arguments) to obtain a directory listing from within a BASIC-PLUS
program. When did "cat" become "dir"?
--Paul
--
This is my address: p...@ama.caltech.edu
This is UUCP: ...!{decwrl,uunet}!
This is my address on UUCP: ...!{decwrl,uunet}!caltech.edu!ama!ph
Any questions?
"Does Emacs have the Buddha nature?" --Paul Hardy "Yow!" --Zippy
Terry Kennedy, Operations Mgr.
> When I used RSTS/E (V06*, mid-70's) the directory command was "catalogue"
> or "catalog" or simply "cat"; however, there was also a system call
> SYS(magic,arguments) to obtain a directory listing from within a BASIC-PLUS
> program. When did "cat" become "dir"?
CAT was the one built into Basic-Plus, which was the normal environment
back then. The other ones are DIRECT (the one you responded to), which is a
Basic-Plus program that reads the disk directly, and PIP/DI, which is what
DIR in DCL calls these days.
Don Stokes
> When I used RSTS/E (V06*, mid-70's) the directory command was "catalogue"
> or "catalog" or simply "cat"; however, there was also a system call
> SYS(magic,arguments) to obtain a directory listing from within a BASIC-PLUS
> program. When did "cat" become "dir"?
CAT exists within BASIC-PLUS; PIP, DIR and any number of others are
external programs. PIP of course does everything in the way of file
management, not just a directory.
RSTS/E has grown somewhat in the last couple of decades or so -- it's no
longer just an overgrown multi-user BASIC. The syscall to get directory
info still exists, however using it isn't the ultimate in speed (there
are tricks to using it effectively). DIR operates by saying "to hell
with syscalls"......
Don Stokes
> CAT was the one built into Basic-Plus, which was the normal environment
> back then. The other ones are DIRECT (the one you responded to), which is a
> Basic-Plus program that reads the disk directly, and PIP/DI, which is what
> DIR in DCL calls these days.
We had something else called DIRECT that did ordinary directory lookups
and optionally searched a file of descriptions. It was v e r y s l o w.
Maybe we're talking about different programs, maybe ours was an older
version (everything was old on these systems -- we were still running
RSTS/E V8 in 1988 on all but two of our 11/70s). We had:
DIRECT (slow) invoked by DIR
PIP (adequate) DIR from DCL
PIP/L or PIP/D from anywhere else
DIR (*fast*) FILES
Mark Crispin
>m...@milton.u.washington.edu (Mark Crispin) writes:
>>There's the alternate problem. DEC assumes that everyone has
>>three-phase power in a wye (not delta) configuation. Consequently,
>>everything that needs more than 110V is 208V or three-phase.
>Uh?
You misinterpreted. I was complaining that DEC assumes that everyone
has 208 (= three phase), not 230 (= single phase).
It was just a side note that it must be a wye secondary, not a delta.
DEC stuff assumes that it can get 115V between any leg and neutral,
which is not true in a delta configuration because the wildcat leg has
a 190V voltage difference.
I also believe that a delta secondary has 240V between each leg?
sys...@iowasp.physics.uiowa.edu
when you attempt to spin up that drive that has the big motor
in it???
Willy
Works wonders on the chain drive motor on the printer...
Bill Pechter
>
>When I used RSTS/E (V06*, mid-70's) the directory command was "catalogue"
>or "catalog" or simply "cat"; however, there was also a system call
>SYS(magic,arguments) to obtain a directory listing from within a BASIC-PLUS
>program. When did "cat" become "dir"?
>
I'm not too deep into RSTS, though. I think DCL runs PIP when directories
are requested, just like RT11...
CAT, I thought was in the Basic Pluis RTS (Run Time System).
My experience with RSTS was with V7.0, 7.2 and 8.0. I did see 9.0 and the
DCL was really very Vax/VMS-like.
/* Disclaimer - I know how to login to RSTS, I can run ERRDIS, but I'm
not a RSTS expert by any means. */
Bill
Joe Morris
>So... What do you say when the electrician swaps phases on you
>when you attempt to spin up that drive that has the big motor
>in it???
I'm not sure what I would call it on a disk drive, but on a (mainframe)
1/2" tape drive we called it "pressure column mode".
Actually, I think most of the boxes which could really be damaged by the
reversal of phase (note I didn't say anything about whether they would
*work*!) had phase sensors in them unless they were the real el cheapo
stuff. There were exceptions, of course; in our case we've had cases
(note the plural) where the low-bid electrical contractor hooked up the
power to our Exide 2045 UPS bass-ackwards and blew out the rectifier
(once) and the inverters (twice). (I say this as an exception to the
comment about "el cheapo". The Exide equipment is generally quite good...
as compared to some other stuff, like a certain well-known T-1 mux vendor
who shipped a non-UL listed power strip with instructions to supply its
5-15 outlets with a 30A breaker...)
Joe
Walter C. Daugherity
3000, I think, but anyway, I seriously considered buying a cheap used RM02
and changing the pulley or the motor to get a 3600 RPM RM03, but I couldn't
find both schematics to make sure there weren't some electronics changes too.
A little more work than the famous "remove that slow-down jumper" approach
to IBM 407's etc.
Bill Pechter
>3000, I think, but anyway, I seriously considered buying a cheap used RM02
>and changing the pulley or the motor to get a 3600 RPM RM03, but I couldn't
>find both schematics to make sure there weren't some electronics changes too.
>
>A little more work than the famous "remove that slow-down jumper" approach
>to IBM 407's etc.
>
Just live with an occasional data late error.
Bill
Robert Bernecky
>
>>In article <16...@helios.TAMU.EDU>, dau...@cs.tamu.edu (Walter C. Daugherity) writes:
>>> at the computer and feed it with a single-phase hot wire!
>
>> Of course, the neutral was designed to carry only the current resulting from
>>any phase imbalances, not the full load. Thus, doing this may result in exces-
>>sive current in the neutral lead. I'd suggest going to at least #10, probably
>>#8, before doing this.
>
>For years the NEC permitted buildings to be wired using a neutral conductor
>size of 50% of the size of the phase conductors on the assumption that the
>phase loads would average out and the neutral wouldn't need to handle much
>current.
>
>The code was updated recently to require that the neutral be sized at 200% of
>phase conductor size. The change is required because of all of the very
>dirty (electrically) loads which have become common in the past decades.
>There have been documented situations where the three phase conductors are
>carrying equal current loads ... and the neutral is carrying a higher
>current. With sufficient harmonic content, you can have the three phase
>legs carrying 100A apiece and still have > 100A in the neutral.
I recently ran into a similar problem with theater lighting:
In residential kitchens and other places where you might be pulling a
good hunk of power from lines, you install "split duplex" receptacles.
These are wired across two sides of a 120-neutral-120 circuit, on the
assumption that the neutral is only carrying the difference between the
two hot lines, and hence can be the same size as the 120 (hot) lines.
However, in a theater hookup, you tend to have junction boxes at BOTH
ends, of such circuits, as one end is going into lighting instruments,
and the other is going into one dimmer or another.
The end result is that you cannot count on both of the hot lines being
on the same circuits (or maybe even the same transformer!), so
you must run separate lines for each circuit.
Bob
Robert Bernecky r...@yrloc.ipsa.reuter.com bern...@itrchq.itrc.on.ca
Snake Island Research Inc (416) 368-6944 FAX: (416) 360-4694
18 Fifth Street, Ward's Island
Toronto, Ontario M5J 2B9
Canada
Richard Wexelblat
>So... What do you say when the electrician swaps phases on you
>when you attempt to spin up that drive that has the big motor
>in it???
Just ask the IBM installers. (Ancient History time!)
When the first 360/65s were installed at Bell Labs in Holmdel (one had
serial number under 10, one under 20) the installers got the phases
backwards on the forced air cooling fans. It took months to figure out
why they were always overheating!
--
--Dick Wexelblat (r...@ida.org) 703 845 6601
Can you accept an out of state sanity check?