Building a PDP-11 for the first time

From: Jerome H. Fine <jhfinexgs2_at_compsys.to>
Date: Mon Dec 15 21:42:17 2003

>Pete Turnbull wrote:

> > On Dec 14, 21:35, Jerome H. Fine wrote:
> > And DEC did as well, as far as I know since the upgrade
> > looks like an official DEC job.
> Very possibly, though it's often impossible to tell who actually fitted
> the replacement EPROMs -- DEC sold them to people on self-maintenance
> and to third-party maintenance companies like the one I worked for.

Jerome Fine replies:

YES!! You are correct.

> > And that is quite a separate issue from whether or not
> > PMI memory was or was not used with the M8190-AB
> > or the M8190-AE boards. RT-11 would report both
> > of these boards as a PDP-11/73B Processor when the PMI
> > memory (or regular memory) was installed below the CPU.
> > If the PMI memory was installed above the CPU, RT-11
> > reported both boards as a PDP-11/83 Processor.
> Yes, because that's about the only useful distinction -- the speed
> increase brought by PMI memory is more than the clock difference
> between 15MHz and 18MHz -- and of course if placed after the CPU, even
> a PMI-capable memory board operates as normal Q-Bus memory.

Many years ago, I finally acquired a M8190-AE and some
PMI memory. I don't know how many are interested, but
I found the following:

Approximate speed increase from M8190-AB (KDJ11-BB)
with normal DEC memory (obviously below the CPU) vs
M8190-AE (KDJ11-BF) with DEC PMI memory above
the CPU was about 33%. An assembly which took about
4.5 minutes was reduced to about 3 minutes. I can't recall
the exact numbers or exactly what I tried, but the test with
the M8190-AB and the PMI memory above the CPU took
about 23% less time than 4.5 minutes. I then tried the
M8190-AE and normal DEC memory below the CPU
and that took about 10% less time than 4.5 minutes. When
the PMI memory was below the M8190-AE, that also
took about 10% less than the 4.5 minutes.

Based on my observations, I can suggest that the use of
PMI memory is the major reason for the speed improvement,
NOT the change of the crystal from 15 MHz to 18 MHz when
the switch was made from the M8190-AB to the M8190-AE.

I am not sure if this conclusion agrees or disagrees with
what you stated above. However, it is what I found
during my rather limited test.

I have heard from others (I think that Megan Gentry once
stated that she did this as well, but it might have been with
the M8190-AE board) who also were able to effect a
substantial speed improvement just using PMI memory
above the CPU board.

Of course, when DEC first sold PMI memory, it was
EXTREMELY expensive compared to normal memory
which was VERY expensive compared to memory for
a PC. Later, I was able to find PMI memory at a price
below that for normal memory since most people at
the time thought that PMI memory could ONLY be
used with an M8190-AE board. Now all DEC memory
seems to be expensive again and PMI memory is also
expensive.

> > Some people like to hear the original fans turning and
> > are not comfortable with a system that runs 100 times
> > as fast as the original hardware.
> Ah, I like to poke the switches, see the lights and hear the fans :-)

The first time I do also. But I still get a bigger thrill
from finding a bug in RT-11 and then fixing it.

Enhancing one of the utilities or adding Y10K to RT-11
is also the same sort of challenge from my point of view
that you probably receive from fixing an M8190-AE
board.

I realize that the majority on this classiccmp list enjoy
the hardware more than the software, but I hope I
am not completely alone!

And then there is the program to locate the PRIME
numbers from the range of numbers from 1 to
4,000,000,000. I have not worked on that one for a
while. I want to see how Ersatz-11 can help by
improving access to memory for FORTRAN
VIRTUAL statements. I need about 500 Kbytes
and the MMU makes things very slow.

With direct access to PC memory and a system
100 times as fast as a PDP-11/93, the system
should be a lot faster. I may be able to locate all
the PRIME numbers from the range of numbers
from 1 to 256,000,000,000,000.

Sincerely yours,

Jerome Fine
--
If you attempted to send a reply and the original e-mail
address has been discontinued due a high volume of junk
e-mail, then the semi-permanent e-mail address can be
obtained by replacing the four characters preceding the
'at' with the four digits of the current year.
Received on Mon Dec 15 2003 - 21:42:17 GMT

This archive was generated by hypermail 2.3.0 : Fri Oct 10 2014 - 23:35:50 BST