• Please review our updated Terms and Rules here

Pdp11/23 max memory confusion...

R

Roe

Experienced Member
Joined
Dec 28, 2012
Messages
240
Location
Regina, Canada
I have an M8186 KDF11 cpu card, etch C (I think) and there is some confusion about max memory.

The 1985 DEC Microcomputer Products Handbook states it is "capable of addressing up to 4 megabytes of main memory." (Pg 3-1)

On the other hand, the preliminary KDF11-A users guide on bitsavers states it maxes out at 256k, with 11/34 style memory management.

I've found references to a KDF11-BA quad width CPU with full 22-bit addressing, but that's a whole different beast.

Anyone have any information? I hesitate to plug this into a Q22 bus with 4MB of memory just to test it.

Thanks!
 
Look to see if your board has the MMU or you not going to get much beyond 28K, also have to wonder just how much memory you need? If you’re just running RT-11 and Basic you will find life’s good with as little as 32K Not one hundred percent on this but think you need the 22 bit wide system to address 4 meg and on the 18 bit system that tops out at 256K
 
Sorry, i should have mentioned in the original post, its an -AA version with the MMU, no FPU chip.

The reason for the concern is that all I have for qbus memory right now is a 1MB board. Definitely overkill for this box, but until I can find an M8047, it is available...

I also plan to stick in a cqd200 to install RT11, and I dont want to hack the board trace for 18 bit mode if I don't have to.
 
When discussing processors it's important to keep track of "words" and "bytes". The PDP-11 can access memory in 8 bit bytes but fetches a 16 bit instruction word and natively used a 16 bit memory address. This used to seem strange to someone who started out on a UNIVAC machine that had a 36 bit word that could be addressed in 12 bit bytes and talked in something called FIELDATA (pre ASCII days). With a native memory address of 16 bits the PDP-11 (without help from a MMU) can address 2^16 or 64K bytes that are also sometimes referred to as 32K words. It's important to specify the all important "bytes" or "words" with the quantity. As the top 8k bytes or 4k words are reserved for I/O use, we get the upper limit for memory of a basic PDP11/03 of 56K bytes or 28K words. And this is before you throw in the processor type and quantity of backplane address lines as variables. A PDP-11/23 can work with more than a PDP-11/03 and a PDP-11/23+ yet even more. And a maxed out PDP-11/70 is a different animal altogether. But everything you need to know to figure out what your particular machine can do can be found on this site, or on the links the people here can and do provide. Have fun with your PDP.
 
Frustrated: thanks for the reply. My only comment on your post is: yep, I know all of that. I started with pdp11 systems in 1973.
 
Thanks for asking, Marty.

I spent all night one night trying to reset the password on my original account. The procedure on this site is as useless as.... well..... use your imagination. I became very..... Frustrated!

But you guys have been no end of help in tips to others that have allowed me to make a lot of progress on my 11/03 (Heathkit H11), my 11/23 (formerly a Bell System Control Console 2A, and the 11/23+ I bought. Thanks to each and every one of you.
 
Roe said:
Frustrated: thanks for the reply. My only comment on your post is: yep, I know all of that. I started with pdp11 systems in 1973.
Click to expand...

Yep, figured you did. ;-)

But please keep in mind that as this hobby (hopefully) expands, new people will be reading these posts as one of their means of learning & getting that old system they found in the dumpster working. They're going to be looking for tips on the posts you experts and near experts provide almost by accident in your conversations with each other. So it helps if you keep it clear and complete whenever possible. Some kid might be in his mom's basement trying to figure out why some PDP-11's max at 28k, and some others at 56k without realizing that they're one and the same.
 
frustrated said:
When discussing processors it's important to keep track of "words" and "bytes". The PDP-11 can access memory in 8 bit bytes but fetches a 16 bit instruction word and natively used a 16 bit memory address. This used to seem strange to someone who started out on a UNIVAC machine that had a 36 bit word that could be addressed in 12 bit bytes and talked in something called FIELDATA (pre ASCII days). With a native memory address of 16 bits the PDP-11 (without help from a MMU) can address 2^16 or 64K bytes that are also sometimes referred to as 32K words. It's important to specify the all important "bytes" or "words" with the quantity. As the top 8k bytes or 4k words are reserved for I/O use, we get the upper limit for memory of a basic PDP11/03 of 56K bytes or 28K words. And this is before you throw in the processor type and quantity of backplane address lines as variables. A PDP-11/23 can work with more than a PDP-11/03 and a PDP-11/23+ yet even more. And a maxed out PDP-11/70 is a different animal altogether. But everything you need to know to figure out what your particular machine can do can be found on this site, or on the links the people here can and do provide. Have fun with your PDP.
Click to expand...

Now there's some words I have heard in awhile. I worked at Sperry Univac in my early days. Which machine did you use?
 
Frustrated (or whoever else may be bored),

When you mentioned that a maxxed out 11/70's memory configuration was even more different, I was intersted to hear more. Although I'll never use an 11/70, I'm curious about the memory structure. Could more than 2MW of main memory be addressed? Is there someplace I should read about this myself?

Lou
 
Lou - N2MIY said:
Frustrated (or whoever else may be bored),

When you mentioned that a maxxed out 11/70's memory configuration was even more different, I was intersted to hear more. Although I'll never use an 11/70, I'm curious about the memory structure. Could more than 2MW of main memory be addressed? Is there someplace I should read about this myself?

Lou
Click to expand...

I'd start with the 11/70 handbook, which is up now on bitsavers.

The big difference between Unibus and QBus systems is now it dealt with memory over 18 bits. The Unibus was an 18 bit bus, but memory was expanded in the 11/70 (and later in the 11/44)
to 22 bits, which required Unibus mapping windows to support 18 bit peripheral controllers to DMA to all 22 bits of memory. The QBus was extended from 18 to 22 bits but the peripheral
controllers were changed to directly address all of memory. Starting with the 11/24, DEC ended up building Unibus map boards that let Unibus controllers work with QBus processors and memory.
 
Al Kossow said:
I'd start with the 11/70 handbook, which is up now on bitsavers.

The big difference between Unibus and QBus systems is now it dealt with memory over 18 bits. The Unibus was an 18 bit bus, but memory was expanded in the 11/70 (and later in the 11/44)
to 22 bits, which required Unibus mapping windows to support 18 bit peripheral controllers to DMA to all 22 bits of memory. The QBus was extended from 18 to 22 bits but the peripheral
controllers were changed to directly address all of memory. Starting with the 11/24, DEC ended up building Unibus map boards that let Unibus controllers work with QBus processors and memory.
Click to expand...

The driving factor I see for the 22b physical address as an update to the 18b physical address architecture (UNIBUS) was that there were only 4 free unused bits in the MMU PAR (page address register) field. In 18b systems the PAR implemented a 12b block number, and the upper 4b of the PAR was not used / reserved. For the 11/70 they extended the block number in the PAR to its maximum limit, 16b, and this then resulted in the 22b physical memory address lmiit (ie, 12b+6b=18b to 16b+6b=22b). This was the best DEC could do without a significant architectural change to the PDP-11 MMU structure.

As Al indicates the 18b UNIBUS could not change, so a mechanism was added for DMA peripherals to map 18b physical UNIBUS addresses to 22b system memory addresses. The UNIBUS map was added to do this. It dedicated a pair of registers (32b) to the base address of each mapped block, but only used the low 22b as the base address. The upper 10b was unused, as they were not necessary. If the full 32b UNIBUS map base address were used it could theoretically support a 32b physical address space (4GB).

So the 22b limit was due to the number of unused bits in the PDP-11 MMU PAR registers. Besides who would ever need more than 4MB of memory on a computer, anyway? :)

Don
 
Lou - N2MIY said:
Frustrated (or whoever else may be bored),

When you mentioned that a maxxed out 11/70's memory configuration was even more different, I was intersted to hear more. Although I'll never use an 11/70, I'm curious about the memory structure. Could more than 2MW of main memory be addressed? Is there someplace I should read about this myself?

Lou
Click to expand...

Perhaps he was alluding to the fact that the 11/45/50/55/70 and J-11 supported separate Instruction and Data space? Thus a program has effective access to 128K bytes rather than the usual 64K bytes? Discussed in Chapter 9 of the PDP-11 Architecture Handbook (pp. 229-230).
 
tradde said:
Now there's some words I have heard in awhile. I worked at Sperry Univac in my early days. Which machine did you use?
Click to expand...

My brief experience with a UNIVAC machine was limited to my time at Florida International University's Computer Science Department. If I remember correctly, and my rememberer is not as good as I remember it being, the machine was an 1100/80. Another quirky thing I remember about it was that it would periodically lock disk access while it "squeezed" the disk. I was told that was a byproduct of the OS requiring that all blocks in a file be contiguous. About a year after I started, the student computer use was switched from the UNIVAC to a VAX cluster environment running VMS. Again this is all from 30 year old memories so some of it may be suspect.

FYI.... some of my posts, including my responses to a couple of questions, have seemed to have failed some sort of screening process and not made it onto the forum. I'm not ignoring you guys. The answers just are not meeting some mysterious required criteria.
 
You must log in or register to reply here.
Back
Top