• Please review our updated Terms and Rules here

Creating a PDP-11. What are my Options?

C

Clyde1992

New Member
Joined
Aug 13, 2013
Messages
5
Hello, everyone:
As I have nearly finished my most recent retrocomputing project (connecting an Apple IIGS to the internet), I have decided to start a new one. While I mostly collect vintage microcomputers, I have decided to try something new and purchase a minicomputer. I have always wanted to own one, and, after much thought, have decided that a PDP-11 would be the best mini for me to own.
I understand that it is fairly difficult to purchase a full PDP-11 system for a decent price, so I am considering building one from parts. However, as my experience with minicomputers is very superficial, I have no idea where to start.
I understand that the PDP-11 range of computers is fairly broad. What I am looking to build would be a PDP-11 that is compatible with the vast majority of PDP-11 software while still being fairly easy to assemble and (preferably) relatively inexpensive. What are my options? From my understanding, all I would really need for a basic system would be a backplane, a cpu card, a serial card, and a power supply. I also do know where I could purchase a J-11 PDP-on-a-Chip. Would it be easy to design a computer around this chip? There is also the FPGA route, but that is my last resort. I would really prefer to own the actual system. Finally, if someone knows of an inexpensive replica of a PDP-11, I would be interested to hear about it.
I would be grateful for any suggestions about where to begin. Thanks, in advance!
 
I think you're going to get a lot of elaborate answers to that question. I'd start with a late 70's early 80's system if you can find a chassis and power supply. Shouldn't be too hard to cobble something like that together. Not sure if I'd want a Rainbow, but that would be another route to take. Good luck. If you have a page/link about the Apple IIGS project I'd be interested to read it.
 
Goodness... how do I add to this response.

Original Hardware -

PDP-11 systems come in primarily two overlapping "generations" defined by BUS type. These are UNIBUS and QBUS. They range in size from somewhat larger and deeper than a "Tower Server" of today, to something the size of a not too small car (tipped on it's side) and weighing a little more.

There are CPUs throughout both UNIBUS and QBUS systems that are fundamentally software compatible with each other. However, if you want to run "any software" there is no option but to choose an 11/70 architecture machine. This would be either a real 11/70 (some would argue an 11/44 fits here too) and any system based on the J11 Chip. (11/73, 11/83, 11/53, 11/93, 11/84)

There are a couple hermaphrodite systems, mixing BUS types. These would be the PDP-11/70 and the PDP-11/84.

The 70 (and 74 if one exists) were "halfway to VAX" - Grafting a 32 bit a memory system and MASSBUS I/O to the CPU(s) through a very efficient CACHE. Also in the picture was a UNIBUS for I/O option compatibility with other family members.

The 84 is actually a QBUS machine, with special memory (PMI) and a UNIBUS adapter. The idea here was to replace those UNIBUS machines that were ageing out with something in the 70 performance class. The comparison worked for many systems, but if your 70 employed primarily MASSBUS disks, you would be very disappointed by the transition.

While a PRO350 or PRO380 "PC" is software compatible with much - it's not entirely so. It's hardware interfaces are severely restricted. So, I do not include any description of these as they are not really applicable to a model of "general compatibility" you requested. (?)


Inexpensive Replica -

What I am looking to build would be a PDP-11 that is compatible with the vast majority of PDP-11 software while still being fairly easy to assemble and (preferably) relatively inexpensive.
Click to expand...

Within this confine, there is only one alternative. Simulation on a PC using ERSATZ-11 by John Wilson. E11 is free to anyone for non-commercial use. If desired, you could purchase E11 and one of the HARDWARE ADAPTERS that connects it to QBUS or UNIBUS to gain access to the physical media which works on those two BUS types. This cost would still be less than an actual hardware system today.

However, this is entirely unnecessary as E11 supports the most amazing virtual emulation of PDP-11 hardware ever. This is so faithful that original DEC Diagnostics run on these emulated devices - without error. It correctly emulates ANY PDP-11 CPU, up to and including the 11/74.

Perhaps the best reason to consider this approach, is that it's totally free to try and you can use it to go directly to your desire of running software. Any software.

Should you decide later to indulge yourself in hardware (be it original or self-made) you'll have the experience to know what the software behavior should be and will have made choices based on intimate knowledge of the same.

The software of a PDP-11 is much more complex and amazing than any of the hardware it runs on. Few have more than a cursory understanding of this today. (unless you happen to be Dave Cutler) E11 will allow you to obtain, exercise and even develop such software on the emulated CPU and system of your choice with no cost to you other than electricity to run a PC and download software from the internet.

New Hardware -

I know of only a couple commercial hardware offerings, and all are far more expensive than actual original systems. Making your own "J11" (JAWS) board, while fun - is not even half the battle. Doing it to be compatible with real BUS I/O is an enormous undertaking and could not possibly be done cheaply. Debugging it and verifying it, certainly couldn't be.

I know there is a working FPGA version or two out there, but once you want to take the next step of interfacing to actual hardware, this must suffer the same difficulty and inherent costs as basing on J11 chips. The only potential advantage is higher performance. (which is in itself "unrealistic" emulation)
 
I am more a hardware person but know from experience that KDF11 (M8186) cards are out there and available fairly cheap along with most of the parts for the 11/03 and 11/23 systems and would think that for under a couple hundred dollars you can get a CPU, memory, power supply and communications card and this will allow you to run ODT or load from the serial port but where things get expensive is getting drives and media for the drives. All depends on what you want to do, software or hardware.
 
There was a period when people were pitching PDP-11’s and would virtually pay you to haul them away. Then the tide turned, although it appears that PDP-8’s are currently demanding higher dollars than most PDP-11’s. If your primary objective is to play with PDP-11 software, then E11, as RSX11M+ mentioned, is indeed an excellent start. If you’re mainly interested in the hardware aspect, then a Qbus system is probably the more readily available, cheaper and smaller alternative to a UNIBUS one. I would think it to be a rather daunting effort to start with the J-11 and then attempt to build around it unless you’re really into that type of challenge.
 
I'd like to respond, not to be argumentative, but to illustrate the consequences of, and rationale behind, some of these choices.

Qbus said:
I am more a hardware person but know from experience that KDF11 (M8186) cards are out there and available fairly cheap along with most of the parts for the 11/03 and 11/23 systems and would think that for under a couple hundred dollars you can get a CPU, memory, power supply and communications card and this will allow you to run ODT or load from the serial port but where things get expensive is getting drives and media for the drives. All depends on what you want to do, software or hardware.
Click to expand...
I too am a hardware centric guy. My colleagues in R&D were always accusing me of this - and they were correct, of course. However, my desires today are for hardware I simply cannot obtain or have. I've used E11 when my adventures took me in those PDP-11 directions.

I absolutely agree there is no doubt a QBUS system is the cheapest real iron. (except for extraordinary luck) As to prices - what's affordable varies on your area, circumstances and opportunities. I've seen a huge increase in QBUS / LSI11 hardware prices recently. He will undoubtedly investigate for himself and decide, but I personally would not call them "inexpensive" any longer.

In order to satisfy his requirement of "compatible with the vast majority of PDP-11 software", a KDJ11 CPU is required if the system is QBUS. A KDF11 would leave him out in the cold on a lot of software - anything that uses separate I&D space for example. However, I do agree that a KDF11 will permit him to run a particular version of most DEC software offerings, so on that basis I guess it could be included. A KDJ11 is pretty much UNIVERSALLY able to run all software that isn't "UNIBUS CPU" dependent - which isn't much. In contrast an 03, for example would severely restrict what he could run as would a KDF11 not equipped with an MMU or FP (both included in the J11 chip).

Probably a moot point since the 03 is more collectible these days than the 11/23. BTW - any 11/23+ (QUAD) is the module to have in that class. It's two serial ports make it a very good choice even at twice the price of the DUAL module. Be sure to get the cabinet kit.


KM11 said:
There was a period when people were pitching PDP-11’s and would virtually pay you to haul them away. Then the tide turned, although it appears that PDP-8’s are currently demanding higher dollars than most PDP-11’s. If your primary objective is to play with PDP-11 software, then E11, as RSX11M+ mentioned, is indeed an excellent start. If you’re mainly interested in the hardware aspect, then a Qbus system is probably the more readily available, cheaper and smaller alternative to a UNIBUS one. I would think it to be a rather daunting effort to start with the J-11 and then attempt to build around it unless you’re really into that type of challenge.
Click to expand...

Your reference to "pitching" is true, and still gives me nightmares. The mind reels at my missed opportunities. :wallbang:

Also to be considered in a UNIBUS choice are the cost of running it (electricity) as well as spares & replacements. A QBUS system is also less "hazardous" to a novice, although expensive hardware destructive mistakes are still possible, it's nothing like the spectacle that can befall one in a UNIBUS machine.

I have always wondered what the all-time production numbers were for PDP-11 vs PDP-8 systems. I have a vague recollection that 11's passed the 5 million mark (not counting Chip products) before being transferred to Mentec, (could that be right?) but do not recall any number for 8's. Does anyone know?

Rarity's not got everything to do with today's market pricing, but it does form a basis of appreciation.

As to self development - if it's your bag of tea, great! But it will not be cheaper than a ready built system, no matter what. Assembling a system from module level parts is possible, but still likely to be more pricey and fraught with danger for the novice. One would require very high aptitude indeed for such an undertaking, and must not be too bothered by the prospect of damaging things in the attempt.

I discounted it for him, on the basis that uncertainty brings cost.


Still a fun discussion though.
 
Last edited:
Quote from previous msg :

Probably a moot point since the 03 is more collectible these days than the 11/23. BTW - any 11/23+ (QUAD) is the module to have in that class. It's two serial ports make it a very good choice even at twice the price of the DUAL module. Be sure to get the cabinet kit.o

Be sure to get the cabinet kit. ???

Please can you provide me with details ?? Many Thanks.
 
This provides the mountable bulkhead hardware for external cable connectors and the internal ribbon cable to connect to the CPU module itself.

Which you need depends on the system box, but the one for the BA23 is DEC P/N CK-KDF1BKA.

Please see EK-245AA- MG-001 Pg 17, Figure 1-2 for an image and further details.

There is a collateral component for the H349 panel of the PDP-11/23+ in the older style BA11-S box referred to as the "KDF11-B Cable Assembly". See it in EK-1T23B-OP-001

The definitive resource is the 83 Systems & Options pg 97 (Components 3-5)

  • CK-KDF1B-KA - Kit for BA23 (includes baud rate selector)
  • CK-KDF1B-KB - Kit for BA11-M (includes baud rate selector)
  • CK-KDF1B-KC - Kit for H349 panel (includes baud rate selector)
  • CK-KDF2B-KB - Kit for BA11-M (Does NOT include baud rate selector)
 
Last edited:
Clyde1992 said:
I also do know where I could purchase a J-11 PDP-on-a-Chip. Would it be easy to design a computer around this chip? There is also the FPGA route, but that is my last resort. I would really prefer to own the actual system. Finally, if someone knows of an inexpensive replica of a PDP-11, I would be interested to hear about it.
Click to expand...

Hi All, I'm not sure if my comment 100% relevant to this thread, but anyway you probably know about famous article "pdp-11 behind the iron curtain". Most of those machines/cpus were too yellow, so gold scrapers already killed them all. However, there were few chips developed by russian ic manufacturer Angstrem back in 80ties (namely k1801vm1,vm2,vm3), those chips are available on ebay and other auctions (however, shipping cost could be daunting). Basically they made them using technology similar to Ferranti's ULA, there is hardware emulator of real DEC cpu inside. However, ODT is not built in feature for mentioned chips, so you need special ROM chips ouside. Also, there are ready to run computers/modules based on those chips are available. vm1 and vm2 is kinda 11/03 systems capable of running rt-11 and probably unix v6. The bk-0010,bk-0010-01,bk-0011 are home computers with 4 color displays and paged memory. The uk-nc educational computer is the system where two k1801vm2 cpus are installed (main cpu and peripheral cpu) also includes graphical chipset with 8 or more(?) colors. There are also DVK computers with the bus similar to q-bus, they use cpu cards ms1201.x where x=1,2,3,4; 1- for 1801vm1 cpu, 2- for vm2; 3,4 - for vm3 (this chip includes MMU but no separate I/D, so it is capable of running rsx-11m).
 
Last edited:
Lou - N2MIY said:
Bigral, you seem to know a good bit about those clone machines. Can you tell us, were there any clone unibus 11s? I've only seen or heard about LSI-11 based clones.

Lou
Click to expand...

The unibus machines are SM-3, SM-4, SM-1300, SM-1420, elektronika-79 (11/70 clone), elektronika-100/25(11/40 clone) (all these made of SN discrete logic), sm-1425 (small factor connector\enclosure was used, looks similar to 11/73 uses clone of j-11). I've seen some of those at different places around 30 years ago, but I was to young to understand WHAT THAT WAS, so most of this information extracted out of wikipedia and web, I never worked with those machines.
 
bigral said:
... There are also DVK computers with the bus similar to q-bus, they use cpu cards ms1201.x where x=1,2,3,4; 1- for 1801vm1 cpu, 2- for vm2; 3,4 - for vm3 (this chip includes MMU but no separate I/D, so it is capable of running rsx-11m).
Click to expand...

Not nit-picking, but clearing up a common misunderstanding for posterity...

The I&D Space feature is invisible under RSX11M. The system is not designed to make use of it, or offer it to applications. Only RSX11M-PLUS with an appropriate CPU (one which includes I&D space support) has the ability to support applications compiled to use it.

The RSX11M-PLUS operating system will itself use I&D space features if sysgen'd to do so. For an OS, I&D space has tremendous advantages. In applications, especially those intended to be simultaneously invoked by multiple users (like a compiler or linker) the system keeps only one copy of the instructions in memory, and shares it with all instances of the app. This not only saves memory, but tends to keep that app's instructions in CPU cache, increasing it's performance. Each instance of the app, can use a full 32kw Data space for it's runtime data. Once code or data is larger than that limit, RSX11M-PLUS supports overlay'd applications by dynamically remapping memory regions, rather than disk based overlay techniques in previous systems - up to the limit of physical memory. Disk based overlay and task checkpointing methods are only invoked when physical memory is exceeded. Unlike the Windows OS, RSX11M can "shuffle" virtual task elements to re-compact memory use, without first committing them to a swapfile on disk.

Finally, RSX11M-PLUS can also be used on CPUs without I&D space. However, the system cannot be sysgen'd to expect I&D space if it is intended to run on such a target. Separate I&D space is not the only reason one might want to run RSX11M-PLUS. It offers a more flexible memory pool mechanism, is more run-time configurable, and implements such features as overlapped seeks on disk controllers with multiple drives, elevator optimization, and disk caching.
 
KM11 said:
...the 22-bit addressing requirement greatly limits the processor options here.
Click to expand...


I did some research before posting to this point. In doing so, I have consulted hardcopies of SPDs for RSX11M-PLUS 2.1, 3.0 and the online copies of 4.6 and the original marketing Guide for M+.

The short response is that RSX11M-PLUS version 2.1 thru 4.6 were applicable to 11/23 CPUs or higher (that is: 23, 23+, 24, 44, 70, 73 and later 70mp, 74, 83, 84, 93). Supported minimum memory configurations began at 128KW and increased over time to a 256KW minimum.


Detail:

My first experience with RSX was RSX11M v3.1. This was considered a "Gold Standard" at the time, and had been since it's release. It was applicable to systems without an MMU (UNMAPPED) with as little as 16KW of memory, and those with an MMU (MAPPED) up to 1920KW. This included all UNIBUS processors at that time, including the 11/70.

However, with the introduction of the 11/70, DEC realized that a single RSX OS would be held back if it were required to retain compatibility with such a wide range of systems. Thus, while retaining upward compatibility with RSX11M programming methods, RSX11M-PLUS was directed to this new class of so called "Large Systems". At that time, this was defined as MAPPED systems with 124KW or more memory.

So it is true, that 'M+ was conceived because of the features present in the 11/70, most jealously I&D space support.

My first experience with RSX11M-PLUS was version 2.1, and was on an 11/70. It would turn out that v2.1 of M+ would be proven over time to be another "Gold Standard". At that time, it never occurred to me to perform a SysGen for a target like an 11/23, 11/24 or 11/34 - all of which we had, and ran or sold systems on.

At this time, I do not have access to my M+ 2.1 SysGen kit, so I can't test it. (on 9-track tape) My belief is that it would not produce a system usable on an 11/34 or older UNIBUS CPU, based on the absence of the MTPS class of instruction, which is present in all the machines on the M+ list.

So, from a certain point of view, the machines M+ is compatible with were all destined to be 22-bit capable, although they were not at the time M+ was begun. However, knowing the evolution of the hardware and this research, makes me think the issue was not strictly 22-bit compatibility and that M+ built without I&D support will probably run on 18-bit 11/23 CPUs.

Even if we take the minimum bar beginning with 11/23 (M8186) CPUs that ARE 22-bit enabled, this accounts for all the PDP-11 systems produced with 1982 designs onward, plus the biggest earlier ones. That is an awful lot of the stuff that is still available and running today.

Except for the 11/70 for which it was intended, M+ does preclude the sexy original UNIBUS hardware machines with the switch register front panels, half of which didn't even have an MMU.


Hobbyists today predominantly run RT-11 in my experience. There are scant few of us RSX types on the forums here. I'm not sure I can think of 5 of us who've admitted to it, and I'm pretty much the only one who is vocal about it at this point.

There are 4 RSX versions I am primarily interested in, 3 of which I used in a career. These are:


  • RSX-11M v3.1
  • RSX11M-PLUS v2.1
  • RSX11M-PLUS v3.0 BL24
  • RSX11M-PLUS v4.6 (the final issue)

An 11/73 can run them all, pretty much regardless of how they are sysgened. When built for a particular target, one of these are suitable for any PDP-11 out there, with the possible exception of an 11/03.

Thanks for a great discussion - really.
 
Hi, thanks for valuable comment. I would like to hear more about rsx11m+ bolts/nuts, but probably we need special forum thread for it. Particularly, about drivers differences between rt-11 and rsx. As you rightly noticed many peoples today are running rt-11 probably because nobody need multi-processing capabilities, but also I suspect that is because rsx driver model is not that simple (And for the exUSSR clones with incompatible peripherals this even more important). Good example is HX disk emulator/console multiplexer available for rt-11 but not for rsx-11 (this is special terminal/protocol/driver developed for windows/rt-11, this allows load rt-11 over console SLU to any pdp-11 system and then mount up to 8 virtual disks in rt-11 while still using same SLU as console).

BTW, forgot to say that mentioned in my previous posts VM3 cpu is not 100% compatible, there are differences so original software should be patched sometimes, but still this is very attractive option to build/run single board pdp-11 computer (however, similar result probably could be reached with m7616 / m7554 DEC boards).
 
bigral said:
Good example is HX disk emulator/console multiplexer available for rt-11 but not for rsx-11 (this is special terminal/protocol/driver developed for windows/rt-11, this allows load rt-11 over console SLU to any pdp-11 system and then mount up to 8 virtual disks in rt-11 while still using same SLU as console).
Click to expand...

Wow! I would certainly like to know more about this!
Can you please provide more information or pointers to information?

Thanks in advance.

smp
 
Well, I suspect this all became too much for our thread starter, as it's been more than a month since he's been here. (so it's just for us now)

However, in keeping with the forum's traditions I did try to address my points to his premise. We've all explored the options and complexities involved in taking various PDP-11 "Startup" directions, and I think it's been pretty thorough.

Whether building on actual DEC hardware, pirated National chips (did any make it out alive?), or Russian Clones - the problems of building your own system from scratch (I'm talking getting chips and designing boards) is for all intents and purposes insurmountable to any but the highly capable, and totally insane. (I say this with awe and admiration)

It is certainly not the easy, fast, or cheap way by any means.

I probably took too literally, his comment about "What I am looking to build would be a PDP-11 that is compatible with the vast majority of PDP-11 software".

This is one of the more interesting aspects of this thread for me. Not too many times has a requirement of broadest software compatibility come up. Most new participants are either targeted at a specific software or OS (how many UNIX adventurers have we seen? Must be hundreds), or are hardware centric types exploring a world now gone and hence finding their way to RT - a very capable single user system.

So it's been fun to try to share thinking about how to select hardware that can run almost anything in the PDP-11 software stable. This is actually what the Architecture was all about - being able to migrate your software to newer and faster machines without modification or recompilation. DEC fulfilled this aim to an extreme, and was the first to do so on this scale.

But we as enthusiasts are in the position to now differentiate what the hardware implications are on various software needs, and vice versa. J11 based systems will provide the best opportunity for such modern day explorers, from a software compatibility standpoint.

RT-11 is not particularly advantaged by PDP-11 CPUs with "extended" capabilities. Yes, CIS and FP make differences, but MMU advances - not at all. (once you have one)



In exploring RSX a little, it was my hope to expose some of the heritage the protected memory management model offered, and indeed how it ultimately connected to our modern world. Being an enclave of RT-11 monks, we don't get that opportunity much. Even regular RSX, with it's preemptive multitasking, multiuser capabilities doesn't force one into the more advanced features of the 11/70 architecture.

Opening up the discussion toward M+ and differentiating it was a way to do so. Back in the day, I was unaware and consequently unappreciative of how far back down the food-chain M+ would reach. It brought along the last generation of non-70 architecture machines further than I might have imagined. It was the basis for MicroRSX and for POS - which made it applicable to both the Pro350 and Pro380 which are 11/23 chip and J11 based products.

Our thread starter probably had no idea how broad that "software" expectation would be, nor any awareness of what "software" that might include. - After all, how many operating systems could a processor have? It's almost not something to talk about. 2 OS "families" on a CPU is the most we have today, and more often it's a 1-to-1 correspondence. (yes - I lump all Windows versions together)



To others who would venture upon this thread in the future, hoping to get started in an ancient and obsolete machine called a PDP-11, I will be gratified if they find some useful connections to topics not oft explored.

RSX is destined to be under appreciated and forgotten. It's clear now that in taking the compute experience away from that architecture model and toward a single user linear model we've lost a cultural awareness. PCs did that for/to us. UNIX is now touted as the Ancestor to our present thinking. The influence of RSX on the NT kernel is almost unknown. The influence of RSX on POSIX is even less so.


But at least here in DEC Land, even though it's for a nutty audience of which I am proud to be included as a member, we can speak the truth and play with things once forbidden, now forsaken and soon forgotten.
 
You must log in or register to reply here.
Back
Top