DEC MKA11 and/or M8261/M8163

[Johnny Billquist]

On 2015-02-15 11:43, Paul Birkel wrote:
> On Sat, Feb 14, 2015 at 2:24 PM, Johnny Billquist <bqt at update.uu.se> wrote:
>
>> On 2015-02-14 19:14, Paul Birkel wrote:
>>
>>> Thanks Eric.  That might explain how I came into possession of one M8162
>>> module.  The 11/74 never entered my mind because encountering even a piece
>>> of one of those is ... unlikely :->.  Using up unsold inventory makes
>>> sense; I guess that an MKA11 wouldn't function without a full complement
>>> of
>>> modules, even if some serve no real purpose.
>>>
>>
>> The MKA11 is really not much different from an MK11, and you can use an
>> MKA11 in a normal 11/70 system as well as far as I understand. Essentially,
>> the MKA11 is just a multiported MK11 box. And the multiporting is achieved
>> through this extra module, that fits (as far as I can remember) in the
>> standard MK11 box.
>>
>
> As I understand the wording in the slim documentation, each port will
> require another pair of M8158/M8159 Address/Data Buffers.  So a dual-ported
> unit would need two of these pairs, and a quad-ported unit four pairs.  The
> M8162 then coordinates the (up to 4) pairs.

Hmm. That might be, I should read through the documentation.

>> All the other modules in the MKA11 are the same as in the MK11.
>
>
> Except the numbers of buffer pairs, there being up to 4 pairs possibly
> installed?


If you need one M8158/M8159 per CPU, in addition to the M8162 (one per 
two CPUs?), that means a lot of cards.

> The MK11 chassis was already full, so they would have had to reduce the
> complement of MOS-modules in the MKA11.  By the time of the 11/74
> presumably it was the case that increased-density ICs would have allowed
> more-RAM-per-module, so the loss in MOS-module count per MKA11 would have
> been acceptable.

No. Unfortunately it will not work that way.
The MK11 have a backplane which uses a special addressing scheme which 
sets the maximum memory size per card to 128Kword. So a fully stocked 
normal MK11 is 2 Mword. If you reduce the number of slots for memory 
cards, you will simply not be able to get the full 2Mword in one box, 
but will be forced to go for 2 boxes.
The original memory cards for the MK11 is 32Kword, which meant you had 
to have a number of MK11 boxes to get to max memory. I'm not sure the 
128Kword cards were even available when the 11/74 was made.

The limitation isn't only the lack of more address pins for the memory 
cards. Each card slot have a card enable signal, in addition to the 
address pins. At reset time, the M8158/M8159 sets up the ECC of all 
cards in parallel, by enabling all the card select pins in parallel and 
writing data. So in essence, all memory cards are written to in parallel 
at reset.

As a small side note, the VAX-11/750 uses the same memory card bus as 
the MK11. So you can move 32Kword and 128Kword memory cards between an 
11/70 and a VAX-11/750. However, the VAX-11/750 also supports 512Kword 
cards, but those can not be used an a MK11 box without both hardware and 
software hacking. (If someone wants to try this, let me know. I can 
provide more details on what is needed to be done in order to do this, 
since I have done it in the past.)

>>   I've looked more closely at the pair of top-edge connectors (the ones like
>>> the 20ma SLU connectors).  Each have 6-of-8 pins wired *thru* a 1K 1%
>>> resistor to a separate pin on tab F.  So there are twelve lines running to
>>> tab F, with resistors in series, and nothing else on those paths.
>>>
>>> The rest of the module is a mix of S and LS SSI chips, and a couple of
>>> what
>>> appear to be multi-tap delay lines, plus 5 diagnostic LEDs (unlabeled).
>>>
>>> Would the KB11-CM have had special cabling to connect in this manner for
>>> out-of-band signaling, and then why series resistors (rather than, e.g.,
>>> pull-ups for an OC line driver)?
>>>
>>
>> I don't think so. There is not out-of-band signalling. But you do have a
>> control panel for the memory box, on which you can control the box. I would
>> suspect this would be for that.
>>
>
> Right (line-diagram Figure 1-1 is pretty crude).
>
>
>> (I think there are some extra controls for an MKA11 compared to the
>> standard panel of an MK11, which is for (I think) setting the memory
>> on-/offline for individual CPUs, and possibly something more. It should be
>> in the documentation I put in a link to.
>>
>
> Beyond the claim that "Port Controls" exist, and where they get installed
> in the racks, I don't see any.
>
> Best available images seem to be at:
> http://oboguev.livejournal.com/2696291.html

Those are copies of some of the pictures Dave Carroll made a bunch of 
years ago. The originals are here: 
https://plus.google.com/photos/104560031285123036919/albums/5632654788488216449?banner=pwa

> I like the user-supplied color-coded dots :->.  Can *almost* decipher the
> front panel labeling, too.
>
> That said, they look to me to be identical to the standard MK11 "Box
> Controller".  No clue there what the extra cabling might be doing.  Maybe
> the extra cabling doesn't actually involve the "Box Controller" on the
> front panel at all.  Perhaps simply cross-linkage among the Port
> Multiplexers in different MKA11 (can't think of a good reason for that,
> offhand)?

I would have to research this more to answer. :-)
But you are right in that it appears each CPU have its own panel for 
each box, meaning the same box could appear at different addresses for 
different CPUs. I had forgotten that. That is pretty cool, and I guess 
that is how they managed to run diagnostics on one CPU while the others 
were running in normal production. (There are special tools in 
RSX-11M-PLUS to do some cool stuff around this.)

I wonder if there might be some connection between the memory boxes and 
the IIST?

	Johnny

>
>          Johnny
>>
>>
>>   On Wed, Feb 11, 2015 at 2:55 PM, Eric Smith <spacewar at gmail.com> wrote:
>>>
>>>   On Sat, Feb 7, 2015 at 7:25 AM, Paul Birkel <pbirkel at gmail.com> wrote:
>>>>
>>>>> I don't understand the role of the M8162 "Port MUX A module" and M8163
>>>>> "Port MUX B module" that the field-guide claims to be part of the MK11.
>>>>> They aren't part of the MK11 as far as any documentation that I've read
>>>>> regarding it. [...]
>>>>> Do these modules really belong to the MKA11, rather than the MK11?
>>>>>
>>>>
>>>> The MKA11 was the multiported version of the MK11, intended for
>>>> multiprocessor (11/72 and 11/74) systems. it makes sense that it would
>>>> have some "port mux" modules, though I've never seen any documentation
>>>> on it.
>>>>
>>>> Reportedly some 11/70 systems were sold to AT&T with the KB11-CM CPU
>>>> and MKA11, apparently just as a way to use up the hardware that was
>>>> built for the cancelled 11/74.
>>>>
>>>>
>>
>> --
>> Johnny Billquist                  || "I'm on a bus
>>                                    ||  on a psychedelic trip
>> email: bqt at softjar.se             ||  Reading murder books
>> pdp is alive!                     ||  tryin' to stay hip" - B. Idol
>>


-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol