Loading ...
Sorry, an error occurred while loading the content.
 

PDP 11/44 11/70 DataRAM 40785 RAM Card with 152 4116's = 128K?

Expand Messages
  • billdeg
    ok...how much ram do you have if you have a 19 x 8 array of 4116 chips? I have a DataRam model 40785 for sale on Ebay, did not find info on the board
    Message 1 of 18 , Aug 14, 2014
      ok...how much ram do  you have if you have a 19 x 8 array of 4116 chips?  I have a DataRam model 40785 for sale on Ebay, did not find info on the board anywhere, but from the specs of the 4116, and assuming there is something unique I dont understand about the way RAM works on a PDP 11, it would be 128K?  This is a PDP 11 class system RAM board given it fits in a UNIBUS backplane..Anyone worked with DataRAM RAM boards that can confirm 128k?Bill

    • David Riley
      ... Are you sure it s not 18x8 chips? That would make more sense; it s 8 ranks of 16 bits + 2 parity, which (not counting the parity) is 256k bytes (128k
      Message 2 of 18 , Aug 14, 2014
        On Aug 14, 2014, at 21:47, billdeg <no_reply@yahoogroups.com> wrote:

         

        ok...how much ram do  you have if you have a 19 x 8 array of 4116 chips?  I have a DataRam model 40785 for sale on Ebay, did not find info on the board anywhere, but from the specs of the 4116, and assuming there is something unique I dont understand about the way RAM works on a PDP 11, it would be 128K?  This is a PDP 11 class system RAM board given it fits in a UNIBUS backplane..Anyone worked with DataRAM RAM boards that can confirm 128k?Bill

        Are you sure it's not 18x8 chips? That would make more sense; it's 8 ranks of 16 bits + 2 parity, which (not counting the parity) is 256k bytes (128k words) for 4116 RAMs.

        I don't know what the additional bit would be for if it were actually 19x8. I've definitely miscounted on boards like that before.

        - Dave
      • Dave McGuire
        ... I don t know this board in particular, but DataRam did a lot of fairly in-depth design of boards for DEC systems, not just cloning, and not just RAM. They
        Message 3 of 18 , Aug 14, 2014
          On 08/14/2014 09:47 PM, billdeg wrote:
          > ok...how much ram do you have if you have a 19 x 8 array of 4116
          > chips? I have a DataRam model 40785 for sale on Ebay, did not find info
          > on the board anywhere, but from the specs of the 4116, and assuming
          > there is something unique I dont understand about the way RAM works on a
          > PDP 11, it would be 128K? This is a PDP 11 class system RAM board given
          > it fits in a UNIBUS backplane..Anyone worked with DataRAM RAM boards
          > that can confirm 128k?

          I don't know this board in particular, but DataRam did a lot of fairly
          in-depth design of boards for DEC systems, not just cloning, and not
          just RAM. They did a very good job of that stuff, too. It's likely
          that it's something "completely different".

          I have many dozens of pounds of DataRam manuals. I'm on the road at
          the moment, but when I get home I'll see if I can find the manual for
          this board. It's likely that I have it.

          One thing to be careful of: some DataRam systems use "hex-wide Qbus".
          Yes, you read that right. It's a backplane that takes dual- and
          quad-width Qbus boards, but is actually six slots wide, in a
          sorta-serpentine arrangement. The danger, of course, is that it looks
          exactly like a Unibus backplane. This is a smoke/flames situation if
          one gets this wrong. The DataRam "A23" system is one of these, for
          example. I don't know if they made any hex-wide Qbus boards to go with
          these backplanes, other than just using them with duals and quads, but
          there exists at least the possibility that they did...so don't go
          plugging that board into a Unibus backplane until you verify that it's
          actually Unibus. This goes for any DataRam board.

          Now, that said, with 4116s on it, that places it more in the Unibus
          era...but still..

          One thing DataRam was very good at was ECC systems. It might be that
          this RAM isn't just ordinary parity (implemented with 74S280s) but it
          might be an ECC setup, which uses more bits. If there are lots of PALs,
          that's likely what it is. It's also possible that the additional DRAM
          chips are spares, to be wired in when in-array 4116s fail. They and
          others did that in a few designs.

          Either way, though, it's in the right range for it to be 128KB, or 64kW.

          -Dave

          --
          Dave McGuire, AK4HZ
          New Kensington, PA
        • Dan Roganti
          On Thu, Aug 14, 2014 at 10:54 PM, David Riley fraveydank@gmail.com ... looking closely at the photo on your listing, ​I actually see alternating rows of
          Message 4 of 18 , Aug 14, 2014



            On Thu, Aug 14, 2014 at 10:54 PM, David Riley fraveydank@... [midatlanticretro] <midatlanticretro@yahoogroups.com> wrote:


            On Aug 14, 2014, at 21:47, billdeg <no_reply@yahoogroups.com> wrote:

             

            ok...how much ram do  you have if you have a 19 x 8 array of 4116 chips?  I have a DataRam model 40785 for sale on Ebay, did not find info on the board anywhere, but from the specs of the 4116, and assuming there is something unique I dont understand about the way RAM works on a PDP 11, it would be 128K?  This is a PDP 11 class system RAM board given it fits in a UNIBUS backplane..Anyone worked with DataRAM RAM boards that can confirm 128k?Bill

            Are you sure it's not 18x8 chips? That would make more sense; it's 8 ranks of 16 bits + 2 parity, which (not counting the parity) is 256k bytes (128k words) for 4116 RAMs.

            I don't know what the additional bit would be for if it were actually 19x8. I've definitely miscounted on boards like that before.

            - Dave

            looking closely at the photo on your listing, ​I actually see alternating rows of 4116's
            starting at row1:
            19x, 20x, 19x, 20x, 20x, 19x, 20x, 19x : 4116's
            16x, 16x, 16x, 16x, 16x, 16x, 16x, 16x :dataword
              3x,   4x,  3x,   4x,   4x,  3x,   4x,   3x :extra bits

            So this would leave alternating 3 and 4 bits per bank
            Not enough for ECC, where a 6 bit hamming code is needed for a 16bit data word.
            So then it would need 22x 4116's per row.
            Maybe they devised an alternate error detection scheme.
            I didn't see a manual for this at Bitsavers, did you have one, Bill ??
            From looking at the manuals for the MS11 PDP Mos Memory boards
            they only indicate Parity was being used on their boards
            And those don't show any more than 2 parity bits.
            Dan
          • Paul Birkel
            I believe that you re looking at a substitute for a MS11-KE. See: EK-0MK11-TM-001_MK11mem.pdf It s a 39-bit double-word module, so two rows of RAM chips
            Message 5 of 18 , Aug 15, 2014
              I believe that you're looking at a substitute for a MS11-KE.  See: EK-0MK11-TM-001_MK11mem.pdf
               
              It's a 39-bit double-word module, so two rows of RAM chips correspond to a double-word.  It can detect double-bit errors and correct single-bit errors.
               
              The DEC part would be 32K-word given 4K RAM chips; in this case it looks like you have 128K-word.
               
              It's for an 11/70.  You need to install them in pairs.  Do you have two? 


              On Thu, Aug 14, 2014 at 9:47 PM, billdeg <no_reply@yahoogroups.com> wrote:
               

              ok...how much ram do  you have if you have a 19 x 8 array of 4116 chips?  I have a DataRam model 40785 for sale on Ebay, did not find info on the board anywhere, but from the specs of the 4116, and assuming there is something unique I dont understand about the way RAM works on a PDP 11, it would be 128K?  This is a PDP 11 class system RAM board given it fits in a UNIBUS backplane..Anyone worked with DataRAM RAM boards that can confirm 128k?Bill


            • billdeg
              I triple checked, yes it s 19x8 rows.I only have the one. Given the 4116 s are 1K, I assumed that 128K from the DEC perspective would be more total
              Message 6 of 18 , Aug 15, 2014
                I triple checked, yes it's 19x8 rows.I only have the one.  Given the 4116's are 1K, I assumed that "128K" from the DEC perspective would be more total instruction space than from say the Altair BUS perspective if the words were larger.  There are not enough chips to make it to 256K given it's a DEC.Hope it finds a good home.Bill
              • billdeg
                thanks for the info, I have revised my Ebay listing. Much appreciated.Bil
                Message 7 of 18 , Aug 15, 2014
                  thanks for the info, I have revised my Ebay listing.  Much appreciated.Bil
                • Dave McGuire
                  On 08/15/2014 01:00 AM, Dan Roganti ragooman@gmail.com ... They have done several over the years. The only data on them was in the manuals; I don t think they
                  Message 8 of 18 , Aug 15, 2014
                    On 08/15/2014 01:00 AM, Dan Roganti ragooman@...
                    [midatlanticretro] wrote:
                    > looking closely at the photo on your listing, ​I actually see
                    > alternating rows of 4116's
                    > http://i.ebayimg.com/00/s/OTM1WDE2MDA=/z/~ogAAOSwq7JT7VFN/$_57.JPG
                    > starting at row1:
                    > 19x, 20x, 19x, 20x, 20x, 19x, 20x, 19x : 4116's
                    > 16x, 16x, 16x, 16x, 16x, 16x, 16x, 16x :dataword
                    > 3x, 4x, 3x, 4x, 4x, 3x, 4x, 3x :extra bits

                    > So this would leave alternating 3 and 4 bits per bank
                    > Not enough for ECC, where a 6 bit hamming code is needed for a 16bit
                    > data word.
                    > So then it would need 22x 4116's per row.
                    > Maybe they devised an alternate error detection scheme.

                    They have done several over the years. The only data on them was in
                    the manuals; I don't think they released any algorithms for the greater
                    good at the time.

                    > I didn't see a manual for this at Bitsavers, did you have one, Bill ??
                    > From looking at the manuals for the MS11 PDP Mos Memory boards
                    > they only indicate Parity was being used on their boards
                    > And those don't show any more than 2 parity bits.

                    Nearly all DEC boards of that era used parity, not ECC.

                    -Dave

                    --
                    Dave McGuire, AK4HZ
                    New Kensington, PA
                  • Paul Birkel
                    If you say so, but the photo clearly shows what I described (4 rows of 19 & 4 rows of 20 = 4x39-bits; see Don s response). And 4116 are 16Kx1-bit (TI example:
                    Message 9 of 18 , Aug 15, 2014
                      If you say so, but the photo clearly shows what I described (4 rows of 19 & 4 rows of 20 = 4x39-bits; see Don's response).  And 4116 are 16Kx1-bit (TI example: http://hardware.speccy.org/datasheet/4116-15.pdf).
                       
                      I stand by my interpretation :->.  I believe it's the only Unibus module like that ... although the somewhat-related PEP-70 is a quite interesting subsystem-replacement.


                      On Fri, Aug 15, 2014 at 9:47 AM, billdeg <no_reply@yahoogroups.com> wrote:
                       

                      I triple checked, yes it's 19x8 rows.I only have the one.  Given the 4116's are 1K, I assumed that "128K" from the DEC perspective would be more total instruction space than from say the Altair BUS perspective if the words were larger.  There are not enough chips to make it to 256K given it's a DEC.Hope it finds a good home.Bill


                    • Dave McGuire
                      ... 1K? No...the 4116 is 16K bits, 1 bit wide. -Dave -- Dave McGuire, AK4HZ New Kensington, PA
                      Message 10 of 18 , Aug 15, 2014
                        On 08/15/2014 09:47 AM, billdeg wrote:
                        > I triple checked, yes it's 19x8 rows.I only have the one. Given the
                        > 4116's are 1K, I assumed that "128K" from the DEC perspective would be
                        > more total instruction space than from say the Altair BUS perspective if
                        > the words were larger. There are not enough chips to make it to 256K
                        > given it's a DEC.Hope it finds a good home.

                        1K? No...the 4116 is 16K bits, 1 bit wide.

                        -Dave

                        --
                        Dave McGuire, AK4HZ
                        New Kensington, PA
                      • Dan Roganti
                        On Fri, Aug 15, 2014 at 7:42 AM, Paul Birkel pbirkel@gmail.com ... ​oh yea, that would make sense You need 7bit hamming code for 32bit data​ ​for ECC to
                        Message 11 of 18 , Aug 15, 2014



                          On Fri, Aug 15, 2014 at 7:42 AM, Paul Birkel pbirkel@... [midatlanticretro] <midatlanticretro@yahoogroups.com> wrote:


                          I believe that you're looking at a substitute for a MS11-KE.  See: EK-0MK11-TM-001_MK11mem.pdf
                           
                          It's a 39-bit double-word module, so two rows of RAM chips correspond to a double-word.  It can detect double-bit errors and correct single-bit errors.
                           
                          The DEC part would be 32K-word given 4K RAM chips; in this case it looks like you have 128K-word.
                           
                          It's for an 11/70.  You need to install them in pairs.  Do you have two? 


                          ​oh yea, that would make sense
                          You need 7bit hamming code for 32bit data​
                           
                          ​for ECC to get SECDED
                          Dan



                        • David Riley
                          ... 4116s are 16k x 1 bit. Similarly, 4164s are 64k x 1 bit, and 41256s are 256k x 1 bit. 41464s are 64k x 4 bit; the numbering breaks down a bit after that.
                          Message 12 of 18 , Aug 15, 2014
                            On Aug 15, 2014, at 9:47 AM, billdeg <no_reply@yahoogroups.com> wrote:

                            > I triple checked, yes it's 19x8 rows.I only have the one. Given the 4116's are 1K, I assumed that "128K" from the DEC perspective would be more total instruction space than from say the Altair BUS perspective if the words were larger. There are not enough chips to make it to 256K given it's a DEC.Hope it finds a good home.Bill

                            4116s are 16k x 1 bit. Similarly, 4164s are 64k x 1 bit, and 41256s are 256k x 1 bit. 41464s are 64k x 4 bit; the numbering breaks down a bit after that.


                            - Dave
                          • Dave McGuire
                            On 08/15/2014 09:55 AM, Paul Birkel pbirkel@gmail.com [midatlanticretro] ... Urr? Maybe I m misinterpreting what you re saying, but the memory bus in an
                            Message 13 of 18 , Aug 15, 2014
                              On 08/15/2014 09:55 AM, Paul Birkel pbirkel@... [midatlanticretro]
                              wrote:
                              > I stand by my interpretation :->. I believe it's the only Unibus module
                              > like that ...

                              Urr? Maybe I'm misinterpreting what you're saying, but the memory bus
                              in an 11/70's MK11 memory chassis bears little resemblance to Unibus.

                              > although the somewhat-related PEP-70 is a quite
                              > interesting subsystem-replacement.

                              Yes. I'd love to stumble across one of those.

                              -Dave

                              --
                              Dave McGuire, AK4HZ
                              New Kensington, PA
                            • Paul Birkel
                              Yes, I was speaking era , rather than literally. My ambiguity .... On Fri, Aug 15, 2014 at 9:55 AM, Dave McGuire Mcguire@neurotica.com ... Yes,  I was
                              Message 14 of 18 , Aug 15, 2014
                                Yes,  I was speaking "era", rather than literally.  My ambiguity ....


                                On Fri, Aug 15, 2014 at 9:55 AM, Dave McGuire Mcguire@... [midatlanticretro] <midatlanticretro@yahoogroups.com> wrote:
                                 

                                On 08/15/2014 01:00 AM, Dan Roganti ragooman@...


                                [midatlanticretro] wrote:
                                > looking closely at the photo on your listing, ​I actually see
                                > alternating rows of 4116's
                                > http://i.ebayimg.com/00/s/OTM1WDE2MDA=/z/~ogAAOSwq7JT7VFN/$_57.JPG
                                > starting at row1:
                                > 19x, 20x, 19x, 20x, 20x, 19x, 20x, 19x : 4116's
                                > 16x, 16x, 16x, 16x, 16x, 16x, 16x, 16x :dataword
                                > 3x, 4x, 3x, 4x, 4x, 3x, 4x, 3x :extra bits

                                > So this would leave alternating 3 and 4 bits per bank
                                > Not enough for ECC, where a 6 bit hamming code is needed for a 16bit
                                > data word.
                                > So then it would need 22x 4116's per row.
                                > Maybe they devised an alternate error detection scheme.

                                They have done several over the years. The only data on them was in
                                the manuals; I don't think they released any algorithms for the greater
                                good at the time.


                                > I didn't see a manual for this at Bitsavers, did you have one, Bill ??
                                > From looking at the manuals for the MS11 PDP Mos Memory boards
                                > they only indicate Parity was being used on their boards
                                > And those don't show any more than 2 parity bits.

                                Nearly all DEC boards of that era used parity, not ECC.


                                -Dave

                                --
                                Dave McGuire, AK4HZ
                                New Kensington, PA


                              • billdeg
                                Meaning you need 16 4116 s to equal 16K, right? Not very accurate on my part.Bill
                                Message 15 of 18 , Aug 15, 2014
                                  Meaning you need 16 4116's to equal 16K, right? Not very accurate on my part.Bill
                                • Paul Birkel
                                  16 x 4116 chips = 16K-words = 32K-bytes
                                  Message 16 of 18 , Aug 15, 2014
                                    16 x 4116 chips = 16K-words = 32K-bytes


                                    On Fri, Aug 15, 2014 at 11:23 AM, billdeg <no_reply@yahoogroups.com> wrote:
                                     

                                    Meaning you need 16 4116's to equal 16K, right? Not very accurate on my part.Bill


                                  • billdeg
                                    RAM is a mystery to me, I really should learn this better, thanks.
                                    Message 17 of 18 , Aug 15, 2014
                                      RAM is a mystery to me, I really should learn this better, thanks.
                                    • David Riley
                                      ... 16 4116s in parallel would give you 16k words (assuming you mean 16-bit words, which DEC usually does) or 32k bytes. 18 4116s in parallel would give you
                                      Message 18 of 18 , Aug 15, 2014
                                        On Aug 15, 2014, at 11:23 AM, billdeg <no_reply@yahoogroups.com> wrote:

                                        > Meaning you need 16 4116's to equal 16K, right? Not very accurate on my part.Bill

                                        16 4116s in parallel would give you 16k words (assuming you mean 16-bit words, which DEC usually does) or 32k bytes. 18 4116s in parallel would give you 16k words with parity bits. The arrangement described by Paul would follow assuming a doubleword-wide interface with some sort of ECC or similar (16k x 39 x 4 ranks, still 256k bytes/128k words in the end).

                                        The idea behind 1-bit RAMs is that you use one for each bit lane, which is why you usually see them in groups of 8 or 16 (or 18, or 39 in this case). That way, you have a 16k x <word size> array that you can address easily. More "ranks" can be added by multiple CAS signals, usually derived from a decoder of some sort that operates on a few address bits, usually the upper ones.


                                        - Dave
                                      Your message has been successfully submitted and would be delivered to recipients shortly.