Conference 7.286::digital

    We make our own microprocessors, memory/disk/video controller chips
    and other support chips.  We buy DRAMs from several sources.
    Semiconductor Engineering puts out a fascinating newsletter
    (used to be called LSI Focus, now something else) that tells the
    rest of DEC what SEG has available and what's coming.  The newsletter
    is restricted-distribution, so it may be hard to get ahold of.
    Also, if the jargon of semiconductor engineers gives you MEGO
    syndrome (My Eyes Glaze Over), you may want to skip it.
    
    					Steve

    Actually, the last time I heard, DEC was the sixth largest manufacturer
    of semiconductor devices in the world. 
    
    IBM, of course, is first.

    re 2.
    I don't know for sure, but I seriously doubt that DEC is the sixth
    largest manufacturer of semiconductors. Take a look at companies
    like Intel, Motorola, TI, and a whole slew of Japanese companies
    that make literally millions of chips. Remember DEC only makes chips
    for their own use. Now, IBM being one of the largest consumers,
    I can see how IBM is the largest. I think IBM might also sell chips
    too. I think SCO here in Hudson tends to be more R & D than mass
    production.

    
    
    re .2  & .3  I think the claim should of been originaly stated:
    DEC is the sixth largest CAPTIVE chip manufacturer, no way were
    even close to sixth in the whole industry, unless of course we have
    about eight fabs the company is keeping a secret from us.
            
    
    Rob

>    I think SCO here in Hudson tends to be more R & D than mass
>    production.

FYI: almost all of the chips produced by DEC are manufactured in Hudson, MA

    
    	re.3
    
    	DEC does/did sell chips on the outside market (T-11, LSI-11
    chip set etc.)
    
    	re Hudson not being a volume manufacturer.... So THATS why we
    cant get uVax CPU chips here!!!    :-)
    
    	-John

    Not all of our microprocessors are manufactured internally...
    
    The Rainbow has an 8088 from Intel and a Z80 from Zilog (?).  The
    VAXmate has an Intel 80286.  I think that some of our peripherals (the
    LN03R) have Motorola 68000 microprocessors.
    
    Who knows, we might even have a machine with an Intel 80386 in the
    works....?
    
    /Brett
    

        Just so there is no misunderstanding, we do produce, in the
        volumes we need, our custom chips.  For MicroVAX and CVAX
        that's no small number.

        Parts like Q-Bus interface are produced elsewhere.

        We do actively sell chips. I provide, through MSD, the technical
        customer support.

        And yes, we don't produce commodity parts.

    Re .7 (Brett):
    
    ... and the DECmate II/III system uses a 6120 chip from (?) [I think
    there are two sources].
    
    Steve Kallis, Jr.

    The 6120 chip is from Harris, and before that, Intersil.  I think
    the J-11 chip was supposed to be from Harris also, but there were
    problems and now we make it ourselves.
    
    					Steve

    Yea, the J11 was initially built by Harris, but that was such a
    black hole for money that Hudson finally decided to build it
    themselves.(a good decision in my mind)
    
    
    For a while, rumor had it that Harris was going to fold up the
    semiconductor operation and  call it quits.  The people in Hudson
    who designed the J11 and were buying them from harris were quite
    nervous at the time.
    
    Do they still make chips at Harris?
    
    -bill

>   Do they still make chips at Harris?
    
        Yes.

Steve Lionel:

  I don't think Intersil ever made the 6120.  They *DID* make the 
  6102 though (the earlier, slower CMOS-8 used in the WT/WS-78).
  They also sold the 6102 commercially on their own.


RE Harris:

  A long time ago, there was a Digital centerfold ad for the "J11
  Family" (back when we intended to sell chips, boards, and systems)
  in some electronics trade newspaper which showed a powerful-looking
  stallion rearing up.  The caption was "Bred for performance..."

  (This was at a time when the "20 MHz" J11 was actually clocking 
  at somewhere between 10 and 400 KILOHERTZ.)

  A copy of the ad was posted on 5-5.  Below the caption, some wag
  had scrawled: "Digital sues for return of the stud fee".

                                   Atlant

        Just so people wont get the wrong idea on the DCJ11...

        We do sell the part externally. Most are 15 MHz parts but there
        are some 18 MHz parts now being sold. Most of DEC's products
        use the 18 MHz part. The (late) PRO380 used a 10 MHz bin.

        If you have an interested customer, contact Kathy Breda
        (RETORT::BREDA).  She is located at DLB in Marlboro. It is her
        group that must handle volume contracts.

        The only parts we use or sell are made by Harris in Florida.
        Hudson has the process and has made some parts. These must be
        qualified in product before volume production and the demand
        increased.

        As far as I know there was never a J11 "family" unless someone
        pointed out that the J11 was really two chips on a hybrid or
        stuck in the DC319 DLART as a family member.

    Ok, yes, Intersil didn't make the 6120.  Didn't Western Digital come
    out with the 6100, the first PDP-8 chip?
    
    We also sell the DCT11 externally, in HUGE quantities.  Atari is one
    of the biggest customers.  And there are of course always rumors that
    we might get around to selling the 78R32 (RTVAX) or a CVAX equivalent,
    but I wouldn't place any bets.
    
    					Steve

    RE:.14
    
    
    Mike,   I also thought that we sold the FPJ11 chip to a few customers.
    I can certainly remember phone calls asking how the damned thing
    worked.  I didn't hang around long enough to see any of the designs
    go into production though.
    
    
    RE: J11
    
    I think if you look at all of the J11 designs (both internal and
    external) you'll find that the Atex machine is the WORLDS FASTEST
    PDP11 that is sold today.  (yup, they beat us)  Their design is
    really a piece of work.  They actually stop the clock and wait for
    the memory to catch up, thus giving them a 6 clock tick read instead
    of the minimum 8 tick read (unless of course you hit cache)  It
    really made for a FAST machine.
    
    
    -bill

    If you are following this note with interest, and I know you are,
    the anonymous "Charlie Matco" column of the current Digital Review
    discusses in less than complimentary terms Digital's chips and PDP-11
    strategy.  It is, of course, incorrect but since 100,000 people have
    read what's written there, it creates its own importance.
    
    I definitely won't post any text from it here as I fear it would be
    extracted and circulated within the company labelled as "my" position.

        Yes, the FPJ11 is sold with the DCJ11. The "warm" floating
        point in the DCJ microcode isn't fast enough for most.

        Pat! That last sentence sounds a bit paranoid :-)

    RE:.16
      If you study the design of the KDJ11-B, you'll find two entirely
    different clock designs - one which stops and one which doesn't.
    The J-11 is clocked by a crystal oscillator at a frequency of either
    15 MHz or 18 MHz and is free-running.  The state machine which inter-
    faces to the PMI/Q-BUS, has an asynchronously started, delay-line
    based clock, which allows the state machine to be "asleep" when no
    bus activity is impending AND to "wake up" in a few nanoseconds
    when a starting event is detected.  If memory serves, the period
    on this clock was about 85 nsec.  Apparently, ATEX actually stopped
    the clock to the J-11, which should be doable with CMOS but was
    not done in any DEC designs of which I am aware.
    
      I have been told that J-11's were used in experiments at liquid
    nitrogen temperatures at frequencies well above 18 MHz.  I believe
    a paper was presented at ISSCC a few years back describing some
    of this work.
    
    Greg Opp 
    

    Yup, ATEX stops the clock to the J11.  They don't use  crystal
    but drive the CLK2 signal from an external source.
    
    If you don't want to know the details, go the the next note now,
    otherwise, a little more about the J11
    
    
    The J11 is somewhat strange when it does reads.  If you can round
    trip a read in less than 4 clock ticks, you do it synchronously
    and the thing goes very fast.  This is where most cache designs
    fit in.  If you miss the 4 tick mark, the machine starts going
    asynchronous and you have a minimum of 8 clock ticks, 10 if you
    want parity.  What Atex did was to stop the clock so the processor
    saw all reads as 4tick reads.   The acutal time of a read was about
    5 ticks or so of the clock if it was running continuously.
    
    
    A really interesting design. And it really GOES!
    
    
    -bill

    RE: .20
      Does this imply that their CPU is entirely asynchronous logic
    circuits with no master clock?  If not, how do they establish
    synchronization between the J-11 and other circuits on board?
    Thanks,
    
    Greg

Greg:

  ATEX had a lot of prior experience hackingthe 11/34(A).  The 11/34
  has no free-running clock -- All CPU timing was established from a
  synchronously-stoppable, asynchronously startable delay-line oscil-
  lator.  All UNIBUS timing was done by a similar, but separate timing
  line.  Then, by adding enough green wires from deep in the machine
  to some external logic and memory, ATEX not only achieved many mega-
  bytes of physical memory (on our 256 KByte machine) but they also
  achieved an effective memory access time equivalent to UNIBUS memory
  running at 50nS access times!!

  All in all, it was a truly wonderful application of DEC Iron.
  (I don't suppose they ever returned many CPU modules under the 
  warranty program, though. ;-) )

  I'd bet you'd find that they adapted the 11/34A-style clocking 
  scheme to the J11.

                                   Atlant

But Vikas...

  I thought about that as I wrote, and I specifically chose the
  title for my reply to tie us back into the "DEC way of working"
  theme!  I'd actually like to see some discussions arise from this.

> < Note 508.22 by ATLANT::SCHMIDT >
>                -< DEC Customers can sometimes teach us a lot! >-

                                   Atlant