Conference 7.286::digital

    
    In short those chips NVAX CPU are only binned and tested to work
    at 14ns (286MHz). It's luck if they run 100% at higher speed. Some seem
    to work fine but may corrupt data in very subtle ways.
    
    Also, To increase the Bcache they must re-work the modules? Adding the
    larger SRAMS.  They would also have to change the Firmware in FLASH ROM
    to set the CCTL register to the correct values. (Or do this from a
    running VMS system by flushing and jamming the  CCTL)?
    
    Jon
    

    The push-the-clock is an old Nemonix trick; that part is easy (and just
    as risky as it's always been).. but the cache part makes me wonder if
    they're taking old boards, shmooing them to see how far they can push
    them, reworking the cache, and reselling them? no, that won't work..
    there aren't enough wires (if memory serves) in the cache array on the
    L4002 to do a 2MB cache.. Did they design a new board? 
    
    Hmm..
    
    ...tom

    Having actually worked with this "company" for about a year in 92...I 
    can safely assume that they:
    
    1. Bought a few boards on the used market and made whatever
    modification was needed to add the cache
    2. Take the customers board in trade at sale time and gives the
    customer the all ready modified board
     
    They are able to offer DEC service, and most likely have made up
    several boards to give to DEC as spares.
    
    This was the approach to a number of their products when I was there.
    I can almost be certain that they did not design a new board.
    
    As .2 stated, "push the clock" is the standard technology at Nemonix.
    I worked a lot better in the older DEC products (8000's, 6000's) than
    in the newer products (4000's) when I was there.
    
    Want to know more....contact me off-line.  

  I've stated before that "On the VAX 8800, in my lab, the
  Nemonix accelerator caused the VAX to be clocked faster than
  the fastest speed at which it would function." The I/O adapter
  failed before the CPU, but a CPU without I/O isn't very useful
  either.

  I have a very pessimistic view of the performance that can
  be gained by "clock-chipping". With a very few, well-known,
  marketing-driven exceptions, when you speed up the clock, I
  believe you trade away a great deal of the "design margin"
  that the engineers built into the system.

                                   Atlant

    
    A prominant VP, Consulting Engineer, and one-time VAX development
    manager once commented on the risks taken by customers who buy
    accelerators - particularly those accelerators which, through simple
    clock-speed increases, reduce the designed-in timing margins used as 
    the only means of data path protection - the accelerated system could
    " ... give wrong answers with a straight face ..." (a near-quote).
    
    I don't know how far we went with the implementation, but there was
    once a plan to have the power-up diagnostics test for the presence of
    an accelerator. If detected for the first time,  the customer /
    operator would be required to acknowledge that Digital would not be
    held liable for system errors or inaccuracies caused by the accelerated
    clock speed. Once acknowledged, the boot process would complete.
    
    
    	mark

The console for VAX 6000 systems tests for clock speed-up at power-up.  If 
the clock is found to be running higher than the speed set by Digital, the 
console prints out a 1-page warning message.


							-Paul

    I know the display of this error msg was a "minor inconvenience" in the
    sales process when I was there...I know they tried to have it removed,
    and they might have been successful once the product was accepted under
    a Field Service contract. Also, I remember some attempts to disable the
    message display through software tweaks - they used to have some
    software engineers come in and reverse engineer some of the system
    software to patch in fixes that they needed. Dont remember if it worked
    in this case or not.

    
    Actually, engineering's decision to put the power-up test was, at least
    in part, due to the fact that service was willing to take the Nemonix
    accelerators under a maintenance contract. They were concerned that if
    "Digital" accepted the maintenance contract, then customers might also
    think that "Digital" supported the accelerators [and their potential
    consequences] from an engineering standpoint. The console message was
    intended to make thier position clear, irrespective of what services
    chose to support.
    
    	mark