Conference ricks::dechips

	The customer is less concerned with what is shipping, tested,
	and working than whether it's .25 or .35 micron? Sounds like
	HP FUD (tm) to me. Counter with proven products that get their
	work done today. Ask them if they are willing to be a field
	test for unproven, untested systems and operating systems.

	I know this doesn't answer your question, I'm sure someone
	will. It just irks me that HP can get away with this time
	after time.

							mike

    I would say that the 600MHz 21164 uses 0.25um transistors
    with 0.35 metal.  It will be shipping in the 2nd half of 1997.
    
    Full 0.25um (CMOS7) will arrive in the first half of 1998.
    

>    I would say that the 600MHz 21164 uses 0.25um transistors
>    with 0.35 metal.  It will be shipping in the 2nd half of 1997.
    
    If you are willing to say this, why not say we have been shipping our
    21164 using 0.25um transistors and 0.35 metal since last year.
    
    
    
    
    

:    ...that HP was releaseing this early because
:    they don't want to wait for the Intel Delays in Merced.

   Merced is a new microprocessor architecture, if I understand the present
   HP/Intel agreement correctly.  This information would tend to imply the
   PA-to-Merced migration has been delayed (again :-), and there's a .25
   micron part planned as a placeholder for the HP PA series to "fill in"
   for the as-yet-unavailable Merced part...

   As for the micron size, nearly every microprocessor vendor -- including
   DIGITAL -- tends to run a "shrink" into the new technology before running
   a new "native" design.  That's effectively what happened with EV4->EV45
   and with EV5->EV56.  I'd bet that's what HP did...

   And unlike HP, we don't have any architectural replacements in the works,
   either -- Merced (if I've got my Intel codeword straight -- this chip is
   sometimes refered to as P7) was to be a wholely new VLIW design, but the
   rumor mill has indicated HP/Intel has decided to use a "more traditional"
   superscaler design -- though it might still have a VLIW core underneath.

   (And DIGITAL has some key patents around superscaler designs -- the DS
   folks in Hudson are very familiar with the limits of the "traditional"
   superscaler design, too.  I'll not comment on futures here, tho'...)

   And unlike HP's situation, Alpha is (already) 64 bit.  (If I were HP,
   I'd definitely want to talk about .25 micron designs long before the
   upcoming Merced and 64-bit migrations...)

   Ask HP (and Intel) what sort of PA compatibility is planned for Merced...

   We've openly demonstrated Alpha systems running at 767 MHz a while back,
   and we have started to tip our hand on EV6.  (Though one would not know
   it by looking at our website.)

    .4 should I tell them that our 767 Mhz chip neede a $20 compressor to
    make it work?... 8-)
    
    

:    .4 should I tell them that our 767 Mhz chip neede a $20 compressor to
:    make it work?... 8-)

    A 767 MHz Alpha (and a $20 chip chiller) is still demonstratibly faster
    than any available Merced...  :-)

The chiller is $1000, not $20. The compressor is structurally similar to
a $20 refrigerator compressor, but it's rated for continuous operation,
and there's more to the chiller subsystem than just the compressor.
Maybe it'll come down a bit if it ever goes into volume production.

--RS

:The chiller is $1000, not $20.

    It's *still* faster...

    (This string reminds me of the "wart on the cheek" comment posited
    by one long-time DEC� manager -- we are an engineering company and
    have a long history of pointing out our warts in great detail, and
    then asking folks "is this ok?", which is an, uh, unusual marketing
    and communications approach.)

    We have shown fast(er) systems, we have faster chips, and Merced
    is (still) vapor...  
    
    --

    �DEC, not DIGITAL.  :-)