Conference 7.286::digital

    I saw this and just stared in amazement...  who the *)&$ are buying
    these chips?  As far as I've seen, there is a complete lack of IBM
    products and Motorola products.  Apple has the PowerMAC, but certainly
    haven't sold a million.  Ford may be putting some in cars, or maybe
    they haven't started doing that yet.
    
    IBM is the king of all FUD.  They must be including all of the
    development chips, prototype wafer lots, everything to account for a
    million.
    
    ..Larry

>>>    sold a million.  Ford may be putting some in cars, or maybe


	In a few  years out, Ford will be buying 5+ Million/year.

	Rick

	

In today's (6/27/94) Wall Street Journal, there's an artcicle on p B1 that says
"IBMis running months behind schedule and now plans a much-lower-profile launch
of its new PowerPC computers ... " 

Brian

    I though the 601 was manufactured by Motorola, but the article says
    "manufactured by IBM" ! What does Motorola do exactly ?
    
    Michel.

If you look at the article, the word "sold" was never used.
1 million shipped. 

How many were functioning? 

How many were just non-functioning samples to board vendors?

Marketing at its finest?

Gim

RE: .4

My friends at Moto told me that IBM is doing ALL the PowerPC fab.  Moto
transferred some process technology that they had been using for Mac �Ps and are
therefore out of the fab business for this �P.

A lot of these devices are probably being sold into embedded apps where people
just need a �P, & not the most technologically (i.e., $$$$) advanced device.

I too, read the WSJ article on IBM's inability to deploy PC's based on PowerPC. 
They're talking about merging the ``old'' PC family with the ``new'' PowerPC
family, and saying that it's just another option in the PC family.  Reminds me
of our PC ads that (IMHO) says, ``Oh, by the way, if you reeeally want, you
can have an Alpha AXP �P...''

Somewhere there is probably a warehouse bulging with these chips.
Or maybe there are a few tractor trailers circling the interstates
with a million chips. They've been shipped, just waiting for someone
to receive them... Maybe IBM should borrow DEC's slogan: We've got
it now (and we sure wish some of you would take them off our hands)!

RE:.7

Your analogy reminds me of the NYC garbage barges that circle the globe looking
for takers, only to re-dock in NYC... :-)

    I always wondered why at nearly 200 Mhz our performance wasn't much
    higher than the other vendors.  I recently read an article on IBM chips
    which explains the differences there:
    
    	In late 1993, IBM introduced its new Power2 architecture for the
    RS/6000.  Its a high-performance, superscalar CPU hat can issue up to
    'six' instructions per clock cycle running with a 71.5 Mhz clock. 
    Remember the AXP only does 'two'.
    
    	The 601, the first of the PowerPC line and the processor used in
    the PowerMac is a pipelined, superscalar RISC microarchitecture that
    can issue up to 'three' instructions per clock cycle.  The instructions
    can be dispatched to a branch unit, fixed-point unit (integer) or  a
    floating-point unit for execution.
    

    
 >       In late 1993, IBM introduced its new Power2 architecture for the
 >   RS/6000.  Its a high-performance, superscalar CPU hat can issue up to
 >   'six' instructions per clock cycle running with a 71.5 Mhz clock. 
 >   Remember the AXP only does 'two'.

There seems to be some confusion here.  Architectures don't issue
instructions, implementations do.   Did they build a six-issue
CPU?

Initial implementations of Alpha are dual-issue.  Others might
be, oh, quad-issue.

It's also highly significant just what instructions can be overlapped.
For example, if you can overlap integer and FP instructions, that
doesn't do your average Windows user much good, since Windows itself
doesn't use floating point (particular applications may of
course be number-crunchers).    I don't know what Alpha's overlap
capabilities are though.

Also, that Power2 CPU was implemented in (I think) 8 separate chips with
incredible tricks pulled to get the speed.  It's generally accepted in the
industry (and pretty much admitted by IBM) that this design is not scalable
and that they'll need to go to single chips in the future.

The other thing to remember regarding multi-issue is that in order for it to
be of any help at all, you need compilers that are smart enough to emit
the instructions in the optimal order AND you need a program that does just
the right mix of things to keep things going 6-way.  Not an easy task.

				Steve

    re .11 "compilers that are smart enough..."
    
    See the Communications of the ACM, Jun 94, V37, N6 "the making of the
    PowerPC".
    
    I'm no compiler expert, not an Alpha expert, nor a RISC expert, nor a
    marketing expert, but I 
    have a working "grasp" of the issues.  After skimming some of the
    articles in this issue, I was, frankly, discouraged.
    
    Motorola silicon, motorola compilers (buyout from ???) targettable to
    AIX, NT, Solaris, MAC, X86 (everything but Alpha), emulators that let
    you test the hardware before there is hardware, native debugger support
    for all these environments, IBM marketing muscle, Apple applications.
    
    I'm overwhelmed.  This consortium seems to have the bases covered so
    well that they can hardly fail at getting the press and momentum that
    they need to succeed.  DEC's Alpha program, is impressive, but is
    dwarfed in comparison.  How does DEC, alone, compete???????
    
    

    According to some discussion on Usenet [a very reliable source, as we
    all know] the number might well be close to reality. Apparently, Apple
    has already sold several hundred thousand PowerMacs, and probably have
    quite a few in the pipeline, so they might account alone for most of
    the 1,000,000 chips. (Of course, IBM might ship chips in July that
    Apple needs for Christmas...).
    

    Just another data point: the "Windows NT on Power PC" section of the
    Motorola setup on the PC Expo floor.  Showing NT running on Power PC's.
    I was on my way out, so I only skimmed: one vendor show Xwindows under
    NT on PPC, and a Taiwanese vendor who will be shipping their own clone
    Power PC's _with_ NT by the end of the summer (HW vendor-supplied
    release, or so he said).
    
    Even if IBM doesn't get off the fence on the NT/OS/2/WorkPlace OS/Power
    PC issue, the clone vendors won't wait.  The clear winner is Microsoft.

re .13 - Ok, so maybe those trucks are loaded with PowerMAC's
instead of just chips ;^)

End user delivery is the real issue. Are there really 1M users
firing up their Power machines within the next few months? I am
curious too about the ratio of embedded systems to "computer"
systems. Any info? How does this compare with other current
chip families?

    
    It doesn't matter who has the better mouse trap. It's who can sell
    enough mouse traps to stay in bussiness.
    
    How do we sell 2M AXP chips a year?  How do we get to 10M where the
    trends suggest IBM/Moto will be???
    
    Jon
    

    We need to get Alpha into thermostats, cars, refridgerators, traffic
    lights, video games, etc.  Computers won't cut it, especially if we're
    the only major player selling them.  Copiers, fax devices, etc. all
    require chips and we should be the supplier.
    
    ..Larry

    In addition, What about having Alpha chips in our storageworks and 
    Printer products, for starters.
    
    If Ford will use PowerPC in their cars, what about all the other
    car manufacturers, starting with our partner mitsubishi (sp ?) cars and
    BMW's, Mercedes, etc., etc. ?
    
      

>In addition, What about having Alpha chips in our storageworks
>and printer products for starters.

>If Ford will use PowerPC in their cars...

Ford puts computers in cars because they have a NEED for them.
Just what is an Alpha going to DO in a storageworks box??  Provide ballast?
Increase the count of "Alpha CPUs shipped"?  Increase the cost of the
storageworks box until it is not competetive?

    re .19
    
    I thought a storageworks box used a microprocessor. It if
    does not (and i have not even seen one), then delete the storageworks
    part of my argument. However, the rest stands.
                               

    The video servers that the storage guys are are shipping for Video on 
    Demand have AXPs in them.
    
    -Bruce  

    DEC seems to be going after the big ticket items.  IBM, however, is
    going after the embedded systems market.  IBM's philosoply is:  The
    success of the PowerPC in desktops will give its use in embedded
    applications an added push.  And conversely, IBM workstations and Apple 
    PowerMacs, can serve as software development platforms for embedded 
    applications.  Writing and debugging application code on a host
    development system, rather than having to cross-compile, download, and
    debug it to execute on a target CPU, gives the PowerPC a significant
    advantage.
    
    If the PowerPC is the embedded microprocessor of choice why would they
    purchase an AXP system for development?  Where is the cross-compilers? 
    One of the most significant items that contributed to VAXes success in
    the development environment was the abundance of cross-compilers
    available for different microprocessors.  I should also mentioned that
    the PDPs also had this robustness of cross-compilers.  When I was a
    chip designer in the 70's, didn't use VAXes, but our PDP had a
    cross-compiler for every microprocessor on the market e.g., Intel,
    Motorola, National, Mostek, etc.
    
    The failure of VAXBI in the embedded market was the reluctance of DEC
    to sell 78032 (MicroVAX chips).  We tried to convience developers to
    buy DEC designer VAXBI modules and card cages; have them design with
    the 68000 or other microprocessor of choice and use a DRB32 to link the 
    VAXBI card cage to a VME environment.  When XEROX heard this they 
    dropped DEC and moved to the SPARC environment for their next generation 
    of copiers fast than we could believe.  
    
    In 1992, 25+ Million 32-bit microprocessors shipped in embedded
    applications:
    	- Over 20 Million were 68000-based
    	- 4 Million dedicated to RISC-type controller made of i960, ARM,
    	  29000, and Hobbit)
     	- 200,000 workstation CPUs made up of SPARC, MIPS, PowerPC, and AXP
    	- 700,000 X86-type products
    
    While most RISC processor makers agree that it doesn't make sense to
    use a 100 Mhz processor for cost-sensitive applications, there are
    embedded applications designs that are making use of high-Mips
    processors.
    	
    Handling ATM protocols with embedded CPUs and next generation laser
    printers.  HP's Laserjet series uses the i960 and QMS 32 ppm uses a
    R3081 RISC core.  Oki Data Systems and Raster Ops use a MIPS core
    processor in some of their current printers.  
    
    You can't use an AXP for software development if a cross-compiler isn't
    available and independent software developers are no longer developing
    cross-compilers for differing architectures.

    re: back a few, embedded chips:
    
    Japanese cars seem to have their entire technical spec written on the
    boot for you to read. 5 speed, 16 valve, DOHC, turbo, ABS and plenty
    more alphabet spaghetti.
    
    If Ford and others are serious about putting processors in cars to
    provide more and more complex functions, when will the processor spec
    start to appear on the boot?  When will you buy a 64-bit 128MB 16Valve
    GTI?
    
    Looking at the games console market, number of bits is a vital
    marketing play.  Look at the 32-bit Commodore CD32, and the (allegedly)
    64 bit Jaguar.
    
    Maybe we should put 64-bit splash logos on our Alpha PC's!
    
    Martin
    

    
    Why use a jackhammer when a feather will do.  The jobs that cars
    CPU's work on aren't that demanding.  Most of these jobs can be
    done with 8-16 bit, slow, low-power computers.
    
    Where would you put a 80W Alpha computer in embedded control??
    
    Jon
    

    re .24
    
    So what's Ford going to use the PowerPC for ? Perhaps, future
    generation of cars will need much more processing power than today's
    models. Also I am not saying use current Alpha Chips. I thought we had in
    the pipeline low-cost, low-power Alpha chips for such applications.
    
    Phil 

    Ford currently uses the 80100 and 80110 chips from Motorola in their
    cars.  This is also the same chip set found in the Data General Aviion
    computers.  They are the only two customers left for this chip that is
    being phased out.  Ford has already announced they are going to the
    PowerPC chipset.
    
    Interesting side note:  DG must go through a chip change soon.  They
    can go to the PowerPC, MIPS or Alpha.  MIPS is doubtful as SGI seems to
    keep this one close at hand for computers and I've heard rumor that we
    turned down DG for using Alpha (unsure of circumstances).  In any case,
    look for DG to announce their adoption of the PowerPC for their new
    computers.
    
    ..Larry

why would we turn down DG fro the use of Alpha? Alpha is for sale to 
anyone who wants it. Actually it will be regarded as discrimination 
since we are selling the chips to the general public.

    
    > Why use a jackhammer when a feather will do.  The jobs that cars
    > CPU's work on aren't that demanding.  Most of these jobs can be
    > done with 8-16 bit, slow, low-power computers.
    
    Maybe for engine management.  Ford is working on a night vision
    enhancement system that would require a bit more poke.
    

    We couldn't even line up Intergraph
      who's Clipper processor ran out of gas last year
      who's CAD software was a good match for Alpha's floating point
      and who's operating system of choice is NT
    for Alpha.  It doesn't get any nicer of a match than that, yet
    (to quote Gordon Bell) "we blew it".  There's only one or two
    (DG, maybe Sun) chances left for Alpha.  kb

     Well Intergraph did port to Alpha/NT.
    
    -Bruce
    
    
     Alpha NT Developers' ...                                 U.S.  DESKTOP
     --------------------------------------------------------------------------
     Alpha AXP/Windows NT Application Census                  Date: 24-May-1994
     --------------------------------------------------------------------------
     Intergraph - MicroStation                                       SHIPPING
    

    yeah, I know Intergraph is porting Microstation to AXP...
    but instead of replacing their Clipper hardware with Alpha, they
    went with Intel.  Their dual 90-Mhz Pentium workstation did beat
    out Digital and land them a $400m Navy contract, so I guess they
    made the right decision.
    -kb
    

   > We need to get Alpha into thermostats, cars, refridgerators...
    
    to keep these 80W babies cool? Talk about a symbiotic relationship...
    
   

    >>How could a DG or SUN switch to Alpha?  SUN hasn't declared their   
    >>architecture dead -- using Alpha would put egg on their face.  DG   
    >>seems to be in a similar space with Digital (workstations -> servers).   
    >>What would differentiate their hardware?  If they were running on   
    >>Alphas, would they compete or collaborate in the OS/software space?
    
    
    SUN could just as well use Alpha as they are going to be selling
    software for the PowerPC.  We used MIPS for some time, what difference
    does it make?  DG has to change, they have no choice if they want to
    remain in the systems business.  They just began reselling Indy from
    SGI, so maybe it will be a MIPS chip instead of PowerPC.  Their
    hardware would be differentiated by their software.  They would most
    likely have their own operating system, layered sw, ISV apps, etc.
    (like Cray) or maybe they will bet on NT.  We have a few doing that
    now:  Nekotech, Alpine, Carerra, etc.  They may or may not use our
    casings and they rarely use our memory, disks, etc.  They merely
    license the chip.  Alpha was made to run any operating system.  This is
    open systems for the latter half of the 1990s.
    
    ..Larry

Back to the original topic, I read in MacWeek that about 750,000 of those
PowerPCs have shipped to Apple.  Apple has said all along they would sell
a million PowerMacs in the first year.  I believe it, I'm typing this on one
of them.  It's an impressive machine.  Anyone familiar with the semiconductor
business knows you sometimes have to buy your parts well in advance to insure
a steady low cost supply.

- Peter

    November 7, 1994 issue of EE Times seems to argue that, once again, we
    should probably not believe all that we read ...  Here are a few
    excerpts from "PowerPC alliance wilts under burden" starting on page
    60:
    
    
    "... Signs of weakness have far outweighed signs of strength during
    what has proven to be a fitful start for the PowerPC and its
    partnerships.  And with a defacto agreement to disagree in place at the
    embedded end, the past year at times made some question whether the
    alliance was beyond repair."
    
    ...
    
    "Hopes for a common embedded application binary interface (API) that
    would allow development-tool makers to create one set of tools to work
    with both forthcoming IBM and Motorola embedded parts are quashed over
    register assignments and other implementation differences.  IBM and
    SunSoft go on to develop a Unix SVR4-based ABI, while Motorola and Ford
    Motor Co. continue to develop their 'deeply embedded' ABI."
    
    ...
    
    "In June, IBM announces it has shipped a million PowerPC 601 chips, the
    bridge chip that implements some legacy Power-architecture
    instructions.  But number-crunching pundits suggest that the company is
    stockpiling the chips somewhere because nowhere near a million systems
    have been shipped."
    
    ...
    
    "Privately, and after weeks of rumors that Apple is upset that IBM has
    made the company's allegiance to the PowerPC Reference Platform
    specification for desktop and server systems a fait accompli [an
    accomplished fact], an AppleSoft vice president confirms the first
    cracks in the PowerPC alliance.  Guerrino DeLuca says at Macworld
    Boston that the company has defined its own PowerPC reference platform
    based on the Macintosh architecture and is ready to license key
    software and hardware elements (including system ASICs) should
    negotiations with IBM over a common platform break down."
    
    "Then in September, though the company says it is still discussing the
    possibility of a common PowerPC reference platform with IBM, Apple
    announces the reference platform and licensing strategy, calling it the
    Macintosh RISC Architecture.  Though it won't name its licensees, Apple
    is reported to be lobbying Toshiba, Fujitsu, Olivetti and other major
    non-U.S. system houses."
    
    "The same month sees Digital Equipment Corp. taking a heavy loss when
    it sells its minority stake in longtime business ally Ing. C. Olivetti
    & Co.  The move fuels rumors that Olivetti has decided to jump to the
    PowerPC for its high-end line, currently based on Digital's Alpha AXP
    processor, and sales of which are a condition of the equity stake."
    
    ...
    
    "All in all, the pending deal is also poised to revive the alliance's
    original promise.  Without such an agreement, the Intel-Microsoft
    hegemony has little to fear."
    
    
    There's more.  Some negative, some positive.  IBM, Apple and Motorola 
    don't appear to yet have their act together, by my interpretation of 
    the article.
    
    Steve