Conference 7.286::digital

fast is nice, but what about price/performance?  where do we stand on that?

I remember years ago there was this company called CRAY who blew the doors off
everyone in performance, but where are they now?

-mark

    
    And those pages currently don't include the AlphaStation 500/500
    announced last week.
    
    SPECint95 = 15.0, SPECfp95 = 19.5, which makes it the
    fastest desktop workstation in the world... period.
    
|fast is nice, but what about price/performance?  where do we stand on that?
    
    We must be pretty bad, huh?  SPEC doesn't include prices of system
    under test for CPU95 results.  That's why Ideas-International doesn't
    include the price.  But you're right, we have a lower price/performance.
    
    Oops, lower is better, isn't it.
    
    								-mr. bill

    Set your web pointers to   http://www.onsale.com
    and you can bid on some Alpha-based PC's being auctioned off...
    going price is about $700.  Pretty good SPEC numbers too.
    Kratz

>    And those pages currently don't include the AlphaStation 500/500
>    announced last week.

Same old question - how about some part numbers? Nice to talk about hot boxes
but all the numbers in the world don't matter a bit if you can't order one, let
alone get a delivery date.

How about some SPEC numbers for the ALphaServer 8400 5/440 - a machine we can 
quote today. Why no performance numbers on that pup?

tim m.

    The AlphaStation 500/500 folks are between a rock and a hard place.
    People like me and .-1 want part numbers. But the rest of the industry
    are all having great fun playing pre-announcement games. The
    AlphaStation 500 folks have taken a slight liberty of attempting to
    gain some publicity a little while before the stuff is orderable. I'd
    love part numbers to be available, but I'd also like to thank the
    AlphaStation folks for playing a bit of the same game that everyone
    else does. 
    
    regards
    john

    ... and I thought OWNING 50% of SPECint and SPECfp top
    20 was a good thing, something a DIGITAL employee could
    be proud of. 

>    ... and I thought OWNING 50% of SPECint and SPECfp top
>    20 was a good thing, something a DIGITAL employee could
>    be proud of. 

It *is* a good thing, but does it put meat on the table?  I'd settle for a
smaller % of something more tangible, like the server market, desktop, etc...

-mark

    re: Note 4770.4 by DECIDE::MOFFITT
    
|   Same old question - how about some part numbers?
    
    The already shipping hot box is the AlphaStation 500/400.  PB56*
    are the part numbers for those.  

|   Nice to talk about hot boxes but all the numbers in the world don't
|   matter a bit if you can't order one, let alone get a delivery date.
    
    The press release said the AlphaStation 500/500 will be available in
    September.   That's just a few weeks away.  You can probably SWAG what
    the part numbers will be.  (The 8MB cache will be available Q4CY96.)
    If you are just curious about the part numbers, please be patient.
    But if you have an immediate need to know, there are better ways to
    find out than here in the Digital conference.
    
|How about some SPEC numbers for the ALphaServer 8400 5/440 - a machine we can 
|quote today. Why no performance numbers on that pup?
    
    There *ARE* SPEC CPU95 performance numbers for that pup.  Again, there
    are better ways to find out than here in the Digital conference.
    If you are just curious, please be patient.
    
    -----
    
    re: Note 4770.5 by BBPBV1::WALLACE "Unix is digital. Use Digital UNIX."
    
|   I'd love part numbers to be available, but I'd also like to thank the
|   AlphaStation folks for playing a bit of the same game that everyone
|   else does.                                                        
    
    Not just playing, playing *AND* winning.  (Making a splash at
    SIGGRAPH'96 just a couple of weeks before all the i's were crossed
    and t's dotted was absolutely the right thing to do.)
    
    So how many of your customers are ready to switch to Digital?
    
    								-mr. bill

         <<< WRKSYS::AXP_TOOLS:[NOTES$LIBRARY]ALPHASTATION500.NOTE;1 >>>
                              -< AlphaStation500 >-
================================================================================
Note 79.6      Highest performance uni-processor machine on earth         6 of 6
BLOFLY::SMITHP "Beware the knights who say "NT"..." 182 lines  13-AUG-1996 04:17
            -< AS500 and PowerStorm announcement with ViewPerf... >-
--------------------------------------------------------------------------------
    Here's the announcement from our geography, which includes Viewperf
    figures for the new 4D40T/50T/60T cards (were these achieved with the
    AS500/500 ??).
    
    We now have a quantitative comparison with SGI Impact series !
    
    
    Cheers, 
    
    Peter.
    
    
IDEAS INTERNATIONAL COMPETITIVE ALERT IDEAS INTERNATIONAL COMPETITIVE 
IDEAS INTERNATIONAL COMPETITIVE ALERT IDEAS INTERNATIONAL COMPETITIVE 
IDEAS INTERNATIONAL COMPETITIVE ALERT IDEAS INTERNATIONAL COMPETITIVE 

                          8th August 1996

         Digital announce new High-end Workstation and 
                More Powerful PowerStorm graphics

*********************************************************************

A few months back Hewlett-Packard announced new PA-8000 workstations 
which wrested the workstation single processor performance crown from 
Digital Equipment's AlphaStation range. Now it appears Digital has 
reacted to HP with its own announcement. Digital has introduced a new 
high-performance AlphaStation 500/500. The desktop workstation uses a 
500MHz Alpha DEChip 21164 processor and 8MB of secondary cache and 
has posted SPEC95 figures which manage to eclipse that of HP's high-
end PA-8000 workstation, the model C180-XP. (see table below)

-------------------------------------------------------------
     Competitive Single Processor SPECint95 and SPECfp95
-------------------------------------------------------------
Company Model                  fp95  int95 Processor     MHz
-------------------------------------------------------------
Digital AlphaStation 500/500   19.5  15.0  Alpha 21164   500
HP      HP 9000 C180-XP        18.7  11.8  PA-8000       180
HP      HP 9000 C160           16.3  10.4  PA-8000       160
Digital AlphaStation 500/400   14.1  12.0  Alpha 21164   400
Digital AlphaStation 500/333   12.5   9.8  Alpha 21164   333
SGI     Indigo2                12.5   8.9  R10000        200
Sun     Ultra 2 Model 1200     11.4   7.7  UltraSPARC-I  200
-------------------------------------------------------------
Note: table is ordered by descending SPECfp95

The AlphaStation 500/500 forms a new high-end to the AlphaStation 500 
models, and as such shares much of the attributes of other 
AlphaStation 500 workstations.

The system comes standard with 2MB of secondary level cache, as found 
in the AlphaStation 500/400. However, to achive even higher levels of 
performance, the AlphaStation 500/500 also has an 8MB cache option 
available.

The AlphaStation 500/500 with 2MB of cache, 128MB of memory, 2GB disk 
drive, 21" monitor, PowerStorm 4D60T, 4MB of texture memory, floppy, 
CD-ROM, keyboard, mouse and Windows NT Workstation licence is 
$US54,660.

The AlphaStation 500/500 is reported by Digital to be available from 
September 1996, with the 8MB cache option coming sometime in the 
fourth quarter of 1996.

New PowerStorm Graphics
-----------------------
In March of this year Digital announced a new family of graphics 
known as PowerStorm. Digital has recently announced additional 
adapters for the PowerStorm line.

Of the three adapters announced, the 4D40T, 4D50T and the high-end 
4D60T, the 4D60T was announced in March, but it was not expected to 
ship until the middle of 1996, and little performance or pricing 
information was available at the time.

The three adapters come into the high-end of the PowerStorm graphics 
adapter range. The table below outlines the major features of the 
three adapters.

---------------------------------------------------------------
                         4D40T           4D50T            4D60T
---------------------------------------------------------------
Price ($US)             $3,995          $7,995          $11,995
Resolution (up to)   1280x1024       1280x1024        1600x1280
Color planes         24-bit DB       24-bit DB        24-bit DB
Overlay planes           4-bit           4-bit            8-bit
Z-buffer planes         24-bit          24-bit           32-bit
ViewPerf CDRS-03         22.40           41.90            46.51
ViewPerf DX-03            6.54            9.07            11.15
ViewPerf DRV-02           3.67            4.86             5.22
---------------------------------------------------------------

----------------------
Texture Memory Pricing
----------------------
4MB           $US595
16MB        $US2,495
32MB        $US4,995
----------------------

The PowerStorm 4D60T graphics adapter offers similar levels of 
performance to Silicon Graphics' Maximum IMPACT graphics option found 
on SGI's Indigo2 workstation. The table below compares the 
performance of PowerStorm 4D60T vs Indigo2 Maximum IMPACT graphics.

------------------------------------------
Benchmark           PowerStorm    Maximum
                       4D60T       IMPACT
------------------------------------------
ViewPerf CDRS-03       46.51       47.67
ViewPerf DX-03         11.15        9.54
ViewPerf DRV-02         5.22        6.72
------------------------------------------

Digital did not disclose what workstation was used in conjunction 
with the PowerStorm adapter to post this level of performance. 
However, it is likely to be the new workstation with a 500MHz Alpha 
processor, employing 8MB of secondary level cache. Silicon Graphics 
has used an Indigo2 with an R10000 200MHz processor. The exception 
being CDRS-03 47.67. This was achieved with a 250MHz R4400 processor.

The Viewperf series of benchmarks are a portable OpenGL performance 
benchmark programs written in C. It was developed by IBM. Later 
updates and significant contributions were made by SGI, Digital and 
other OPC project group members. Currently, the program runs on most 
implementations of UNIX, Windows NT, Windows95, and OS/2. The 
following is a brief description of the three Viweperf tests featured 
in the table above.

CDRS-03 -  contains seven different Viewperf tests, is a modeling and 
rendering application for computer-aided industrial design. 

DX-03 -  has 10 different tests, is a visualization application. 

DRV-02 - has 10 different tests, is a 3D computer model review 
package. 

Availability
------------

Digital has announced that the three graphics adapters will be 
available September 1996.

******************************************************************

IDEAS INTERNATIONAL COMPETITIVE ALERT IDEAS INTERNATIONAL COMPETITIVE 
IDEAS INTERNATIONAL COMPETITIVE ALERT IDEAS INTERNATIONAL COMPETITIVE 
IDEAS INTERNATIONAL COMPETITIVE ALERT IDEAS INTERNATIONAL COMPETITIVE 

################################################################
#  The attached information is provided by IDEAS International #
#  as part of their industry research services and is intended #
#  for the internal use only of the recipient company.         #
#                                                              #
#  Copyright remains with IDEAS International and further      #
#  distribution/forwarding/modification outside of the scope   #
#  of your agreement with IDEAS International is prohibited.   #
#                                                              #
#  The information contained in this document has been         #
#  researched and compiled from sources believed to be         #
#  reliable at the time of writing, but is not warranted as    #
#  being correct in any way by IDEAS International Pty. Ltd    #
################################################################


    -------------------------------------------------------- 
    Phone:  +612 9482 8900            Fax:    +612 9482 8877
    [email protected]   http://www.ideasinternational.com 
    -------------------------------------------------------- 
    I  D  E  A  S                    Suite 18               
    =============                    14 Edgeworth David Ave 
    International                    Hornsby NSW  Australia 
    -------------------------------------------------------- 
    *             Computer Industry Analysts               * 
    *        Competitive Profiles System Selection         * 
    --------------------------------------------------------




    
    

    
    AlphaStation 500 no longer fastest desktop workstation....
    
    Now it's the fastest uni-processor on the planet.
    
    

    This doesn't look good.  From a customer's prespective Digital had to go
    to 500 Mhz to beat a 180 Mhz.  Digital doesn't have much head room but HP 
    seems to have lots of room for growth to 500 Mhz.  I can already
    anticipate the questions at the next quarterly Business Partner Update.
    The response will be - UM???? cuz I won't get any appropriate messages
    from Corporate.

|    This doesn't look good.  From a customer's prespective Digital had to go
|    to 500 Mhz to beat a 180 Mhz.

	This is like comparing Gas RPMs against Diesel RPMs. Anyone who
        compares MHz accross computing platforms, or even CPU generations,
	for the purpose of determining speed, is nuts. Treat MHz as part of
	the model number, nothing else.

	Compare the machines using industry standard benchmarks, for example.
	If HP can run "X" clonks per clink with glow plugs and we can do a
	bit better, for less money, who cares about the MHz.

	The only thing that would possibly create a question in people's mind
	is if 500MHz is somehow near some theoretical limit and that HP, for
	that reason, can easily grow from 180 to 500 using proven technology
	while we have to break serious new ground. I am not a HW guy, but I
	suspect that the difficulties of doubling the 180 versus our 500 may
	have little to do with the clock itself.

	>Per

re: Note 4770.11 by tennis.ivo.dec.com::TENNIS::KAM 
    
    Have your customer ask HP how many watts a PA-8000 would run at if they
    could get it to work at 500Mhz.
    
    Don't assume 500Mhz is a limit for EV56.  
    
    -Bruce

>       This is like comparing Gas RPMs against Diesel RPMs. Anyone who

  Could be but If I make up a chart that looks like this, obviously ours has
  to work harder and theirs has more of a future.


  CPU      |    MFM*     |      Clock
  =========:=============:===============
  Ours     |    100      |      500 Mhz
  ---------+-------------+---------------
  Theirs   |    100      |      180 Mhz
  ---------+-------------+---------------
  * Meaningless Figure of Merit or "industry standard benchmark" depending on
  how you look at it.

  	mikeP

More like ignorance.

If HP (or Intel, or Sun) could run their CPUs at 500MHz, don't you think they 
would? "Yes, we can run at 500MHz, but we're artificially crippling our product 
line because we enjoy the competition". And if they did suddenly multiply their 
clock speed by 3, would their bus, memory, and disks do the same?

The simple explanation (and it's *very* simple) is that Alpha has chosen to do 
simple instructions quickly, whereas others have chosen to do more complicated 
instructions more slowly (speed demon vs brainiac, I believe are the cute 
names).

Because Alpha's instructions are simple, they can, and do, go faster. But 
because they are simpler, they can do less.

Because PA-RISC's (x86's, etc) instructions are more complicated, they can do 
more, but they can't be executed as quickly.

When it all comes out in the wash, different trade-offs end up giving roughly 
the same performance, with Alpha arguably coming out ahead.

Other things also apply. For instance, PA-RISC does instruction reordering, 
whereas Alpha doesn't. If Alpha did, it might go faster for no increase in clock 
speed.

In summary, our's *isn't* working harder, it's working differently. Their's 
doesn't have more (or less) of a future, because it's working differently.

As .12 says, comparing MHz between architectures is nuts.

If you want better explanations, go to the DEChips conference.

PJDM

    re:.11  Who said anything about MHz?  
    	    Spec fp and int DOMINATION amongst uni's.
    	    

    re: our 500 MHz vs. their 180 MHz, and "working harder"
    
    Just tell the customer that our machine is a RISC (Reduced Interval Slice
    Computer) -- that We've got smaller hertz than Them, so our hertz run
    faster.  Gee, they bought the reduced instruction set thing, didn't
    they?   :-)  :-)
    
    -Mark
    
    P.S.  Be sure to spell it out -- so they don't think you said our
    computer is a risk!

I'm a software guy.  I don't really know much about the details presented in 
the last few replies.  I don't know who's right or wrong.  I hate to say it 
but I don't think I care.

But what did mean something to me was .14.  Here it is again:


>
>  CPU      |    MFM*     |      Clock
>  =========:=============:===============
>  Ours     |    100      |      500 Mhz
>  ---------+-------------+---------------
>  Theirs   |    100      |      180 Mhz
>  ---------+-------------+---------------
>  * Meaningless Figure of Merit or "industry standard benchmark" depending on
>  how you look at it.


Real simple, and very clear that Thiers is better because it does the same
work with less 'something'(MHZ in this case).

Right or wrong doesn't matter.  Its how the customers percieve it.

Competition will run ads like this and sell boxes.  Lots of them.  Much more 
than us.

We will hit them with lots of techie talk.  Probably all correct, no 
lies/deception.  We will sell some boxes to those customers that have techies 
writing the POs.  Big deal.

That table from .14 is called marketing.  As I have said before, we don't do
it.  Never have.  Our competition does.  We may have the superior products
but if the competition has better marketing (any marketing) they'll win the
sales.  I can only hope our marketing people don't think like the folks 
authoring the previous replies.  Not that I think the previous replies are 
good/bad/right/wrong.


Brad

Yes, 500MHz is the processor internal clock speed (i.e. dependent add
instructions can issue at 2ns intervals, and the theoretical peak
for the "perfect" instruction mix is ~2BIPS.)

Re several:

  Fred, we actually clock the Alpha chip with a clock signal
  that's substantially *SLOWER* than the internal on-chip
  clocks. When we say "500 MHz", that's the internal chip
  speed. This is pretty consistent across the entire industry.

  (Most external clocks are 33 MHz or some multiple. This is
  handy because it cycles the PCI bus at its top speed. Then,
  a phase-locked-loop on-chip then synchronously multiplies
  the clock frequency. The actual multiplication factor for
  Alpha is set by SROM code (firmware).)

 -=-=-=-=-=-=-

  With regard to slower-but-bigger Hertz (e.g., PA RISC) versus
  faster-but-smaller Hertz (e.g., Alpha), this is often referred
  to as "slow tick" versus "fast tick" architectures and yes, it's
  been discussed extensively and heatedly in RICKS::DECHIPS. If
  you don't already have this conference in your notebook and
  would like to add it, press <KP7> or <Select> or type "SELECT"
  and the conference will be added to your notebook.

                                   Atlant

    Gosh, aren't we good at spreading the SPECfoo95/MHz FUD!
    
    It's about as meaningful as SPECfp95/bit (we have been shipping
    64-bit UNIX, our compeition is talking about "outlining" their
    64-bit strategy.  Since we've got twice the bits, clearly we are
    losing the SPECfp95/bit race!
    
    It's like the people who claim the Pentium PRO 166 is the "slowest"
    Pentium PRO.  No, it's not.  The Pentium PRO 166 with 512KB L2 cache
    can be *faster* than the Pentium PRO 200 with 256K L2 cache.  (See
    the large number of TPC results published for the 166MHz/512KB and
    the lack of results for 200MHz/256KB.)  But of course, the P6/166
    loses at SPECint95/MB-cache metric, another absolutely meaningless
    metric.
    
    
    Many technical customers accept that SPECint95 and SPECfp95 are metrics
    that mean something.
    
    It should be *trivial* to explain that SPECint95/MHz and SPECfp95/MHz
    means *nothing* to these customers.  (If you really need a marketeer
    to help you explain that what are you doing in this business?)  It
    should be *trivial* to explain that only a marketeer could attempt
    to convince a techincal customer to use such a meaningless metric
    as SPECfp95/MHz as a "selling" point.
    
    
    But a simple analogy for here.
    
    A car race.  All the rules are laid out, and whoever finishes
    the race first wins.
    
    But after losing the race, the 2nd place finisher says "but see,
    I really won the race based on the speed/cylinder, since the winner
    was driving a V12 and I was driving a V8."  Implicit assumption -
    next year they'll be driving a V12.
    
    
    Finally.
    
    I've got news for you.  Talk with some people who work for our
    competitors.  The news that we are sooooo damn fast at SPECint95
    and SPECfp95 is admired by them, but also greated with groans
    at HP, Sun, IBM and SGI (and even Intel).
    
    Who would have thought that such great news could be answered with
    groans at Digital?
    
    								-mr. bill

    The 166/512 is the slowest PPro for servers.  The desktop-based 150/256
    isn't being made anymore anyhow.
    
    The TPC/C NT numbers on the 4100 (with the fastest Alpha - 400Mhz -
    Digital currently offers in a server) couldn't beat the 166 PPro
    based servers, so they don't get published, eh?.  And yet you
    complain that there's no 200 PPro server numbers?
    K 

    re .22 What does it take?  MARKETING and it been worn out in this
    notesfile over and over again.
    
    I suggest that you go back and re-read .18.  PERCEPTION IS EVERYTHING
    and it looks like DEC has to work 2.77 times harder than an HP system
    e.g, 500MHz vs. 180MHz.
    
    Every quarter when we meet with the Business Partner's and display a
    chart showing 500 vs. 180 and performance is marginally different it
    starts all over again.  And we're dealing with individuals that were
    mostly engineers or have engineering backgrounds.  The Competition
    might be admiring DEC back in their facility but on the street they're
    eating out lunch with FUD.
    
    The word is: what engineering efforts does it take to design, develop,
    and deliver a system at 500MHz vs. 200MHz - A LOT more effort.  Things
    are more complicated at 500MHz.  When HP is delivering systems at
    500MHz DEC will be at 1000GHz.  Therefore, if your microwave breaks
    just use your Personal Workstation until it's repaired.

    What it really comes down to is speed per dollar since most don't buy
    the fastest that is made, but only the fastest they can afford.  It's
    rare to see prices in performance studies, or to see the systems 
    configured the same.  I saw one performance study Digital did that had
    our machine configured with two ethernet cards to the competition's one
    ethernet card - in a web-server test.  It's getting very confusing
    trying to figure out if one should suggest a dual Pentium Pro 200 or an
    AlphaServer 400 or 1000. 
    
    bill

    re: PERCEPTION IS EVERYTHING
    
    The last similar run around was the panic about "what if HP could
    run a PA-7200 at >>140MHz, if they matched our >200MHz we'd be in
    real trouble!"  (Back then it was SPECfoo92/MHz that was the
    meaningless metric.)
    
    Recent history lesson.
    
    HP announced the PA-7200 chip at 140MHz.
    They shipped their first PA-7200 system at 100MHz.
    They will ship their last PA-7200 system at 120MHz.
    
    Makes the old worry about the PA-7200 running at >200MHz seem kind of
    silly, doesn't it?
    
    What makes you believe that the PA-8000 will be any different this time?
    
    (If you have credible information that the PA-8000 is going to be going
    out anytime soon at greater than 180MHz, let alone much much greater
    than 180MHz, please share.  Please include a source for the cache that
    goes with it.  Otherwise, I'm more inclined to assume that HP is having
    enough work right now with rumored thermal issues with the 180MHz
    PA-8000, it's cache, and the Visualize-48XP graphics in
    a desktop box.)
    
    
    								-mr. bill

Re .24
>    The word is: what engineering efforts does it take to design, develop,
>    and deliver a system at 500MHz vs. 200MHz - A LOT more effort. 

Nonsense.  Nothing outside the microprocessor (and the little device that
wiggles its clock inputs) sees anything approaching 500 MHz.  The cache loop
runs at something closer to 60 MHz (that is, PC speeds). (I'm sure some HW
folks can supply the exact rate).  The memory bus even slower.

Do you have any idea of the hoops HP jumps through to run its processor at
180 MHz?  (Hint -- there's no on-chip cache.)

Perhaps we should quote SPECfp95 per package pin -- HP needs a lot more pins
to get the same level of performance.  This is a metric that actually relates
(marginally) to a real cost.

Of course, price and availability are what really matter to the customer.
Pundits have been saying "Alpha can't maintain its lead" -- but it has.

Has anybody, over the last few years, kept track of CPU announced release dates 
and performances, and actual release dates and performances?

It might make an interesting chart.

PJDM

And how about end-user price and price/perf to go with it...

>Right or wrong doesn't matter.  Its how the customers percieve it.

That was my point Brad.  Thanks.  Thanks too for:

>That table from .14 is called marketing.  As I have said before, we don't do

Maybe I've found a new line of work.  Nah.  Probably not. 

	mikeP

    OK it's like this
    
    It sits on a desktop
    It occupies a space slightly bigger than a briefcase
    It is faster than the refrigerator-sized beasties from the competitors
    
    You say you can't sell it, because the chip is 500 MHz vs. their
    nearly-as-fast 200 MHz system.
    
    I say fine, please go sell shoes.  
    Please don't try to sell Alphas.
    
    You say only a technoweenie would understand that its the fastest.
    
    Fine.  Sign me
    	/technoweenie and proud of it

    
    most will buy on emotion NOT logic, therfore .14/.18/.26/.30 hit the nail
    on the head it's the perception.../marketing that has always won
    in the end...
    
    why do you think microsoft/intel are number 1 in there field...
    
    marketing/perception...
    
    others may have a better product but WITHOUT perception/marketing
    it doesn't mean squat to the people who buy cause 90% buy based on
    emotion NOT logic...
    
    IMO

    perceptions are real
    
    but so are facts
    
    One might almost gather that the naysayers in this stream would prefer
    not to have to bother with the inconvenient fact that Alpha is fastest.

My experience is that many customers never get past megahertz, ie. DEC is 
delivering 500 MHz (best) while others are at only 200 MHz or less.  I don't 
go into the other details unless I see a glimmer of comprehension in their 
eyes.  Otherwise we do our best to go straight to discussions on applications 
and benefits.

If they've taken the time to listen to HP's fud regarding microprocessor clock 
rates they'll spend a few minutes listening to the other side of the story.

I appreciate this string and others on the topic here and in DECCHIPS because 
it prepares me with answers to their questions.

D.

    The analogy about 6cyl vs 12 cyl:-
                                                 
    Haven't we started to qualify *our* claims with the addition of
    *uniprocessor* ?  How is this different then?
    
    If our competitors can approach our performance by using SMP, then we
    should not be talking about unfair comparison.
    
    - Vikas

or 'Why engineers shouldn't be marketeers'


RE .33

   
>    One might almost gather that the naysayers in this stream would prefer
>    not to have to bother with the inconvenient fact that Alpha is fastest.

Perhaps I'm one of the naysayers?  I don't really mean to be.  Anyways, 
the fact that ALPHA is fastest/best/pick_your_superlative is great, but
I think companies like Intel and Microsoft have proven that you don't 
need to have the best product to take over a market.  Although it does 
make it easier, alone it gets you almost nowhere.

We may have the best chip and the best boxes but if the customers are buying 
HPs then it just doesn't matter. We need to focus on _convincing_ the 
customers that our products are the best regardless of whether they are or 
not.  I know it doesn't sound good but that is business, IMO. 
  

> One might almost gather that the naysayers in this stream would prefer
> not to have to bother with the inconvenient fact that Alpha is fastest.

  One of {Ferrari|Lamborghini|Masserati} are probably the
  fastest, too. But I'd rather own the {Honda}Toyota|Nissan|
  Volkswagenwerk} corporation than any of them.

  Fast is fine. But cheap and easy to understand sells.

                                   Atlant

    I use Windows 95 on my desktops because is does what I need and costs
    $90. Is Windows NT better and OpenVMS the best? Yes, but what do THEY
    cost?
    
    Ken....

re Note 4770.36 by PATRLR::MCCUSKER:

> I think companies like Intel and Microsoft have proven that you don't 
> need to have the best product to take over a market.  

        Alternatively, the success of Intel and Microsoft may show
        that our conventional notions of what constitutes "best" are
        wrong.

        Bob

    
    Best for what?
    
    Personal productivity tools?  Get a Mac.
    
    Personal productivity tools, wide selection of of system vendors, and
    wide selection of offbeat/free/hobbiest software?  Get anything that
    runs Windows 3.1/DOS/Windows 95.
    
    Almost everything above in a more expensive package, with some long
    term potential?  Windows NT
    
    A traditional multiuser production environment?  Get VMS.
    
    A 3D workstation?  Get UNIX on SGI ;-)
    
    

    re: .32
    
    >>    why do you think microsoft/intel are number 1 in there field...
    
    >>    marketing/perception...
    
    Well, that and a ***TREMENDOUS*** amount of luck.  Where would 
    Microsoft be today if Gary Killdall (sp?) hadn't blown off IBM when
    they wanted to license CPM-86 for their new PC.  What IBM had
    decided to go with a chip other than the 8088?  Both companies have
    done a good job in capitalizing on their good fortune, but the
    initial good fortune was not created by virtue of their marketing 
    acumen.
    
    That doesn't mean Digital doesn't need to vastly improve its
    marketing ability.  
    
    
    My 2 cents,
    Jim

> Well, that and a ***TREMENDOUS*** amount of luck.  Where would 
> Microsoft be today if Gary Killdall (sp?) hadn't blown off IBM when
> they wanted to license CPM-86 for their new PC.  What IBM had
> decided to go with a chip other than the 8088?  Both companies have
> done a good job in capitalizing on their good fortune, but the
> initial good fortune was not created by virtue of their marketing 
> acumen.

  Well, at least Bill Gates had the good sense to change the
  name of the operating system he bought from "QDOS" ("Quick
  and Dirty Operating System") to just plain "DOS".

  We sold an operating system called "P/OS". Many thought,
  fairly or not, that it was.

                                   Atlant

RE .41 about the luck of Microsoft and Intel 

My point exactly.  Both comapnies overtook a market
without neccessarily having a _superior_ product, 
(whatever the heck that is).

Just ask Motorola/Apple.  
(Yeah, but the Mac is better ;^))

    re: 4770.34 by Vikas Sontakke
    
|   The analogy about 6cyl vs 12 cyl:
    
|   Haven't we started to qualify *our* claims with the addition of
|   *uniprocessor* ?  How is this different then?
    
    You are somewhat correct.
    
    You will find *no* SPECint95 faster than the AlphaStation 500/500.
    No matter how many CPU's you put in a box.
    
    For SPECfp95, there are indeed faster systems available.
    They are:
    
    Digital AlphaServer 4100 5/300 (4)
    Sun Enterprise 6000/12
    Digital AlphaServer 8x00 5/300 (>4)
    Digital AlphaServer 8x00 5/350 (>4)
    Digital AlphaServer 8x00 5/440 (>6)
    
    You'll see *one* system *not* from Digital that beats the
    AlphaStation 500/500.  All of the above systems cost quite
    a bit more.
    
    Honestly, someday HP will submit multiprocessor PA-8000 SPECfp95
    results.  But they have not yet done so.  But when they do they
    may not take number one away from us with any of their current
    (only up to 4 CPU) PA-8000 systems.
    
    (And finally, for what it's worth.  No desktop system, no matter
    how many CPU's are in it, pizza box or mini-tower, has a higher
    SPECint95/SPECfp95 than the AlphaStation 500/500.)
    
    
    Why talk about single CPU performance?  Tongue-in-cheek reason:
    To give some of our competitors a sporting chance.
    
    With current results, we take 9 of the top 10 if we don't restrict
    to single CPU.  Sun places 8th with the Enterprise 600/12.  Restricting
    to a single CPU, we only take 6 of the top 10.  Today HP grabs 2nd, 3rd
    and 4th with the PA-8000s, and SGI with the Power Challenge R10000
    [2MB L2] grabs 9th.
    
    The more serious answer - because some technical customers know that
    their application is best characterized by uni-processor performance.
    They know that throwing more CPUs at the problem gains them little if
    anything for their application.
    
    								-mr. bill

>                     <<< Note 4770.33 by PERFOM::HENNING >>>
>                             -< facts count too. >-

>    perceptions are real
    
>    but so are facts
    
>    One might almost gather that the naysayers in this stream would prefer
>    not to have to bother with the inconvenient fact that Alpha is fastest.

  Facts are facts.  I'm not arguing facts.  I'm aguing what sells.

  If I'm about to purchase based on what I've heard, I want you to walk into
  my office with nothing but your uh er hat in your hand and convince me
  quickly and efficiently that your machine is best.  My time is valuable.

  By the way you should probably be the last presenter on my list for it to
  stick but that's a problem you can't well control.

  	mikeP

RE: .40

>>A 3D workstation?  Get UNIX on SGI ;-)

	That's rapidly turning into "Get a DEC (sic) Alpha 
	running Windows NT" Especially if it's an Alphastation 500/500.

	We turned alot of heads at SIGGRAPH from what I've been reading.

							mike

>  By the way you should probably be the last presenter on my list for it to
>  stick but that's a problem you can't well control.
    
    Thanks :-)
    
    Seriously ... I should like to apologize if my remarks in this notes
    stream have been taken as personally insulting to any individual.  
    
    I am not saying that the perceptions are not real.  It just seemed like
    there was insufficient attention being paid to the little fact that
    Alpha's first place (at X MHz vs. competitor's second place at Y MHz) is 
    STILL FIRST PLACE.  
    
    Thanks for listening.
    	/john

    re: A 3D workstation? Get UNIX on SGI :-)
    
    Or an Alpha with PowerStorm running UNIX...
    Or an Alpha with PowerStorm running Windows NT...
    
    The new PowerStorm 4DT graphics (announced at SIGGRAPH) have better
    capabilities than SGI, better performance than SGI, and lower (much!)
    cost than SGI.
    
    We have a $4,000 graphics card that does true color double buffered,
    high performance 3D, and supports up to 32 MB of texture memory.  The
    $13,000 SGI High Impact graphics card does 16 bit color/alpha double 
    buffered, and supports a maximum of 4MB of texture memory.
    
    Our top of the line PowerStorm 4D60T priced at $12,000, and is
    delivering graphics performance identical to or better than the $23,000
    SGI Maximum Impact.  The 4D60T also supports 1600x1200 resolution,
    while the Maximum Impact is limited to 1280x1024.
    
    The AlphaStation 500/400 and 500/500 are delivering much better
    application level performance than the SGI systems.  One of the
    demonstrations at SIGGRAPH was a visual simulation of a F1 racetrack.
    This simulation was running at high resolution and fully texture
    mapped.  I don't know how they did it, but a 500/500 and 4D60T
    combination was getting over 30 frames per second -- the engineer
    developing this commented that this was twice the frame rate that they
    were getting from their SGI Reality Engine2.
    
    Much of the 3D world is moving to NT -- this was the strongest trend at
    this years SIGGRAPH.  SGI has been claiming that no NT system can do
    real 3D graphics, publishing ViewPerf benchmark numbers to support this
    claim... Alpha/PowerStorm systems running NT blow away the SGI numbers.
    
    This time around, Digital is actually on the leading edge of the
    next revolution, and is in an excellent position to exploit these new
    developments.
    
    Russell Doty
    PowerStorm 4DT Product Manager

Russell:

> > re: A 3D workstation? Get UNIX on SGI :-)
> 
> Or an Alpha with PowerStorm running UNIX...
> Or an Alpha with PowerStorm running Windows NT...
> 
> The new PowerStorm 4DT graphics (announced at SIGGRAPH) have better
> capabilities than SGI, better performance than SGI, and lower (much!)
> cost than SGI.
    
  How much of the vast base of SGI software has been ported to Alpha?
  Is it at rev-parity with the SGI versions?

  How much is in the process of being ported?

  What's not being ported?

                                   Atlant


P.S.: I honestly don't know the answers to these questions, but
      I know they're questions any Mac- or SGI-based shop would
      certainly ask before they even let you in the door.

http://infopad.eecs.berkeley.edu/CIC/ is a good source for CPU benchmark
information accross the industry.


							-Paul

    Excellent questions!  In fact, the very questions we were asking
    ourselves several months ago.  As a reference, we announced the
    PowerStorm 4DT on August 6, and will be volume shipping in
    mid-September.
    
    Now, relative to SGI and applications...
    
    First, we are in much better shape than SGI on NT applications ;-)
    (As a reference, I'm running on my desktop an Alpha with a 4D40T and
    Windows NT.  I run most of the target applications and demonstrations,
    plus MS-Office, Corel Draw, Netscape, Access, MS Project, etc.  Not
    only can I run Intel apps (FX!32), but Intel apps that use OpenGL run
    with fill hardware graphics acceleration!)
    
    We targeted 3D technical applications.  We have had hardware on site
    for several months with active testing and optimization at the key
    ISV's, and have begun the formal qualification and certification
    process, with the goal of having applications certified by FRS.
    
    Mechanical CAD
    	Pro/Engineering by PTC
    	Unigraphics by EDS
    	Euclid by Matra Datavision
    	SDRC
    	Computervision
    
    	We have done some testing with Bentley MicroStation.
    
    Visual Computing
    	AVS
    	OpenInventor by Portable Graphics
    	World Toolkit by Sense8
    	Vrealm (VRML) Builder and Browser by IDS
    
    Entertainment
    	SoftImage (extensive work with them; EXCELLENT opportunities here)
    	Lightwave3D by Newtek
    
    And, since the PowerStorm 4DT series uses the same interfaces as all of
    our other graphics (OpenGL and either X-windows/Motif or Win32), all
    applications that run on Alpha will run with PowerStorm 4DT.
    
    We believe -- based on market trends -- that the biggest growth
    opportunities lie with WindowsNT.  Virtually all of the new
    applications development we are familiar with is originating in the NT
    space, and SGI has no presence in NT.
    
    SGI does have a number of applications that we don't.  We are going
    after most of these applications.  The vast majority of these application
    vendors are either moving to NT or adding support for NT.  Many of them
    are responsive to the Digital NT story (both Intel and Alpha).
    
    So, application availability is the most critical question to ask.  We
    have the core applications we need to succeed, we have more work to do,
    and I believe we are in amazingly good shape for a brand new product
    that has not even shipped yet!
    
    (also: If you know of any critical applications that we should be
    targeting, send me mail.)

    
    Stick 'em in the arse with a fork, without a ;-) and see what it gets
    us... a serious argument about DEC vs SGI as a new rathole under the
    wrong topic ;-)
    
    Oh well.  We'll own the 3D desktop in 3 years. ;-0  Nah, we'll spin
    the business off if it looks like a threat to SGI ;^)
    
    You guys are way too serious.
    
    'scuse me while I morph out of here.
    
    

    FWIW, Kinetix (developers of 3D Studio Max) put on hold porting their
    application to Alpha.  Cadence, a major provider of CAE software,
    basically dropped their support for Alpha.  Other CAD vendors have 
    threatened to drop support unless they get a wad of cash.  I don't 
    like this trend, if it is a trend.  These can all be dismissed as 
    just a few, tiny markets, but as a deccie who loves working with 
    Alphas, 3D and CAD, this hurts.  
    
    I saw Alpha boxes being auctioned off on the internet for less 
    than $1K.  If I knew that 3D Max was definitely headed for Alpha, I 
    might have gone for one.  Instead, I've gone with adding another P5 
    system for about the same cost (rolling my own) because I *know* it 
    will run all my NT applications, including future applications.  This
    visible (to me) reticence from some software vendors to support Alpha
    makes me shy about buying one for my own general use.
    
    I was enthusiastic about Alpha for 3D work since I see a lot of 3D work
    heading toward garage shop outfits and being done on machines and
    software that don't require investing a small fortune.  I think that is
    where a large part of the 3D market is headed.  But, I see Intel boxes
    as more a threat to SGI than Alpha boxes for 3D in the future.  This is
    partly because I think the 3D market, small as it is, is headed toward
    being more of a high-volume, low-margin market that both SGI and
    Digital are unprepared to dominate.
    
    Steve

    ...it should probably be mentioned that Kinetix canned
    3D Studio Max after evaluating Alpha's performance relative
    to P6 boxes.
    
    Time for DS to throw some money at them.
    Kratz

    On 2 different contracts this year for 2 different companies
    I looked into using an Alpha as a server for intranet applications.
    In both cases Alpha lost because there was not an Alpha version
    of some piece of software we needed.  
    
    Yes, there may be nn,nnn products ported to Alpha, but there always
    seem to be one or two other critical programs that haven't been
    ported, so the whole house of cards falls and we use an Intel box.
    
    When you are arguing about performance you are just polishing turds.
    
    To munge up a phrase from the last US presidential election:
    
    "Its the _applications_, stupid."
    
    -Vin
    
    

Low volume: moving from "threatened" to "endangered".

Our hardware-centric management has frequently seemed to come up short
with respect to software in the past.  (Back in the mid-80s there were
questions about how to make our software development environment succeed.
One response [which was ignored] was competitive price/performance
platforms with our competitors at the time [e.g. Sun].  There was
a substantial "price and price/perf gap" for many, many years.)

Things are really no different now, except we've going from "enslaving"
(extract profit, volume-limiting, under-investing in) our own software
(which we were capable of doing at the time), to being incapable
of providing platforms where ISVs now see this (correctly) as an externality.
If you were an ISV, why on earth would *YOU* want to port to a very
low volume, very high priced platform when you could sell 10x or more
on higher volume platforms??  Why bother with a tiny niche platform??

At one point, it looked to ISVs that Digital was actually serious about
making Alpha NT a reasonable volume platform such that a wide variety
of ISVs might find it interesting.  It has become painfully evident
to them that in fact Digital is NOT persuing a strategy that has systems
that are either price and/or price/performance competitive with Pentium Pro
(for example), and are NOT going to come close to achieving even a small
fraction of the volume (unless something changes radically, and based
on a dozen years of history, the ISVs have ZERO reason to believe that
such a radical change will happen.)  This is a classic *negative* feedback
loop where ISVs are halting or withdrawing products in a number of cases.

Volume is NOT a cure-all to be sure.  However, lack of it (and obvious
lack of pursuit of it) means that you don't even get into the game
for the long term.

    Re: .55
    
    Did you contact the FX!32 people to see if they could work those
    problems?
    
    There's nothing new about application vendors demanding big bucks to
    port their software.  They see a way to make major money and they try
    it.  That's another reason FX!32 is important, because it reduces
    their leverage.
         

    ...or FX!32 will give ISV's *less* incecentive to port a native
    version to Alpha, since the Alpha users can always run the x86/32
    image under FX!32 and Digital, not the ISV, then gets to shell out
    the bucks to support it when it is run under FX!32.
    Kratz

>>> This is partly because I think the 3D market, small as it is, is 
>>> headed toward being more of a high-volume, low-margin market that both 
>>> SGI and Digital are unprepared to dominate.

Yes, I see this happening, too!

Microsoft is pushing 3D into the genaric PC systems in the form of 
Direct-3D API, for use in video games.  Microsoft is targeting the 3D PC 
against "console" game systems, like Sega Saturn, which support dedicated 
3D hardware.

The PCI S3-virge 3D cards are available now for about $150.00.  The 
"serious" 3D folks laugh at the Virge performance, but sales of a bazillion 
cards will fund the next generation.  Almost every video card manufacturer 
is preparing low-cost 3D cards for Microsoft's Direct-3D API.

The drop in RAM prices will also make an 8Mb VRAM card more affordable, 
where half of the VRAM can be used for 3D buffering.

Unless we have extreamely high performance 3D cards *and* the right 
application, we will be over-run by the fast moving PC crowd.


							-Paul

    We are just now entering a new (to us) market: entertainment and
    animation.
    
    We have the hottest application -- SoftImage -- running on Alpha,
    optimized, and fully supporting our newest graphics (including 3D and
    texture mapping).  We have the number 2 application, Lightwave 3D,
    running, including our newest graphics.  We received tremendous
    response at SIGGRAPH, including several of the largest animation houses
    starting serious conversations and people literally trying to purchase
    our demonstration systems off of the show floor.  I worked with several
    of the SoftImage corporate animators, who are used to high end systems.
    They all, without exception, asked "how can I get one of these 
    systems for my office?"  It was kind of fun, watching them get
    blown away by the performance of a 400 Mhz Alpha -- and then apologize
    for giving them one of our slow workstations!
    
    We don't have a committed port date for 3D Studio Max (the real issues,
    ummm, are not technical or performance...).  Our recommendation is
    to sell SoftImage and Lightwave 3D on Alpha, and 3D Studio Max on
    Intel.

    Russ,
    
    I worked with SoftImage at IBC95 last September (i work for CSS) when
    they first demonstrated their product (behind locked doors) to their
    customers/prospects and they were blown out by a 300Mhz AlphaStation
    600, so i can appreciate the feedback you are now getting with 400 and
    500 MHz machines.
    
    Gerald Connolly @SBP

    Gerry,
    
    You really need to see these new systems running SoftImage!
    
    In the past, graphics has been the major bottleneck on Digital
    workstations -- literally any Alpha processor could over-run our best
    graphics.  With the new PowerStorm 4DT boards, the PROCESSER is the
    bottleneck.  Even a 500 Mhz EV5 doesn't max out the graphics (although
    it looks like it is getting close).
    
    In addition, the new graphics support shading and up to 32 MB of
    texture memory, which SoftImage fully exploits.
    
    My personal belief is that graphics will do for our workstations what
    VLM has done for our servers -- provide a clear and compelling reason
    to buy Digital systems.

Russ,

	Well it looked pretty good at 300, I think i saw it at 400 at NAB this
year but with faster graphics i not surprised the SoftImage people are really
impressed!

Gerald

    See the following URL for a peek at Intel plans for commodity 3D
    graphics controller targeted for 2H97/98.

http://www.intel.com/pc-supp/platform/gc97/


    							-Paul

    Read a book called "Accedental Entrepreneurs. About the Boys from
    Silicon Valley Who Created an Empire, Fought Off the Japanese and Still
    Can't Get a Date".  Its a fun, quick read.
    
    The author's point is that every computer architecture needs a
    compelling application.  If you get one, the other applications
    will move too.  Compelling is defined as something that many people
    need that is only available on your box.  (It could just be stunningly
    better too.)  Apple 1 had Visicalc.  IBM PC had Lotus 123.  Apple MAC
    struggled until it got Desktop Publishing.
    
    Alpha Servers finally found Data Warehousing.  With proper brand
    marketing, we might be able to keep the lead.  We need to establish
    the image that Data Warehousing = Alpha.  We won't get it done with
    Specmark charts.  
    
    Alpha workstations haven't found a way to exploit it's 64 bit
    exclusive yet.  3-D rendering may be it, but it seems like a small
    market segment.  32 bit NT in 64 bits sounds wastefull.  It will be a
    long time before the market can use 64 bit NT. Success in data warehousing 
    won't pull the workstation applications.
    
    Wish I knew the answer.  Does anyone have a nominee?
    
    

> Read a book called "Accedental Entrepreneurs. About the Boys from
> Silicon Valley Who Created an Empire, Fought Off the Japanese and
> Still Can't Get a Date".  Its a fun, quick read.

  Make that "Accidental Empires ..." by Robert X. Cringely, the
  fellow who for years was the "Info World" gossip columnist.
  (The column still runs under his name, but "Bobby and Pammy"
  are now ghosted by someone else and lawsuits are underway.)

  Cringely has also done a very compelling TV series for
  Public Television.
                                   Atlant

  Personally, I think "Data Warehousing " and "VLM" are both
  way too esoteric to make it as the compelling apps that save
  our butts. Artistic 3D rendering is pretty obscure too�.

  I mean, who can't understand and productively use spread-
  sheets? Or desktop publishing? Who *CAN* use data warehouses,
  VLM data-bases, or 3D modelling?

                                   Atlant


� In contrast, the introduction of 3D APIs by both Apple
  ("QuickDraw-3D") and Microsoft mean that lots of pc's will
  be rendering 3D under the direct control of applications
  programs (and applications programmers) rather than creative
  graphic artists. But this is a market where lower performance
  3D rendering, well within Intel's capabilities, will probably
  be more than sufficient. No need to pay for an Alpha just to
  run "Doom, ripped limb-from-limb in 3D!"

    >>Who *CAN* use ...VLM data-bases
    
    Well millions of people a day use Alta Vista, and this lead the
    competition to throw away their free-of-charge Sun boxes and pay the
    asking price for Alphas. They get high on the number of hits, not the
    technology underneath.

> Well millions of people a day use Alta Vista, and this lead the
> competition to throw away their free-of-charge Sun boxes and pay the
> asking price for Alphas. They get high on the number of hits, not the
> technology underneath.

  You've answered the rhetorical question that I asked.

  So now let's complete the answer: VLM can be productively used
  by people/institutions with very large databases. And as we've
  seen in the example of Alta Vista, a relatively small server can
  service the entire world full of consumers _of_the_database.
  And there just aren't that many mega-databases.

  So now let's move onto the meta-question: Will VLM sell enough
  machines to act as a "Killer App" for Alpha? Or even for 64-bit
  computing in general?

  For Alpha: I think not. For 64-bit computing: No more than a "Maybe".

                                   Atlant

    >  Personally, I think "Data Warehousing " and "VLM" are both
    >  way too esoteric to make it as the compelling apps that save
    >  our butts.
    
    How large is the total market for these applications?  As a
    mainstay of mainframes, Data Warehousing should be a large market.  
    
    VLM is a technology.  Will data mining and other applications that use VLM
    become large before IBM and HP can deliver VLM?
    
    The key to a compelling application is that the customers pull other
    applications to the computer architecture.  For some customers, data 
    warehousing and and data mining could be important enough for them to 
    put heat on other application suppliers.  Are we getting this type of
    pull for Alpha?  Are the application suppliers complying or fighting?
    
    
    
     
    

     >>>Will data mining and other applications that use VLM
       >become large before IBM and HP can deliver VLM?
    
    Well the point is, isn't it, that they are going to have to put their
    best blue suits on and iron their ties and explain to their customers
    that if they want these nice new applications they will have to migrate
    to a 64-bit world. And by the time they have finished doing that, and
    getting the know-how and organisation together to do it, the customer
    may be thinking, do I need all this? Isn't there someone who has done
    this already? And the hard bit for them is, timing when to take the
    plunge. Too soon and they lose share to those still pushing the cheap
    32-bit boxes. Too late, and Digital extends its lead. Just remember
    those Sun boxes going out the window.  It gives me a warm glow.  my
    guess is there will be a shakedown in the 32-bit world and Digital will
    thrive in the next 10 years. It would help if we got our management act
    together though. 
    
    On the question, will VLM sell enough boxes, I think so. It gives a
    competitive advantage. Those Sun boxes HAD to go, or the company using
    them went. My line is that we have achieved a tremendous amount in the
    64-bit world, but that the extent of the achievement is not yet
    apparent.
    
    Kevin 

    VLM, data mining, data warehousing... are they "killer" apps?  Well,
    the answer is quite possibly - yes.
    
    A group of companies trying not to be swallowed by Microsoft are trying
    to sell a alternate universe than the Windows/Intel one that we seem to
    be heading towards.
    
    The group of companies, spearheaded to a great degree by Oracle, have a
    view that the new world of the "information superhighway" will be one of
    inexpensive clients (see the recent Netscape aliance) which access vast
    information - which of course needs to be organized in a manner that it
    can be made useful.  Anyone who surfs the web can tell you how hard it
    is to navigate through a disorganized jumble of hypertext links --
    without aids such as AltaVista to provide "some" amount of
    organization.  Now think of being able to go to a web page that would
    allow you to pose interesting "research" questions that might sift
    through the terabytes of data floating around -- to determine things
    like future market trends.  Yup.  It might be a killer app.
    
    So, big fast relational databases are a key element of such a universe.
    But PC servers are not very interesting.
    
    Now.  Quite beyond this new world order, VLM, data mining, and data
    warehousing are all very good high margin, low volume business -
    which after all is the only business DEC has traditionally succeeded in.
    
    Where DEC is missing the boat is not being the vendor developing the
    cheap information appliances... although I've noticed a definite lack
    of information flow lately -- for all I know we may actually have plans
    to do something with StrongARM.  That's where the high-volume,
    low-margin product will be.
    
    

> So, big fast relational databases are a key element of such a universe.
> But PC servers are not very interesting.
> 
> Now.  Quite beyond this new world order, VLM, data mining, and data
> warehousing are all very good high margin, low volume business -
> which after all is the only business DEC has traditionally succeeded in.

  Thanks Fred.

  I hate to trot out the old, discredited "The world will only need
  three computers!" argument, but even if we're talking about tens
  of thousands or hundreds of thousands of these VLM data mines,
  we're still not taking about a business that will sustain Digital_
  as_we-know_it or make Alpha a success.

  VLM = Interesting niche, profitable for the smaller and/or nimbler.

  But Alpha still needs that killer app.

                                   Atlant

    What was that old diagram... the "Wheel of Re-incarnation" (or
    something like that) which showed technology moving in a cycle
    between centralized and distributed computing.
    
    I doubt that there is any single product that would provide a
    sustainable level of profitability more than a few years.  By
    being the leader, you get a short grace period before the rest
    of the pack leverages your work into a lower margin business.
    
    So.  Ask the question, what comes next?  And place your bets.
    And also ask what business DEC has a credible shot at winning.
    
    Some people are guessing that large distributed database servers,
    and cheap clients (sounds a lot like timesharing and terminals ;-)
    is the next wave.  If you aren't going to build the cheap clients,
    then you had better have the software and hardware for the servers.
    
    

RE: StrongARM

	From what I've been reading, this is ramping up very quickly as
	one of the top contenders in the NC race. (NC = Network Computer)

	Do I think we should product a NC of our own? Probably not. We
	could re-badge someone elses, if they take off.

	Do I think we can be successful producing some of the components?
	Yup.

							mike

    
    As discussed here and about, we need ALL of the applications for a
    usage segment.  I remember losing a workstation bid a few years back
    because we lacked a radar antenna simulation application.  The supplier
    was a guy who worked out of his house in the hills of Silicon Valley.
    SUN lent him equipment when he started, so his application worked only
    on SUN.  
    
    Despite low volumes, Data Warehousing, Mining, etc. could be a killer
    application if it causes the customer to force other application
    suppliers to port.  Data Warehousing, etc., buyers are big IT buyers.
    Thay have clout, if they want to use it.  Will they pull enough other
    applications to Alpha to build momentum?
    
    I wonder if Digital has built a market application map for this use
    segment and how much we have to fill in yet.
    
    It is like starting at the point to build the great pyramid.
    
    
    

  StrongARM will also figure in the next-generation Newton coming
  from Apple around Christmas time.

  You might argue that this next-generation Newton could be one
  example of the Network Computer. It's also a lot more, and
  powered by StrongARM, it should finally be able to shake
  many of the woes and jokes that the earlier Newtons provoked.

                                   Atlant

    I read with interest recently (I think it was Business Week)
    about a useful PDA (the Newton is less than useful) concept...
    The concept was that it was a limited purpose, extension to
    a PC.  You connected it simply to your PC, and the PC would
    then download things like -- your updated phone directory, and
    calendar.  And upload things to your PC that had been input
    since the last docking.  Pretty cool stuff.
    
    Gone were the things that even remotely made it try to mimic
    a pocket PC.  And it even fits in a pocket.
    

    StongARM will power Newton???  That is great!  
    Second generation products can take off if the producer listens carefully. 
    (LA36 vs LA30, VT100 vs VT52, Apple Mac vs Lisa, MicroVax 2 vs MV1,
    Windows 3.1 vs Windows 1&2).  Newton hit the market and messed up once.  
    Second try might work.  Could give Hudson some volume.
    
    I have less hope for the NC terminal.  If it boots off of the net,
    start up will be very slow until the network infrastructure improves.
    If it boots internally it will need non volatile memory or disk to
    store the start up program.  Both are getting cheaper, but still
    expensive for a $499 device.   I think we will see several attempts
    at an NC before someone gets it right.  A cellular phone may be the
    best model, were the connection service subsidizes the hardware.
    It will take a while be for the NC hits consumer driven volume.
    

RE: Note 4770.80 by SUBSYS::JAMES 
    
>    StongARM will power Newton???  That is great!  
    
    Apple has always been the first customer for StrongARM.  Ed Caldwell 
    was showing off a StrongARM based Newtons in Hudson a couple of weeks
    ago.
    

	It strikes me as ironic that a Digital powered Newton could 
possibly help ressurect Apple (and Digital) years after, from what I have 
read/heard, we didn't strike a deal with apple to use Alphas when they were 
looking for a processor for their new computer architecture.

	C'Ya,
	Wayne

>    Apple has always been the first customer for StrongARM.  Ed Caldwell 
>    was showing off a StrongARM based Newtons in Hudson a couple of weeks
>    ago.
 

Does he still have them? What is the handwriting recognition like now? 
Can it keep up with normal writing?

Mark
 

Mark:

  Even without Strongarm, the newer Newtons have much-improved
  recognition. (More RAM was a big help.) Also, now they can also
  do "deferred recognition", storing your writing as "Digital Ink"
  until you do recognition later. Thus, they can keep up with you,
  recognition or not.

  StrongARM, though, should make things even better.

                                        Atlant

    
    
    The best little ooooh ahhh we got from the demo of the new
    Newtons was the voice/text  text/voice recognition.  The 
    Newton read a memo from some Apple big wig directed to Ed
    thanking him for Hudson manufacturing the StrongArm.
    
                                                  db3 

re: .84

I have a Message Pad 110, which has "magnetic ink" and deferred recognition.
With those, if you very carefully craft a single word, it can get it after
a few seconds.  As soon as you start writing at "normal" speeds it
falls apart pretty rapidly. Even with deferred recognition, your writing
is so different at speed that it can't recognize it. Pen and paper are
much quicker. 


Text to voice may make a good demo but it's commonplace on some sound 
cards. Even a 386 can do well at that. Now voice-to-text is different
matter!

Mark

Mark:

> I have a Message Pad 110, which has "magnetic ink" and deferred recognition.

  Note the models I claimed in my title. Your 110 doesn't have the
  advantage of the newly-increased RAM, does it?

                                   Atlant

I missed that, good point. Time for a demo.

M