Conference 7.286::digital

From another string:
    
    >the attitude of senior management was "Hey, we're DEC, we're great,
    >we're sure to win!" .. 
    >
    >Sound familiar?  8-(
    
    					andrew

    The worksystems group is selling the NT/Alpha boxes near their cost
    (don't forget the 20%+ reseller markup their "list" prices carry).
    Their costs are high due to a couple of reasons: they carry excess
    baggage (either directly or thru the groups they have to fund) and
    consequently aren't terribly efficient, and their volume is low.
    One could, of course, argue that their volume is low because their
    prices are so high.
    
    For a low cost, hi performance Alpha/NT box, its either got to come
    from the outside or from elsewhere in this company (for example, the
    PC group or terminals).  However, worksystems is also politically very
    powerful, and won't tolerate someone stepping on their toes.  If
    you're from Massachusetts, it's the equivalent of the control the
    police have doing the flags at construction sites... yes, it's
    terribly inefficient, but there's nothing you can do about it.
    .02 kb

    Would some one please answer me a question: Why would some one need a
    266 Mhz computer? What does one do with it? Why does some one need 1.2
    million bytes of information per second for? I know it sounds nice to
    say "I own one of those Alpha`s" but why would I need anything like
    that. 
    
    No smilies here for this.

For that matter, why would you need a computer at all?

Just hire some more people to do the work manually... :-)

/Mario

RE: .3

	Why do I "need" a 120Mhz Pentium PC? It'll run PC-based
	Flight Simulators really fast and smooth. That gives me enjoyment.
	Hence, I "need" it. I also feel I "need" a Mazda Miata too BTW :)

	Need is relative. Don't question a customers need. Need for me could 
	be profit for you.

						mike

re: .2
>    For a low cost, hi performance Alpha/NT box, its either got to come
>    from the outside or from elsewhere in this company (for example, the
>    PC group or terminals).  

	I do not believe this is true.  If we move the build of a system to 
an outside company then we just increase the cost on everything for that
division and ultimately to DIGITAL.  If we assume that we own the equipment
already and that we don't layoff the people in that facility, then we have just
added the payroll and profits of the outside company or division for building
that system to our costs. 

	It is my guess (with very little data) that our largest source of cost
is that we don't sell what we are capable of making.  (That is NOT a dig at the 
sales force.)  In other words, underutilization cost.  If marketing, sales,
design and manufacturing work together to more fully utilize our facilities,
then we might be able to make a low cost, hi performance Alpha/NT box.

> 	I do not believe this is true.  If we move the build of a system to 
> an outside company then we just increase the cost on everything for that
> division and ultimately to DIGITAL.
    
    It's a fact.
    
    Nechotech Mach 2-289-T (a 289 MHz 21064A): $11,995
    AlphaStation 400 4/222 (a 233 MHz 21064A): $16,394
    
    The above prices are from the March issue of Byte where they tested a
    few Windows NT based RISC workstations. The configurations are almost
    identical (disk size, RAM, SCSI etc.). Graphics adapters are slightly
    different.
    
    Olav

I don't know all the details of the Windows-NT system we were selling, but
given that it runs NT I am speculating that it does share a good deal of
components with our PCs. Graphics card? Disk controller? Memory? Monitor?
Mouse? Keyboard? CD reader? CD controller? Hard disk itself? Enclosure 
(box)? Cables? Power supply? SIMMs? Ethernet card? Generic documentation 
for the above? What is left? Backplane (isn't it shared too?). The only 
thing I can think of that is unique is the mother board and the chip ..

Ashame that CSS isn't around any more .. they could buy some Celebris
down at Sam's club, rip 'em appart and get some Alpha chips from Tadpole.
Get some 16Mb SIMMs from Kingston, a Radio Shack mother board and hey
presto a $5,000 Alpha-PC? Maybe I am generalizing too much :-)

>Per

    Re: .8
    
    Compare a Celebris XL 590 and a AlphaStation 400 4/233. What is the
    difference? The CPU board *only*. In fact you can upgrade a Celebris XL
    to a 233 MHz Alpha (card available from the PCBU). You will end up with
    a system that costs about 60% of the equivalent AlphaStation 400 4/233
    from the SBU.
    
    Olav

RE: .8

     CSS is alive and well!  As for the rest of the smily... :-)

Kevin

So, in other words, our pay-freeze and lay-off savings have subsidized
the development of a system that already exists! In addition, we have
used some more corporate funds to tell the world about it ... when they 
could have used the money to sponsor work to upgrade the PCBU board and 
create a system that would actually be worth bragging about ..

Lets hypothesize a bit here. The Alpha chip costs X dollars more than 
the Pentium. Upgrading the PCBU board to 266MHz and adding 16Mb of memory
would be an additional Y dollars. The cost of Y goes down as we sell more
systems. If we were able to sell 10% of our Pentium volume of such Alpha
PCs, what would that X+Y be? I know this is generalizing, but would anyone 
care to make a guess? BTW, has anyone done any performance testing of the 
speed of the Intel emulator on the 266MHz?

>Per

    > Would some one please answer me a question: Why would some one need a
    > 266 Mhz computer? What does one do with it? Why does some one need 1.2
    
    Because it builds my compiler faster than a 190 Mhz Jensen and MUCH
    faster than the VAX in our cluster.  Thus, I spend less time waiting
    for a compiler to test so that I can figure out what new bug I just
    introduced.
    
                                august

    The need for speed ...  How about rendering graphics?  Some claim that
    the Alpha is becoming the standard for rendering.  Alpha boxes get
    regular press in the rags I see like Digital Video and Computer
    Graphics World.  
    
    Steve

> Lets hypothesize a bit here. The Alpha chip costs X dollars more than 
> the Pentium.
    
    I don't know about the chip prices, but isn't RISC chips supposed to be
    *cheaper* than the CISC chips like the Pentium. The PCBU 233 MHz 21064
    upgrade CPU board costs about the same as the Pentium 90 CPU upgrade
    board (both for the XL).
    
    Wouldn't it be great if you could buy the XL with an Alpha CPU board
    without paying for the Intel CPU first? Then we might have a killer.
    2-3 times Pentium 90 performance for the same price!
    
    Olav

  The recent Dataquest presentation in Hudson (only attended by
  about 20 people, BTW) suggested that no, conventional wisdom
  hasn't worked out and that no, RISC isn't even *FASTER* than
  CISC. They showed curves with Pentium performance crossing
  conventional RISC performance (e.g., PowerPC, Sparc, and
  PA/RISC) soon.

  They did throw a bone our way saying that "Aplha was way off
  the curve" in terms of its performance, but that was one of
  only two times in the entire presentation that Alpha was
  mentioned; it seemed to be way off their radar screens.

  They also mentioned what I believe was 25,000 8" wafer starts
  *PER WEEK* on the Pentium line. That drives the Pentium costs
  real low real fast, so there goes the RISC/CISC areal cost
  advantage, too.

  (The videotape is available inthe Hudson library.)

                                   Atlant

>  The recent Dataquest presentation in Hudson (only attended by
>  about 20 people, BTW) suggested that no, conventional wisdom
>  hasn't worked out and that no, RISC isn't even *FASTER* than
>  CISC. They showed curves with Pentium performance crossing
>  conventional RISC performance (e.g., PowerPC, Sparc, and
>  PA/RISC) soon.

Interesting.  How did they explain this?

- Intel has been more successful at increasing Pentium performance
  than expected.

- Intel has been able to successfuly apply RISC techniques to
  to the Pentium architecture.

- The conventional RISC/CISC argument is is invalid because...

I thought the RISC performance benefit was fundamentally based
on the fact that memory systems have improved much faster than CPUs.
With large fast memory systems, it is more efficient to retrieve
instructions that can be executed quickly in parallel than
to spend time decoding and sequencing more complex instructions.

I remember some argument about theoretical limits within the CPU itself?

Than again, the difference between theory and practice is that practice
takes all of the theory into account :-)

- Peter

    
    The SAD thing is that if we did a state of the art VAX CPU after
    NVAX we would be ahead of Intel in CISC too...  Not sure it would
    have mattered but at least we have a better commercial CISC machine
    that runs VMS.  We could even port Digital-UNIX to it...
    
    The 4 year old NVAX chip has run at 100MHz since the day we powered it
    on (CMOS-4)...  The Pentium is just getting there.
    

Back to the original question:

>What I do not understand is why the Windows NT based system,
>that appart from memory size and CPU has a pretty average Pentium config,
>is so darn expensive. It lists for about $13,800. What, appart for the
>CPU, do I get for the extra $10,000? A 486/66Mhz emulator?

One thing you get is 2Mbytes (I thnk) of fast cache RAM -- I haven't seen
a Pentium configured like that yet.

    Alpha needs the large cache to perform to potential (take a look
    at Multia's 21066 limping along with 256k); Pentium doesn't.
    Alpha's architecture needs more money thrown at it; hardly a feature. 

There are two different issues being dealt with at the same time here,
which makes a straightforward comparison more difficult:

    1) A state-of-the-art RISC implementation vs a state-of-the-art CISC
       implementation targetted toward the same market.  (Note that
       unfortunately, Alpha wasn't initially targetted at the same
       market, otherwise some things would have been done differently -
       for example, code size would have been recognized as being
       important for the mass market, as well as things like driver
       portability, so byte/word memory accesses would almost certainly
       have been included.  Even things like 4K page size would have
       been deemed more important for compatibility.)

    2) The resources available by particular vendors to produce particular
       parts, and to produce them at particular volumes/prices

Ignoring #2, you can argue that RISC is in fact considerably either
faster or cheaper (or both) in the same technology if designed
to the same market constraints.  (Basically, the complexity of CISC
has a cost you take take in some combination of chip space, design
time, cycle time, etc.)  In this context, RISC *is* faster/cheaper -
BUT it ignores "initial conditions", which are important.

The problem is that in comparing x86 and RISC in the real world, #2
factors in significantly, and Intel has an enormous amount of resources
compared to most of the RISC vendors, plus they are able to draw
on near-monopoly profits to finance future versions, and "throwing
money at the problem" does in fact work to some degree.  Combined
with the pain of architectural incompatibility (which can be addressed
to some degree via translation/emulation), it isn't at all clear that
RISC has an advantage, particularly given:

    - The RISC vendors are very fragmented (It isn't like there was
      a single RISC architecture competing to displace x86)

    - Most of the RISC vendors are abysmal at marketting compared to Intel

    - Most are pushing OSes that are not user-interface compatible with
      the majority of systems in the world (PCs)

The combination makes it extremely difficult to see how RISC is going
to gain any inroads (unless the RISC processor comes from Intel, and
given the number of alternative x86 manufacturers likely to be around
at the point where Intel may try to do that, it isn't even clear that
Intel will have the clout to introduce a new architecture at the point
where there is really increasing competition.)

                  <<< Note 3791.8 by KOALA::IRIE::hamnqvist >>>
                            -< Whatever it takes? >-

>Ashame that CSS isn't around any more .. they could buy some Celebris
>down at Sam's club, rip 'em appart and get some Alpha chips from Tadpole.
>Get some 16Mb SIMMs from Kingston, a Radio Shack mother board and hey
>presto a $5,000 Alpha-PC? Maybe I am generalizing too much :-)

    1. Celebris is not available at Sam's Club.
    
    2. Celebris is 5 volt only (perhaps you mean Celebris XL?)
    
    3. Radio Shack does not carry AXP compatible motherboards.
    
Mark.
    

>    - Most of the RISC vendors are abysmal at marketting compared to Intel

How do you think Motorola will do with PowerPC? By many considered to
be the Intel alternative.

Olav

Dataquest assessment was essentially that the PowerPC is a mediocre RISC
design with performance limitations and will be overtaken by X86, that IBM
is doing a miserable marketing job for PowerPC with an unfocused OS strategy
and virtually no application software, and that PowerMacs won't be enough
to make up the shortfall.  In other words, PowerPC will eventually fail.

Mac

Personally, I don't expect PowerPC to do all that well.  (Apple "may" not lose
market share as fast if it passes along cost savings.  I don't expect PowerPC OS/2
or PowerPC WNT to take significant market share, since they don't provide any
significant advantage to switch either in terms of price, price/performance, or 
performance.  One advantage Alpha could provide would be significant performance
advantage at the high end - IF price at the low end system prices were more in line
with reality, and IF the price/performance curve were a little below x86, instead
of above it.)

Note that if you consider shipping systems, and look a year ahead as well, Alpha
is the ONLY RISC processor that has a significant integer performance advantage
over x86 (measured using SpecInt, which is definitely a flawed metric, but it
does bear some small relationship to delivered integer performance.)

    Ahem...lest we forget, Apple is closing in on 1M PowerMacs shipped.
    
    Mark.

And all totalled, Apple has less than 10% of the PC market, right?
Apple/PowerPC is "extremely unlikely" to break Microsoft's hold
on the market (note that I didn't say Intel's hold on the market.)
Odds are much more likely that Apple will gradually wither and die
(the "Multics" of PCs.)

What matters in general is NOT "an architecture" (e.g. PowerPC).
That provides silicon economies of scale only, which can, at least
in theory, be provided by multiple architectures for a given foundary
(e.g. Digital => Alpha + AMD x86.)  What has mattered, and matters
increasingly, is the *binary compatible installed base* - i.e. what
ISVs see when they go to produce software.  (i.e. Win16/x86 has a very
dominant position, potentially moving to Win32/x86 relatively quickly.)
Note that that Win16 ABI encompases several OSes (DOS+Windows, OS/2,
a variety of emulators, etc.), and several hardware platforms
(predominantly x86, and of that, predominantly Intel.)

If any RISC is really going to succeed (in the mass market), it's
probably going to require:

    - The ability to reliably run win16/x86, DOS, and win32/x86 applications
      at some reasonable fraction of then-current upper-end x86s, plus
      at least some fraction of significantly faster native applications.
      (An alternative would be a lot more native applications, if you
      have the $s to throw at ISVs, and the ISVs take you seriously.)

    - A reasonably PC-compatible user interface (Win apps in Unix windows
      with a Unix desktop UI doesn't count...)

    - Buy-in from Microsoft in some form (probably in the form of the OS)

    - True multiple source competition.  (Single-vendor solutions
      or even "limited" competition is unlikely to compete with
      the nearly pure competition already present in the PC market.)

    - MARKETTING

    - Better price/performance and higher absolute performance than x86

    - Go almost all of the way down the x86 price band (e.g. <$1500 today,
      less tomorrow)

And that's probably just the price of entry *for a shot* at a franchise.
(You also have to be able to deliver a steadily increasing volume, etc.,
and do the various things required to generate a positive, rather than
a negative, feedback loop.)

Intel can probably pull off most of these, though 4 is where it's likely
to come up short, and that could cause the market to reject it, even if
it's "superior" in some form (e.g. "New Coke")

    Apple had 12.3% of the '94 US desktop market, second only to Compaq.
    
    However, my point is that saying that the PowerPC will not succeed
    while lauding the capabilities of Alpha is specious considering that
    Apple has shipped a cool million in one year, while the AXP lags far
    behind.
    
    Mark.

    Hey, keep facts out of this discussion, huh Mark?
    ;-)

One reason to pump up the volume?
    
            <<< DECWET::DOCD$:[NOTES$LIBRARY]NT-DEVELOPERS.NOTE;1 >>>
                         -< MS Windows NT Developers >-
================================================================================
Note 2443.0                 Gartner Group on NT/RISC                   3 replies
SSDEVO::PARRIS "RAID-5 vs. RAID-1: n+1 << 2n, in $" 187 lines   6-APR-1995 11:12
--------------------------------------------------------------------------------
From VTX IR:

Subj:	GG-MCS The Risk of NT/RISC (MR018YSC.TXT, ASCII Document)

--------------------------------Summary-----------------------------------
While Microsoft's advanced OS is gaining momentum on Intel-based servers,
Alpha- and MIPS-based demand has been minimal.  We believe it is unlikely
that any NT/RISC combination will reach critical mass in the server market.
--------------------------------------------------------------------------

Copyright (C) 1994, Gartner Group, Inc.

Key Issues

Which operating systems will survive and prosper during the 1990s?

How will software issues affect the market for microprocessor-based systems?

Strategic Planning Assumptions

RISC implementations will capture no more than 10 percent of the Windows NT
market through 1997 (0.8 probability).

Technical computing applications will dominate NT/RISC acceptance through 1997
(0.8 probability).

Most commercial application and tool ISVs will avoid support for RISC
implementations of Windows NT through 1997 (0.8 probability).

Microsoft Corp. has compelling reasons to offer Windows NT on RISC
microprocessors.  However, while NT is doing well on Intel processors, it has
not yet been successful on RISC, and we believe success will remain elusive -
even considering Microsoft's market power and motivation.

Dependence on a single vendor.  Microsoft does not want its future tied
directly to the fortunes of Intel Corp.  While happy with Intel (together they
have wrested computing leadership from IBM), Microsoft fears the ramifications
of depending on any single company.  Intel is a market leader and, hence, a
target.  Were it possible for IBM or others to crack Intel's desktop hegemony,
Microsoft would be exposed if its operating system was not portable among
microprocessors.


Note 1

ISV Porting Costs

Supporting multiple platforms in Windows NT has a lower porting cost for ISVs
than does multiple platform costs in Unix.  In Windows NT, the port is
typically nothing more than a re-compile.  However, the overall costs
associated with multiple platform support go well beyond just the cost of the
port itself.  Additional costs include staff training, quality assurance and
testing, support infrastructure, documentation, packaging, and channel
development.  Though some of these costs will be lower with Windows NT than
with Unix, ISVs are reticent and are generally waiting to see customer demand
before committing to alternative platform (RISC) support for NT.

Intel does not own the server market.  NT cannot succeed without winning
significant market share of commercial application servers - a market in which
Intel has done well, but has not dominated, as it has with desktop systems. 
Compaq Computer Corp. ships the most Intel-based servers. However, its image is
more as a PC and file server vendor than as a business-critical computing
systems vendor.  In the growing market for commercial Unix servers, no
Intel-based system achieved top-tier status. Intel's lack of market leadership
in this segment has little or nothing to do with the technological differences
between RISC and CISC processors; the failure of Intel-based server vendors
like AT&T Global Information Solutions (formerly NCR), Unisys Corp., ICL Inc.
and Sequent Computer Systems Inc. to rival the market share of Hewlett-Packard
Co., IBM and Sun Microsystems Inc. has furthered the market's incorrect
perception that Intel processors are a PC phenomenon, and are not suitable for
large enterprise tasks.  Microsoft is concerned that there are too few
credible, Intel-based, server vendors.

Why fight perception?  RISC delivers substantive floating-point performance
benefits for technical workstation applications, yet RISC's lead over Intel in
the integer-intensive commercial server applications market is small.  Intel
systems have often compensated for the deltas by exploiting the scalability of
SMP designs.  Better RISC SMP systems are emerging, but most commercial
computing problems (e.g., database serving, messaging, groupware and the like)
can be exceeded by the aggregate power of six Pentiums in an SMP configuration. 
Yet, the impression that RISC is necessary remains a market perception. 
Microsoft is concerned that Intel may not quickly reverse the perception -
hence its desire to legitimize NT/RISC.

Follow the ISVs.  Microsoft has lined up ISVs to port to Windows NT on Intel -
where they see a market opportunity.  Even though also supporting NT/RISC is
easier than porting among Unix variants, the majority of ISVs are not doing
this.  The ISVs are waiting for an NT/RISC market to develop before committing
resources (see Note 1) and, if it develops, they will wait to see which RISC
platform(s) emerges as a leader (see Note 2).  ISVs also hear little demand
from their customers and prospects for RISC support.  Despite Microsoft's
concerns, early NT customers seem happy with the Intel-based SMP server
vendors:  Compaq, AT&T and Sequent are seeding many NT trials and early
rollouts.


Note 2

Which RISC Platforms Will Win?

MIPS:  The NT/MIPS combination dates back to the days of the failed ACE
consortium.  The combination has little ISV support beyond CAD, and little
hardware manufacturer support beyond start-up NeTPower.

Alpha:  Digital began its NT/RISC effort with a strong market development push
for a broad set of ISV applications.  Despite its early commitment, Digital has
failed to attract the support of general-purpose desktop and server NT
applications for Alpha.  Digital is focusing its NT efforts on engineering
workstation applications, with which it is more likely to achieve a critical
mass of ISV applications.

SPARC:  The NT port for SPARC is led by Intergraph (Huntsville, Ala.), an
electronic design vendor moving away from the proprietary hardware business. 
Intergraph's well-regarded design applications are featured on the SPARC
platform (as well as Intel), but are not joined by many other ISV offerings. 
Sun is still belligerent toward NT.

PowerPC:  IBM's PowerPersonal division has looked toward NT as a growth vehicle
for its forthcoming RISC PCs.  IBM is institutionally incapable of expending
efforts to lure ISVs to the NT/PowerPC combination, because it spends its
energy on the same task for PowerPC/OS/2 and Intel/OS/2. Neither is going well
for IBM.

PA-RISC:  It is well known that HP has NT running in its labs on PA-RISC
systems.  HP is also aware that, without full ISV support, any PA-RISC/NT
combination would be doomed to market failure.  HP, whose Intel-based
superserver is gaining market momentum, may elect a RISC/Unix and Intel/ NT
strategy, avoiding the expense and uncertainty of NT/RISC.  The new HP/ Intel
alliance furthers this notion - ISVs can stay with Intel and get the benefits
of RISC when the architectures eventually merge.

The only application vendor that is fully committed to NT/RISC is Microsoft
itself.  Its SQL Server, Exchange Server (EMS) and Systems Management Server
(Hermes) teams are working on MIPS and Alpha versions concurrent with Intel
versions.  Though Microsoft will supply more of a complete software solution
environment over time, it will never be self- sufficient in the commercial
server market, and it will always need ISV middleware, tools and applications
to field a complete solution.

Users considering NT/RISC for application servers should anticipate, at best,
fewer choices of ISV offerings.  Though simple database servers can be deployed
initially without much supporting software, increased use will lead to needs
for more applications, tools and connectivity software; NT/RISC users will find
these scarce or unavailable.

Some system vendors market their platforms as ready for Unix today and NT in
the future, but, without committed ISVs, this claim is empty for RISC
platforms.  Hence, Digital's Alpha should not be viewed as a future mainstream
NT platform.  Alternatively, users can view AT&T's 34XX and 35XX systems as
Unix today and NT tomorrow, because they are based on an Intel SMP design -
thus assured of a full complement of NT software.


Glossary

   CISC Complex instruction set computer

   ISV Independent software vendor

   OS Operating system

   PC Personal computer

   RISC Reduced instruction set computer

   SMP Symmetric multiprocessing


For a reprint of this document, contact Jane Katzen at Gartner Group,
FAX: 617-229-2208.  Reprint prices are $1 (one dollar) each.  ALLOW THREE
WEEKS FOR DELIVERY.  Please include the following information.

@TITLE:94/06 GG-MCS  The Risk of NT/RISC
@SERVICE:GARTNER-MCS
@NUMBER:MCS Strategic Planning Assumptions, SPA-967-1500
@DISPDATE:June 29, 1994
@DATE:X940629
@AUTHOR:S. Winkler
@BEGIN:SUMMARY

This publication is published by Gartner Group, Inc. Reprints of this document
are available. Reprint prices are available upon request. Entire contents,
Copyright (C) 1994 Gartner Group, Inc. 56 Top Gallant Road, P.O. Box 10212,
Stamford, CT 06904-2212. Telephone: (203) 964-0096. Facsimile: (203) 324-7901.
This publication may not be reproduced in any form or by any electronic or
mechanical means including information storage and retrieval systems without
prior written permission. All rights reserved.
    

    re: .-1
    
    > "Microsoft fears the ramifications of depending on any single
    company."
    
    If that's not irony, I don't know what is.
    
    	Steve

If we now recognize that the road to market acceptance
of the Alpha on the desktop is trough NT and UNIX, why
do we turn potential customers off with a high price
tag?

Why not use our now streamlined PC operation to build
some really hot, low-end, workstations that will blow
the socks off SUN, HP, et al? Perhaps $13,800 is
competitive in that market, but I imagine that a $5,000
would be even more competitive. Or are we concerned that
a too competitive product will kill the margins our own
higher end workstations? 

>Per

> Or are we concerned that a too competitive product will kill the margins our
> own higher end workstations?

Yes, I think we are. However the Alpha "clone" makers don't care and will
give some competition here. Here is one sample:

Path: pa.dec.com!decuac.dec.com!haven.umd.edu!purdue!lerc.nasa.gov!magnus.acs.ohio-state.edu!math.ohio-state.edu!cs.utexas.edu!swrinde!emory!news-feed-1.peachnet.edu!news.duke.edu!godot.cc.duq.edu!newsfeed.pitt.edu!uunet!ycrdi!Steve_Lee
From: Steve_Lee@ycrdi.com (Steve Lee)
Reply-To: Steve_Lee@ycrdi.com
Newsgroups: comp.cad.pro-engineer
Distribution: world
Subject: Re: Pro/E on Unsupported Alpha Clones
Date: 06 Apr 1995 18:38:09 GMT
Message-ID: <3253010430.124530322@ycrdi.com>
Organization: Young Chang R & D Instititute
Lines: 23

We will be getting a Nekotech 289MHz in a couple of weeks.  If you would like
to, please send me mail around the end of April (just to remind me) and I
will give you a progress report.

I found that DEC was a lot more expensive (27K vs. 20K) for less performance,
compared to a Nekotech (or other clone).  There is a great article in last
month's Byte magazine issue that compares a lot of these clones and a DEC.

The first DEC reseller I talked to did not even know that clones existed!  He
spent a whole morning telling us how an Alpha gives you better performance
than an SGI at a better price.  But his quote was within ~$100 of an SGI
Indigo II.  (He did come back _weeks_ later with a better price, but it was
too late for him by then.)  So I went to another DEC reseller.  He could not
give me a good reason to pay more for a slower DEC Alpha over a faster
Nekotech.  I got a call from him a couple of days later, and guess who is now
a distributor for Nekotech!!  This DEC reseller!  So I ordered the Nekotech
through this dealer since he was local and would support both the machine and
Win NT.

Wish I had better data for you, but I should have some hands-on in a couple
of weeks.

Steve

    Something I have been grappling with.... can someone help ?
    
    We *can* make low cost Alpha NT / Unix systems. Trouble is, when we
    crank the price down, we also depress the $$ that someone running
    OpenVMS has to pay - 'same' hardware n'est ce pas ? Now, as far as
    I can tell VMS users tend to be quite happy with VMS. Due to our
    'advantage' of being able to run 3 O/S on the same tin, we reduce
    the absolute number of $$ we can extract (name of the game) from
    customers who are truely 'installed base'.
    
    Yes, I know customers expect declining prices constantly. yet, our
    'legacy' customers are getting their prices driven down even faster
    because:
    
    1) same tin
    2) need to compete (ours) using same tin with PC type systems for NT.
    
    
    
    Next thing: Price/performance is all well and good. What about 'Cost to
    make / Performance' ? This determines the margin that can be got from a
    box, given that 'Price' is usually determined by others (ie the
    'market'). Our manufacturing guys do strip downs of competitors H/W to
    estimate the cost to make for systems that are comparable (vis-a-vis
    Price / Performance) to systems we try to sell. Someone said to me that
    the best computer engineers these days are mechanical and packaging
    engineers, not semi-conductor types. I didn't understand what he meant
    until I saw the low end HP workstation taken apart - cheap as hell to
    make. I have a cup of strong coffee, can you smell it ?
    
    
    Rgds,
    
    AW

    
    We've lost a lot of deals because we are simply too expensive, why
    should the customer pay for more power then he needed.
    
    re PowerPC. Compaq and Zenith all talk about using PowerPC as a future
    server chip, nobody in this arena is talking about Alpha as I see it.
    
    Greg

    And Zenith Data Systems, Toshiba, TI, Vobis, Olivetti, Motorola, Radius
    and others have announced plans to offer Mac clones, thus expanding the
    acceptance of the MacOS and PowerPC even further.
    
    Mark.

PC Week has figured this out too:

see http://www.ziff.com/~pcweek/columns/cvn/cvn0403.html

some quotes from Mark Van Name and Bill Catchings:

In recent conversations, we've run across a few annoying bits of
conventional wisdom: RISC CPUs are cheaper and
faster than their CISC counterparts. Your future dream server will
run on RISC.

...

Here's an example of how rarely that notion seems to materialize. 
A couple of weeks ago we read a brochure from
DEC selling two Windows NT PCs, one with a 166MHz Alpha CPU and 
one with a 90MHz Pentium. The systems
otherwise held the same key components: 32M bytes of RAM, a 
1G-byte hard disk, a CD ROM drive, an Ethernet
adapter, a 17-inch monitor, and a Windows NT development 
environment.

Conventional wisdom told us the Alpha box would be the better 
bargain. Wrong. Its list price was nearly $4,000
more ($11,000) than that of the Pentium box. Even the "on-sale"
prices of the two systems left a gap of nearly $2,000.

...

Maybe, though, this example is not an indictment of RISC. Maybe 
it's just another case of Dinosaur Equipment Corp. getting things 
backward.

  "PC Week" probably has seen the Dataquest presentation too. After all,
  isn't that how new "conventional wisdom" gets manufactured? :-)

  By the way, yesterday's IDC* presentation in Hudson placed Apple
  at the very top of the home market shares, with something like
  16% or 17%. "IBM" was next, with 15%, but they downplayed IBM's
  real share as being 4 to 6 points because many people with
  "j random PC clone" would report to IDC's surveyors that it
  was "an IBM of some sort". I think Compaq was the next vendor
  with something under 5%. (Sorry, the notes are in my car.)

                                   Atlant


* IDC, International Data Corporation, is another well-known manufacturer
  of new conventional wisdoms (similar to Dataquest).

re: RISC vs CISC

I guess I'm not that surprised the performance gaps are narrower
than expected. When RISC became popular in Digital it was said to
gain performance by two methods:

    o  Fewer cycles per instruction (instructions simpler to implement)
    o  Consequential reduction in design time so that your architecture
       could be implemented on a newer technology.

The combination of the two was supposed to guarantee RISC a 2x performance
benefit over CISC. 

I think this all falls apart because each RISC chip we produce is
the most complex Digital has ever built. I.e. AXP chips have more
transistors than VAX chips. On the other hand Intel seems to get
impressive gains by just getting their existing chips to run
faster!

Does anyone whether we ever did achieve a substantial chip technology lead?

Cheers
Mark

re .2

>>> The worksystems group is selling the NT/Alpha boxes near their cost
>>> (don't forget the 20%+ reseller markup their "list" prices carry).

This statement is false for all NT/Alpha workstations that we are currently
selling. These systems have considerably higher margins than the margins
of typical PCBU products, and the average margin on NT/Alpha machines has
actually gone up in the past year.

The topic of Alpha/NT pricing and positioning is attracting considerable
management attention these days. Personally, I hope the VPs opt for lower
pricing across the board for the Alpha/NT machines. I think we'll be able
to sell a ton of 'em, despite their various remaining warts.

--RS

    The simplest indicator [imho] of how much margin there is in the Alpha
    products is the fact that the IEG markdown on "systems" (read:
    non-Intel boxes) is still 70% off List, while for PCs and everything
    else it's only 40% off.
    
    Sounds like there's still mucho fat in the server/workstation pricing
    model to me, I think...
    
    
    

    The topic of Alpha pricing has attracted attention on the internet
    newsgroup comp.sys.dec, with a couple of additional twists beyond
    Alpha-vs-Pentium and DEC-vs-reseller pricing:
    
       - Alpha/Unix vs Alpha/NT
       - Commercial Unix vs. free Unixes
    
    /john
    
Article: 19403
Newsgroups: comp.sys.dec,comp.unix.osf.osf1
From: katcher@netcom.com (Jeff Katcher)
Subject: Affordable Alphas AND Affordable Unix (A Clarification)
Organization: Me, Myself, and I
Date: Sun, 9 Apr 1995 17:34:08 GMT
Sender: katcher@netcom6.netcom.com
 
Last week I posted a query about reasonably priced Alpha systems.  For fear
of seeming more clueless than I like to think I am, I'd like to clarify and
amplify my question.
 
FACT: There are fast Alpha (21064A) systems available from the clone makers
and from Digital (discounted) for reasonable prices.
 
QUESTION: However, when the prospect of running Unix instead of NT is raised,
all prices increase 2 to 3 thousand dollars.  Why?  Does Digital Unix have
a future?  If Digital is weighting a system purchase with this kind of
a price difference, aren't they discouraging Unix/VMS purchases in favor
of NT?  I realize that large companies probably do not care, but small
developers such as myself really need to pinch the pennies.  Even SGI,
as bloated with hubris as they are, has a saner pricing policy.
 
I do not mean to criticize Digital or even Windows NT with this posting.
I do not think that NT is evil (I've used it a bit), but simply want a
very fast Unix platform for a reasonable price. 
 
 
Jeffrey Katcher
EMail: katcher@netcom.COM
    
--------
    
Article: 19409
Newsgroups: comp.sys.dec,comp.unix.osf.osf1
From: danpop@afsmail.cern.ch (Dan Pop)
Subject: Re: Affordable Alphas AND Affordable Unix (A Clarification)
Sender: news@news.cern.ch (USENET News System)
Organization: CERN European Lab for Particle Physics
Date: Sun, 9 Apr 1995 20:45:40 GMT
 
In <katcherD6s3Gw.Lv@netcom.com> katcher@netcom.com (Jeff Katcher) writes:
 
>FACT: There are fast Alpha (21064A) systems available from the clone makers
>and from Digital (discounted) for reasonable prices.
>
>QUESTION: However, when the prospect of running Unix instead of NT is raised,
>all prices increase 2 to 3 thousand dollars.  Why?  
 
Because NT is a Microsoft product, while DEC OSF/1 (recently renamed
Digital UNIX) is a DEC product.  Different companies have different 
pricing policies.
 
>Does Digital Unix have a future?  
 
I left my crystal ball home :-)
 
>If Digital is weighting a system purchase with this kind of
>a price difference, aren't they discouraging Unix/VMS purchases in favor
>of NT? 
 
DEC marketing doesn't have a particularly brilliant reputation.
>
>I do not mean to criticize Digital or even Windows NT with this posting.
>I do not think that NT is evil (I've used it a bit), but simply want a
>very fast Unix platform for a reasonable price. 
 
Have you considered running Linux or *BSD on a high end PC as a possible
solution to your problem?
 
Dan
--
Dan Pop
CERN, CN Division
Email: danpop@afsmail.cern.ch 
Mail:  CERN - PPE, Bat. 31 R-004, CH-1211 Geneve 23, Switzerland
    
----------
    
Article: 19411
From: peterm@alfven.eecs.berkeley.edu (Peter Mardahl)
Newsgroups: comp.sys.dec,comp.unix.osf.osf1
Subject: Re: Affordable Alphas AND Affordable Unix (A Clarification)
Date: 9 Apr 1995 21:35:15 GMT
Organization: Plasma Theory and Simulation Group
 
In article <katcherD6s3Gw.Lv@netcom.com>,
Jeff Katcher <katcher@netcom.com> wrote:
>
>QUESTION: However, when the prospect of running Unix instead of NT is raised,
>all prices increase 2 to 3 thousand dollars.  Why?  Does Digital Unix have
 
You are paying for the "DEC UNIX" licence.  It's quite a lot for an OS.
 
Have no fear, though.  I think the NetBSD and the Linux people are
working on a port to the Alphas.  Their product will be free, not
 ~ $3000.  But it probably won't be as stable and certainly won't
be supported.
 
 
>a future?  If Digital is weighting a system purchase with this kind of
>a price difference, aren't they discouraging Unix/VMS purchases in favor
 
It might be that DEC wants to get out of the OS business.  Hummm.  
Microsoft UNIX?  Scary thought.  I hope that someone just writes a
nice front-end to UNIX like xfm (x file manager) so that people
can achieve instant productivity.
 
Actually, I wonder why Microsoft hasn't done something similar to
this.  Windows could run on top of UNIX as well as it could run on
top of DOS.  In this way, people could take advantage of the great
body of free software out for UNIX as well as all the Windows applications.
 
>I do not think that NT is evil (I've used it a bit), but simply want a
>very fast Unix platform for a reasonable price. 
 
Someone else suggested a Pentium-box.  They're quite fast, and Linux is
free.
 
Regards,
 
PeterM
 
--------- 

Article: 19413
From: gendalia@buchanan07.res.iastate.edu (Tracy J. Di Marco)
Newsgroups: comp.sys.dec,comp.unix.osf.osf1
Subject: Re: Affordable Alphas AND Affordable Unix (A Clarification)
Date: 9 Apr 1995 16:55:13 -0500
Organization: who, me?
 
Dan Pop <danpop@afsmail.cern.ch> wrote:
}In <katcherD6s3Gw.Lv@netcom.com> katcher@netcom.com (Jeff Katcher) writes:
}>I do not mean to criticize Digital or even Windows NT with this posting.
}>I do not think that NT is evil (I've used it a bit), but simply want a
}>very fast Unix platform for a reasonable price. 
 
}Have you considered running Linux or *BSD on a high end PC as a possible
}solution to your problem?
 
NetBSD has a port to the alpha system, as does Linux, I believe.
There is more information on the NetBSD port available in:
ftp://ftp.iastate.edu/pub/netbsd/NetBSD-current/src/src/sys/arch/alpha/README
-- 
Tracy Di Marco                        gendalia@buchanan07.res.iastate.edu
'"Actually this is a common misconception...I do *not* in fact have a lot
  of time on my hands at all! I just have a very very very very bad sense
  of priorities."       --Dean Engelhardt'    -- David Sweeney
    
---------

Article: 19414
Newsgroups: comp.sys.dec,comp.unix.osf.osf1
From: katcher@netcom.com (Jeff Katcher)
Subject: Re: Affordable Alphas AND Affordable Unix (A Clarification)
Organization: Me, Myself, and I 
Date: Sun, 9 Apr 1995 22:50:12 GMT
Sender: katcher@netcom16.netcom.com
 
I realize that a suitably loaded PC is an alternative, but I've wasted
enough of my life fighting weird incompatibilities between PC components.
I've been spoiled by workstation packaging.  I buy the system, and it
works.  The problem with Linux and NetBSD is that they are not here for
the configurations that are interesting (if at all), and I have no time to
port them.  I could just by another SparcStation, but if I am forced to go
to Solaris 2.xyzzy, I might as well go to another Unix on a faster
platform.  I'm not complaining about the premium so much as about the
magnitude of it.  I'm not buying a 1000 user license, just a single
workstation.
 
 
Jeffrey Katcher
EMail: katcher@netcom.COM

-------------
    
Article: 19415
Newsgroups: comp.sys.dec,comp.unix.osf.osf1
From: katcher@netcom.com (Jeff Katcher)
Subject: Re: Affordable Alphas AND Affordable Unix (A Clarification)
Organization: Me, Myself and I 
Date: Sun, 9 Apr 1995 22:54:09 GMT
Sender: katcher@netcom16.netcom.com
 
The Linux release is not yet self-hosting on old hardware and NetBSD only
supports the old stuff as well.  If this has changed, I'd be happy to be
corrected. 
 
 
Jeffrey Katcher
EMail: katcher@netcom.COM

---------
    
Article: 19416
From: peterm@alfven.eecs.berkeley.edu (Peter Mardahl)
Newsgroups: comp.sys.dec,comp.unix.osf.osf1
Subject: Re: Affordable Alphas AND Affordable Unix (A Clarification)
Date: 10 Apr 1995 00:23:19 GMT
Organization: Plasma Theory and Simulation Group
 
 
 
In article <katcherD6sI3o.9B0@netcom.com>,
Jeff Katcher <katcher@netcom.com> wrote:
>I realize that a suitably loaded PC is an alternative, but I've wasted
>enough of my life fighting weird incompatibilities between PC components.
 
Unless prices on Alphas drop very seriously, I'm going to buy a PC and
turn it into a UNIX box, probably via Linux and XFree86.  Many people have
done this, and there are FAQ's on which hardware is supported, etc.
 
Some computer stores are even willing to install Linux FOR you and then
you do not have to worry about weird incompatibilities at all.
 
>I've been spoiled by workstation packaging.  I buy the system, and it
>works.  The problem with Linux and NetBSD is that they are not here for
>the configurations that are interesting (if at all), and I have no time to
 
This is true, I just checked on the state of the Alpha port of NetBSD,
and it's not there yet.  Linux I'm not sure of, but I suspect it is
not as far along as NetBSD.
 
 
>platform.  I'm not complaining about the premium so much as about the
>magnitude of it.  I'm not buying a 1000 user license, just a single
 
Er, the ~2-3k for DEC UNIX is purely for a 2-user license.  OSF-BASE.
Nasty, isn't it?
 
 
PeterM
 
    

I get a big kick out of the rather naive, cowboy attitude implicit in
these conversations - "just get a PC and put NetBSD or Linux on it. 
They're supposed to be porting it now, so it'll be out soon.  You won't
get support, but it's free"....all to save $2K- $3K.  And of course it'll
work right off the bat...and if it doesn't, it'll be obvious and a fix
will be available right away... The guy even *says* he doesn't want to
deal with the component incompatibilties of do-it-yourself systems, but is
willing to spend the time and energy putting together a one-of-a-kind box
with no O/S support to save two week's pay.

He'll spend twice that money on making the thing work in the first year -
time that he could be spending making more money for himself.  What a
waste of time and money.  It's depressing sometimes to think of what an
unprofessional and impractical attitude much of our 'technical' colleagues
have - except, of course, when I think of them as the competition...;-)

Who are these guys, college kids?  Oh, yeah, it looks like they are from
the SMTP mail addresses...figures.

It's far better to save time and spend a little more money - you can
always make more money, but you can never make more time.

tim

Of course, one underlying thing going on here is related to economies
of scale.  Windows development (including Windows/NT to the degree that
it uses Windows "artifacts") and applications get amortized over a MUCH
larger base than all Unix-derivates combined.  If all of the Unix vendors
with their partially proprietary source pools amortize the same amount
of development over 1/10th to 1/100th of the installed base, things are
going to cost more per unit.  (This isn't meant to imply that Digital's
higher prices are totally related to this factor.  However, over time,
this will be an increasing issue for the Unix vendors, or any other low-
volume software products by PC standards, like VMS.)

    Re: .43 Windows Nt vs UNIX volume
    
============================================================================
SUBJECT:  UNIFORUM HOSTS UNIX VS. WINDOWS NT ROW
SOURCE:   Network World via First! by Individual, Inc.
DATE:     March 26, 1995
INDEX:    [1]
----------------------------------------------------------------------------

  Network World via First! : Dallas Microsoft Corp. really is everywhere.
Even at last week's UniForum '95, a bastion for hard-core Unix fans, many
attendees were buzzing about the challenge Unix faces from Windows NT.

  Featured at the show were tools for helping users migrate from Unix to
Windows NT, and a host of speeches and seminars highlighted differences
between the operating systems and ways to make them coexist.

  Although Unix has been around much longer than Windows NT, sales of Unix
are only twice those of Windows NT, according to figures from International
Data Corp. in Framingham, Mass. Robert Enderle, an analyst with Dataquest,
Inc. in San Jose, Calif., predicts Windows NT will outsell Unix by six times
in 1998.

  It appears users are coming at the Unix vs. Windows NT issue in two ways:
There are those seeking to integrate the operating system environments on
their networks and those who want to go with one or the other.

  David Neill, manager of systems administration for Southwestern Bell in
Bethany, Okla., is a staunch Unix supporter, but uses both Windows NT and
Unix in his enterprise network. Windows NT is replacing DOS and Windows on
many desktops and is being used on servers to a limited degree. Neill spent
time at the show exploring tools to help manage networked Unix and Windows
NT systems.

  On the other hand, Dennis Geason, a systems analyst with Reno, Nev.,
mining company Placer Dome US, Inc., said his company is thinking about
migrating away from Unix to Windows NT because a far greater number of
applications are available for Windows 3.1 and Windows NT than Unix.

  Another reason for the switch is that most of Placer Dome's employees
prefer using personal computers to workstations, he said.

  FILE  AND  PRINT  PROWESS One of Windows NT's main selling points is it is
an application server platform that also ably handles file and print
services. While Unix is a strong application server, many Unix shops also
run NetWare to handle file and print services to compensate for Unix's
weakness in that area, said Patrick Higbie, chairman and chief executive
officer of DataFocus, Inc., a Fairfax, Va., software development firm
providing Unix-to-Windows NT migration tools.

  Having a single operating system to support both applications and file
services can greatly simplify desktop and server management, he added.

  Kevin Libert, director of high-end platforms in the worldwide server
marketing group at AT&T Global Information Solutions, agreed. While many
users are interested in both systems, many like Windows NT's ease of use, he
said.

  Another advantage is that Windows NT works with the large installed base
of PCs running Windows, providing users with a common desktop and server
operating system, Higbie said. This can save money on retraining end users
who change departments, he said.

  The Unix community, however, is fighting back. The Common Desktop
Environment (CDE), which debuted at the show, promises Unix workstation
users a common look and feel no matter what version of Unix they are
running, said Buzz Walker, program manager for open systems at Hewlett-
Packard Co. CDE is backed by major Unix vendors including HP, IBM, Novell,
Inc. and SunSoft, Inc.

  Even as Unix makes inroads with that common front, Windows NT is
attracting users because of the many applications already running on it,
including Microsoft's own BackOffice suite of management, messaging and
other products, said Michael Goulde, editor in chief for Patricia Seybold
Group, Inc.'s ``Open Information Systems'' newsletter.

  However, Windows NT lacks the maturity to handle mission-critical
applications, whereas Unix was built for just such programs, said Eric
Schmidt, chief technology officer for Sun Microsystems, Inc. in Mountain
View, Calif.

  Another of the most common criticisms about Windows NT is that it does not
scale as well as Unix. The operating system works reliably with up to four
processors, while Unix can handle up to 4,000 on a high-end Tandem
Computers, Inc. server, said Steven Jan, Tandem's product manager for future
systems.

[03-26-95 at 12:08 EST, Copyright 1995, Network World, File: x0326073.4dn]

    .42 has some statements about LInux that are not correct.
    
    You can get Linux on CD, from about a dozen sources, for about $20 to
    $30. It boots off the CD and runs. You can build a custom config from a
    simple menu choice. Updated versions appear on the net, and on new CDs
    every couple months. They work. Bug fixes are reported on the net, and
    patches are provided much faster than any software from commercail
    vendors. 
    
    And these are not just home-use toys, see THE LINUX JOURNAL for
    application stories of commercial uses, including Utility companies and
    a couple hospitals. 
    
    Have a look at Linux, it is an amazing example of the power of the Net
    and cooperative software development. And a lot less than we charge for
    Unix.

	And Linux for Alpha is coming or is already here.

							mike

But who'd nickel and dime with a $20 Linux when you've just coughed
up $13,800 for the system to run it on? BTW, didn't someone use to offer
mahogany enclosed VT100s back in the old days? Maybe we should give 'em
a call. Perhaps it would appeal to those who are willing to go that
extra mile for a true status symbol?

>Per

    
    You can buy support on Linux too. I know of a few relativly large Swedish 
    companies that does. As the company delivering the support is quite small 
    (9 employees) I don't think the user groups at the buying end are very big,
    but they do exist. 
    
    So even major companies seems to accept to use 'free' software for some 
    business purposes. Infact it was an article about it in this weeks 
    "Computer Sweden". But I suspect that it's usually not a first option to 
    use 'free' software for business.
    
    /Kjell

re Note 3791.44 by OSL09::OLAV:

>   However, Windows NT lacks the maturity to handle mission-critical
> applications, whereas Unix was built for just such programs, 

        Really?  Unix was built for mission-critical applications?

        Bob

        (Please note, I am talking origins, not "Digital Unix".)

>>   However, Windows NT lacks the maturity to handle mission-critical
>> applications, whereas Unix was built for just such programs, 
>
>        Really?  Unix was built for mission-critical applications?
    
    You forgot to include: "said Eric Schmidt, chief technology officer for
    Sun Microsystems, Inc. in Mountain View, Calif."
    
    This is FUD from the UNIX camp. Windows NT was built to be a reliable
    operating system. Old software isn't neccessarily "mature" and bug
    free. I think Windows NT will "mature" more in two years than UNIX has
    done in twenty.
    
    Olav

>>        Really?  Unix was built for mission-critical applications?

Absolutely.  The "mission" was document preparation at Bell Labs,
and I'm sure the implementors thought it was quite critical.

    That's stretching it a bit. The "mission" was to play (darn, I used to
    know this, it's something like Lunar Lander) and create an interesting
    operating system where they could host some of their research projects 
    to make up for the loss of access to Multics. Document prep was the first 
    real world application they hit on for it but I don't think any of the 
    players would suggest they had it in mind when they started. 

  [The mission was to play...] "Spacewar", I believe.  That program
  with two spacships nominally orbiting around a sun_with_gravity.

                                   Atlant

    Since all early unix systems had teletypes for terminals, I doubt
    anyone played spacewar on them. Spacewar required a graphics device
    like a GT-40, or PDP-12.
    
    Many unix 'features' can be traced to the early use of tty's. On an
    ASR-33 you really didnt want to be typing long commands and file names.

    That's right, spacewar. The original, assembly language V1 UNIX ran on
    a PDP-8 that had a VDT of some sort. It was a single user operating
    system (hence the name UNIX, a pun on Multics).

PDP-6, wasn't it?

		Steve

Split the difference

    DECUS, but you probably knew that.
    
    			Steve

    Ratholes-R-us

    The list prices for our systems are list prices.  Our distributors and
    resellers have significant margin to work with and discount.  However,
    many of the channels have been hurt by the brutal competition the past
    few years, so they are not likely to discount much now for the kinds
    of sales that we're likely to see reported here.
    
    Another factor that leads to the high list prices besides providing the
    high margins are the need to provide good margins for the clone makers
    while they offer significant discounts.  Apple is facing this issue
    as the first clone comes out.  Will Apple allow a clone vendor to price
    systems 20% less than Apple's own pricing?  Sun wouldn't allow this to
    happen with the SPARC clones so that today there are only niche
    markets.
    
    And the pricing for OpenVMS and Digital UNIX reflects the availability
    of TurboLaser.  If you don't need the scalability and features offered
    by all the systems from the desktop to a turbolaser cluster, then buy
    Windows NT.  If you don't need someone to help you configure the
    system, select the software, then buy Microsoft Backoffice or
    Quickbooks, but if they don't solve your business problem and you need
    SAP or SAS, then expect to pay for all the time that some sales person
    will spend with you, and for his training and experience.

From:   3049::"agl@redline.ru" "Anthony Graphics"
To:     xdelta::hoffman

In article <3n5le3$o79@jac.zko.dec.com> you wrote:

  <snip>

:    ... Suggestions or specific complaints are welcome.  (Via
:    e-mail please.)  I will forward these along to the internal
:    news groups or, when appropriate, to specific individuals.

Suggestion: OSF/1 with no support at all but at NT price :~-(
People do not want NT on their workplaces but OSF is a bit expensive I
guess. You know how they are making out of this situation?
Right, just using PCs running FreeBSD and Linux as a workstation
instead of Alphas: the guys who hate bill gates won't buy even cray loaded
with windoze :-I Recent discussion in the freebsd mailing lists shows just
that.
AGL