Conference decwet::windows-nt

	This has already been answered in another notesfile.

	In future, please note what other notesfiles you have
	cross-posted in.

							mike

    It was answered but the cross-posted notesfile is a closed conference.
    
    The jist of the response was that this could be a network card
    bottleneck.
    
    Gary
    

    You can read details on how a 2100 got clobbered by a PPro system
    with Serverbench in the 6/19/96 issue of PC Week, too.

    
    For the benefit of other and future readers .. I have included my
    reply (with a few notes[] added) from the other conference where this 
    was originally posted.
    
    p.s. check out www.enorex.com for good example of 3rd party vendors
    now selling 500Mhz Alpha's at same prices as some PPRO's
    
    :-)
    
    / Kerry
    -----------------------------------------------------------------------
    
    Since it is likely that we (Digital) have limited expertise in this PC
    mag's benchmark, it is very likely that they have little experience in
    optimizing their PC benchmark for the Alpha platform. I may be wrong
    (perhaps someone can jump in here if they have additional info) here,
    but was Digital involved in the port of this benchmark to Alpha?
    
    Given that entire benchmarks can be swayed by as little as 1 or 2
    parameters in the applic and/or OS, I would suggest taking a closer
    look at what this benchmark was doing and where it was peaking out..
    
    Also, keep in mind that these benchmarks are all based on a 32bit OS
    (NT) with no (or very few) 64bit optimzations in the OS for the Alpha
    platform.
    
    If this Customer plans to use SQL Server or Oracle on NT, then remember
    that all current Intel systems (including MMX) will NOT be able to
    take advantage of the planned 64bit optimizations in NT5.0 (now in
    early beta). P7 will be available later in '98 depending on system
    availability.
    
    For those who think Oracle and Microsoft do not consider this 64bit
    issue to be a big thing, reference:
    
    http://www.pcweek.com/news/1209/09ent.html
    http://www.microsoft.com/corpinfo/press/1997/Feb97/PowerPr.htm
    http://www.microsoft.com/corpinfo/press/1996/jun96/nt64btpr.htm
    
    For non DIGITAL view of 64bit importance, reference:
    http://www.digital.com:80/alphaserver/news/64abst.html and
    http://web133.bbnplanet.com/cs116f5.htm
    
    For WS specific issues, see PC Week review of latest 433a WS as
    compared to PPRO : http://www.pcweek.com/reviews/0127/27alpha.html
    [remember that this review was before recent huge price drops on 
    Alpha platforms]
    
    Regards,                                             
    
    / Kerry
     
    
    

    As a follow on to .4 reply, it would appear that the Intel P7 chip has
    been delayed for quite awhile. This means that the the Alpha
    competition will likely not be the 600-750Mhz Alpha (later this year),
    but rather 1Ghz that will likely be here in 1999-2000 timeframe.
    
    Reference:
    http://techweb.cmp.com/iw/newsflash/nf618/0224_st3.htm
    
    Interesting note is that 64bit NT code is being done on Alpha and will
    be ported to P7 when these systems finally do become available. Now,
    assuming DIGITAL maintains it's Alpha pricing at PPRO prices policies
    and 64bit NT on Alpha has been in place for 12-18 months (time to
    heavily optimize etc) with associated applications such as SQL Server, 
    Oracle etc and assuming that the 850-1Ghz Alpha's are the competition 
    for P7 at it's time of release, then I would say that the P7 folks 
    have their work cut out for them ... 
    
    [DB servers with 64bit NT are going to love Alpha - especially with
    memory prices dropping through the floor...]
    
    :-)
    
    Should be interesting ..
    
    / Kerry

>>    As a follow on to .4 reply, it would appear that the Intel P7 chip has
>>    been delayed for quite awhile. This means that the the Alpha
>>    competition will likely not be the 600-750Mhz Alpha (later this year),
>>    but rather 1Ghz that will likely be here in 1999-2000 timeframe.
    
	At the risk of being a wet blanket... The article did say that
	the delay was not for technical reasons but for marketing reasons.
	Perhaps this is spin control on their part to cover technical
	problems - who knows.  But let's assume it's true.  If they started
	to lose real market share to us in the VLM space - assuming we 
	see it start to mature quickly - I suspect that their marketing
	would change it's mind very quickly.

>>    Interesting note is that 64bit NT code is being done on Alpha and will
>>    be ported to P7 when these systems finally do become available. ...

	The real question to ask here is: "If there were no P7 on the
	horizon, would VLM competition to SQL Server alone be enough of
	an incentive for MS to develop the WIN64 API on Alpha?"

	I would like to think so.

>>                                                                ... Now,
>>    assuming DIGITAL maintains it's Alpha pricing at PPRO prices policies
>>    and 64bit NT on Alpha has been in place for 12-18 months (time to
>>    heavily optimize etc) with associated applications such as SQL Server, 
>>    Oracle etc and assuming that the 850-1Ghz Alpha's are the competition 
>>    for P7 at it's time of release, then I would say that the P7 folks 
>>    have their work cut out for them ... 
    
>>    [DB servers with 64bit NT are going to love Alpha - especially with
>    memory prices dropping through the floor...]
    
	Let's hope this much is true!

	- norm

    >>
    At the risk of being a wet blanket... The article did say that
    the delay was not for technical reasons but for marketing reasons.
    >>
    
    Interesting approach to marketing ie. sell the positive, de-emphasize
    the negative. Our approach to marketing would likely have been
    something like "we are sorry, but we have run into technical problems
    and for this reason, our P7 systems will be delayed at least 18months."
    
    >>
	The real question to ask here is: "If there were no P7 on the
	horizon, would VLM competition to SQL Server alone be enough of
	an incentive for MS to develop the WIN64 API on Alpha?"
    >>
    
    Keep in mind that Microsoft wants 64bit just as much (possibly more)
    as we do - they want to get at the TP market (online transaction
    processing for large Corp db's and over Internet etc) and they will not
    be able to do it with a db that can only do 7k TPS. The big guys are
    in the 20-30K range (our top 30k number was done with old old 350Mhz
    system cluster). With EV6 due soon, and memory channel technologies 
    these 64bit numbers will likely be in 35-40K range by end of year.
    
    Also, while the initial intent of the 64bit goodies are VLM, I 
    strongly suspect that these 64bit optimizations in NT will allow
    better performance in other area's as well.
    
    Reference:
    http://web133.bbnplanet.com/cs116f5.htm
    
    With Alpha and Intel the only NT platforms now, DIGITAL is now
    positioned to take a leadership role in that we can supply both
    platforms. IMO, with Alpha prices now in range of PPRO's, sales 
    are going to go through the roof.
    
    Regards,
    
    / Kerry
    

    Not sure where 64bit NT got into this note, but it won't help the
    base noter's 128Mb system one bit (64 bit NT API's require that
    the physical memory be there; no page fault mechanism).  Most
    Ziff reviews that use Serverbench, like the 6/17/96 PCWeek review
    where Alpha got clobbered, won't have >2Gb machines either.  For
    the vast majority of server sales (i.e. <2Gb), we need to show Alpha
    outperforming equivalently priced Pentium Pro NT servers.
    .02 K