[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference ricks::dechips

Title:Hudson VLSI
Notice:For Digital Chip Data - CHIPBZ::PRODUCTION$:[DS_INFO...]
Moderator:RICKS::PHIPPS
Created:Wed Feb 12 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:701
Total number of notes:4658

642.0. "Request for Alpha/Intel comparison information" by GLDX02::ALLBERY (Jim) Wed Jan 29 1997 13:15

    I'm preparing an internal course that will include segments on 
    architecting solutions under Windows-NT.  I'd like to include a
    segment on choosing between Alpha and Intel.  Any pointers to 
    current information on the relative positioning of our Alpha and 
    Intel products would be most helpful.
    
    To put it another way...  I think I can handle the arguments for
    why you'd pick Intel-- what are the arguments for choosing Alpha?
    What are specific applications where Alpha excels?
    
    Thanks in advance,
    Jim
T.RTitleUserPersonal
Name
DateLines
642.1If performance matters, buy AlphaWRKSYS::SCHUMANNWed Jan 29 1997 14:4215
o  Fast floating point. FP is much faster on Alpha than on Intel.
   This typically helps on modelling apps, such as spice, weather forecasting,
   molecular modelling, EMI/RFI modelling. It also helps on 3D graphics apps,
   because the graphics pipeline requires floating point for rotating and
   positioning objects in 3d. 

o  Good memory bandwidth. 433a/500a memory bandwidth is much better than 
   200i.

o  Best overall performance. If you app ever does anything that takes time
   to complete, it will almost always be faster on the Alpha, assuming an
   Alpha native version of the app is available. If you like to twiddle your
   thumbs, buy Intel!

--RS
642.2PCBUOA::KRATZWed Jan 29 1997 15:361
    don't forget to mention 64 bit NT coming on Alpha first.
642.3.1 Numbers PleaseXFMV01::AJITKThu Jan 30 1997 02:166
ref.1 

If you could quantify the points in your note, it 
will help. 

Ajit
642.464 bit NTRDGENG::WILLIAMS_AThu Jan 30 1997 06:553
    re 64 bit NT, first on Alpha. Is this to be 'real' 64 bit, or WIN32
    APIs with some horrible 64 bit cludge nailed on ? Curious, and
    customers asking too. [wrong conference maybe ?]
642.5Quantification is in progressPERFOM::HENNINGThu Jan 30 1997 07:1425
    Re: memory bandwidth - We're working on quantifying the memory
    bandwidth into "publishable" form.  But Reinhard has run a variant of
    the same benchmark and the final results will certainly confirm his
    point.
    
    If you want the preliminary numbers drop either of us an email off line.
    
    Re: floating point: I think we (SPEC itself + Digital) are still not
    ready to publish NT SPEC numbers.  But the decision to publish might be
    accelerated a little  if we separated the publication of SPECfp95 from
    SPECint95.  The former is running on NT a whole lot like it runs on
    Unix; the latter is still doing a bit of porting and tuning.  If you
    think that separating the two is a good idea, drop us a line.
    
    For a really loooooooose estimate, take a Unix 400 MHz number for
    SPECfp95 and suggest to your customer that the 433a will not be
    disappointing once we do publish NT SPECfp95.
    
    The best Intel SPECfp95 number at the SPEC website is 6.75; Alpha 400
    Mhz SPECfp95 results are 14.1 (AlphaStation 500/400) and 17.2
    (AlphaServer 4100 5/400).  The SPEC web site:
    
    	http://www.specbench.org/
    
    /john
642.6Re: Request for Alpha/Intel comparison informationQUABBI::"everhart@star.enet.dec.com"Glenn C. EverhartThu Jan 30 1997 12:5010
What folks tell me from the MS conferences is the 64 bit APIs will be
very limited, Alpha only. That sounds to me like enough to make marketing
statements about 64-bitness in NT (they after all get competition from
systems that DO have full 64 bit functionality). When a 64 bit Intel
followon comes, though, this threadbare state of the Emperor's wardrobe
may prove too much of an offense to their sensibilities...

(Mind too, there may be futures they have that I have not heard.)
[posted by Notes-News gateway]
642.7DECCXL::OUELLETTEThu Jan 30 1997 15:554
The 64 bit NT work is more than marketing hot air,
but you are right that the API is minimal.
If you need to know the details, you should
contact DEC West for an athoritative answer.
642.8MPOS01::nqsrv201.nqo.dec.com::CerlingCall on the CalvaryThu Jan 30 1997 21:406
	DEC West has actually put together a paper (internal use only) on
	what to expect.  I think there is a pointer in the NT notes 
	conference.

tgc
642.9Alpha/Intel Design Comparison?PHXSS1::MEAGHERCube with a view of the blues ..Thu Feb 27 1997 12:4623
    Greetings - 
    
    I've tried all the title searches I can find and this is the
    closest note to my concern.  I've also 
    *scoured* the Intranet to find this paper [that I read once in 
    hardcopy] written by one of our Engineers [of this I'm sure] that
    compared our Alpha design philosophy with that of Intel.  He talked 
    about superscalar, superpipelined approach with Alpha and why our MHz 
    ratings are higher than Intel -- what that really means.  I'm sure he
    went into floating point performance, etc.  The paper was about 4-6 
    pages, if I remember, and I can't remember whether it was .PS or
    .DOC or what -- I do remember it was in very small print -- but
    I might have printed it out that way somehow.
    
    Now, from the looks of this conference, some of you could probably 
    address the topic straight out, but I'm not looking to take up that
    much of your time.  If you know the paper I'm speaking of, it will
    come to you.
    
    Thanks in advance,
    
    Kent
    SBU Sales Support, PHX
642.10CAMPY::ADEYIs there a 'Life for Dummies'?Thu Feb 27 1997 14:505
    re: Note 642.9 by PHXSS1::MEAGHER
    
    It was probably something written by Richard Sites. Cantact him.
    
    Ken...
642.11Might be difficultKAMPUS::NEIDECKEREUROMEDIA: Distributed Multimedia ArchivesFri Feb 28 1997 06:511
    Richard Sites is no longer with Digital (he went to Adobe).
642.12dsites@adobe.comPHXSS1::MEAGHERCube with a view of the blues ..Fri Feb 28 1997 22:093
    OK, I got his address at Adobe.  Thanks.
    
    /Kent
642.13Why did Sites leave?COPCLU::TRIERJens Trier Rasmussen @ DMO 857-2264Sat Mar 01 1997 12:173
    Anyone knows why Sites left?
    
    - Jens
642.14Did you really expect an answer ?KAMPUS::NEIDECKEREUROMEDIA: Distributed Multimedia ArchivesThu Mar 06 1997 14:082
    Sure, but that certainly isn't a viable topic for a public notes
    conference.
642.15MSE1::PCOTEJust say yes to 'thin clients'Thu Mar 06 1997 15:3512

>    Anyone knows why Sites left?

     

    TFSOed.

    
    Excuse the sarcasm. Yet, it seems that digital's biggest 'export'
    is our indispensable talent pool.

642.16if we only had Photoshop...PCBUOA::KRATZThu Mar 06 1997 16:256
    Not to mention that a native Alpha version in Photoshop V4 would have
    been a serious coup for DIGITAL; perhaps the single biggest opportunity
    to substantially increase market share in the hi-end commercial
    desktop.
    .02 K
    
642.17SAYER::ELMORESteve elmore@mail.dec.com 4123645893Wed Mar 12 1997 16:371
    When did Richard Sites leave?
642.18SMURF::DENHAMDigital UNIX KernelThu Mar 13 1997 00:221
    Last fall seems familiar. October-ish?
642.19limited info on VLM for NT5.0MSE1::PCOTEJust say yes to 'thin clients'Thu Mar 20 1997 15:3813




>	DEC West has actually put together a paper (internal use only) on
>	what to expect.  I think there is a pointer in the NT notes 
>	conference.

     ftp://ntnone.zso.dec.com/doc/NT-64bit/nt64white.doc
     (albeit somewhat terse)

     This is an internal only document!