[Beowulf] [craig.hunter at nasa.gov: Re: Intel?]

Robert G. Brown rgb at phy.duke.edu
Wed Jun 8 09:46:09 PDT 2005


> > I was just wondering what the general view was about the move to Intel?
> > 
> >  From my own point of view, it will be a very good thing. It will
> > (hopefully) mean that all the lovely engineering apps. from windows can
> > now run at full (or nearly full) speed on my mac, assuming of course
> > that Microsoft get VPC working well.
> > 
> > Thank you very much Mr. Jobs.
> > 
> > The problem for Apple is that now everyone will wait for Intel Macs
> > before buying a new one. I know I probably will.
> > 
> > It occurred to me that rather than being more of a rival to windows, it

...

> current and near-future PPC systems seem like lame ducks.  Instead, they
> announced a very dramatic shift which sent out a huge ripple.  It just
> doesn't make sense to me, so I think there must be a lot more going on
> behind the scenes that we don't know about (Apple/IBM politics).  I just
> can't think of a rational reason why they would make such a dramatic
> 
> * lots of time/money spent on AltiVec tuning which is going to be wasted.
> Yeah, Intel has MMX/SSE/SSE2, but if you believe the claims Apple made over
> the years, AltiVec was always supposed to be better.
> 
> * what multi-processor 64-bit CPUs does Intel offer that are a compelling
> alternative to the G5 for desktop and cluster systems, in terms of cost,
> power, efficiency, heat, etc?  (keep in mind that I continue to think the P4
> is a great 32-bit CPU).  Again, Apple has always touted the superiority of
> the G5, so this represents a pretty big shift in their posture.

....

> * On a related note, scientific developers and users are different than
> commercial developers in the sense that we're not driven by sales.  When I
> spend a week porting a code and tuning it for G5, it's a major investment in
> my job and my research capability, and the idea is that the software
> development becomes a tool for long term use.  With tight budgets and
> limited time, switching platforms/CPUs and supporting other platforms/CPUs
> is not easily justified.
> 
> * the issue of software compatibility with x86/Windows is speculation at
> this point; we don't know if the Mac switch to Intel will be a good or bad
> thing for software and apps.  I hope to see more commentary on this from
> experts in the field.

Just a quick remark -- I think that all of this has to be viewed in the
light of the major sea-change that is underway in COTS computer design.
Starting this year, NOBODY'S computers are going to be the same anyway.

I speak of:

  a) 64 bits -- the P4 and all its friends are dead as a doornail,
although it will take a couple of years to die all the way and although
speciality markets e.g. laptops will likely linger until they can cool
off the 64 bit entries.

  b) Multicores.  There are three very, very different approaches to
multicore CPUs (which I predict will totally dominate HPC starting
basically "now").  Intel's on top of Hyperthreading, good for
multitasking and HA, not necessarily so good for HPC (although Intel
will argue otherwise and of course COULD be right -- we'll see:-).
AMD's on top of HyperTransport, which is basically replacing bus and
bridge with a switched network on the motherboard, AFAICT.  This looks
to be GREAT for HPC in the short run -- very high total bandwidths and
quite short latencies on top of DDR-2 -- but again the proof is in the
benchmarking and application timing, which will be showing up on list
Real Soon Now.  IBM's CELL, which is to put a single core GP CPU and
surround it by a cluster of pipeline processors (AFAICT there -- it
isn't too easy to really see what or how it will all work yet).

  c) Software maintenance.  Like it or not, it is expensive and a PITA
to maintain multiple versions of applications, each tuned for a
DIFFERENT vector paradigm for suitable parts of the application space.
If you don't reach and maintain a critical mass there, you get eaten
alive my maintenance costs and porting costs with an every shrinking
market.

So like SGI before them (and many others) Apple is simply reading the
writing on the wall and making a move now before they are completely
marginalized.

I'm actually impressed.  Apple has ALWAYS been a software company
masquerading as a hardware company, and I've been absolutely astounded
that they've survived this long while maintaining this illusion.  There
are only a few others like them surviving -- e.g. Sun -- and Sun was
never mainstream PC.

This just makes it official.  Apple will from now on be in direct
competition with Microsoft, unless they are completely daft and make
enough architectural changes in x86 that their software isn't portable
to vanilla box PCs.

This is a smart move.  For a while, they may even appear to win.
However, they CAN'T compete with linux and freebsd -- nothing can.  They
have to make a LOT of money to keep shareholders happy, where the linux
folks are happy to make money period, with salad days coming inexorably
in the future...

Editorially yours,

   rgb

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu





More information about the Beowulf mailing list