[KLUG Members] Re: Yo! -- Such consumers made Microsoft

[Bryan J. Smith]

--=-Ze7amJygv3jPEZlMILFc
Content-Type: text/plain
Content-Transfer-Encoding: quoted-printable

On Fri, 2002-11-29 at 14:11, Adam Tauno Williams wrote:
> Which means it is slower running the software people use, which means it =
is
> slower.  Benchmarks belong in the same catagory as claims made on informe=
rcials.

The "problem" nowdays is that AMD has a 30% marketshare.  Furthermore,
the P4 is now incompatible with some software because of the opcode
reuse issue.  Intel keeps adding more and more SIMD instructions out of
100% pure marketing.

Thank God AMD is "taking control" of x86 with x86-64.  Most people in
the PC OEM and developer worlds agree.

> But there isn't one.  Hence AMD is slower.

Well, the K6 _was_.  But that chip is long gone.  _That_ was my main
point.

The Athlon is _far_ different.  It's core is pretty much superior in
every aspect to the P4, as the P3 was far better than the P4 as well.=20
But unlike the P3, which was designed for 200MHz+, and "topped out" at
1.4GHz, the Athlon is designed for 600MHz+, and will "top out" at around
4GHz.

Sure, the P4 will scale to somewhere between 5-10GHz, but it's an
ultra-simplistic core with a lot of design issues.  Hence the 33% slower
ALU performance than the P3/Athlon, MHz for MHz, and over a 50% slower
FPU, MHz for MHz.  Again, why did Intel do this?  Simply because their
engineering resources have been tapped to IA-64, and they didn't think
IA-32 (aka x86) would be around in 2002.

Building a powerful core takes 3-5 years typically.  Intel "threw the P4
together" in about 18 months.  Quite a feat, but the core is pretty
ineffective.  But they made the pipes deep enough, and they enjoy a 6-12
month fabrication technology lead over AMD.  But that still doesn't make
up for the die size differences completely -- AMD still leads there.

Not surprisingly, because the majority of current PC technology, from
system interconnects to memory to peripherial controllers came from
Digital.  Most Digital engineers made a mass exodus to AMD years ago.=20
And various technologies from AMD have resulted.

> I "know" this, as lots of people have told me.  And I don't much care.   =
It's
> about DELIVERED performance.

DELIVERED performance in _Linux_ is quite different than Windows.=20
There, the "playing field" is leveled.  Even for servers -- actually,
especially for web servers where there is mix of dynamic content
creation and I/O.

Besides, many developers are finding it "difficult" to "keep up" with
all the new instructions Intel is adding.  For what?  Many times some
new instructions and the new Intel optimizing compilers _reduce_
performance.

I don't run Athlon for nothing.  And engineering firms don't run it for
nothing either.  Linux users should look at actual performance of the
Athlon, it's not like on Windows where everything is Pentium-optimized.

Although I _do_ find it humor that many "Pentium optimized" binaries for
Windows run _faster_ on the Athlon than the P4.  Again, Intel threw the
P4 together in about 18 months.  Add in the various 386/486 code
incompatibilities now in the P4, and it's not pretty.

> 512Mb
> Blue Curve uses hundreds of megs more of RAM than Crux?!

I didn't say that.  (I don't know personally)

> Huh, that must be why they still sell so many.

Marketing !=3D Engineering

IA-64 has taken far longer to "mature."  Intel canNOT "turn back the
clock" to 1993 and go, "oh wait, we probably need to look at IA-32
lasting until 2005."  It's too late.  Intel realized that IA-64 wasn't
going to be ready in 1998/1999, and had to do what they could.

A simple pipe extension, and a whole slew of SIMD instructions with most
expensive advertising campaign outside of Redmond is what we got.  And
it worked!  People like yourself are sold!

> Ask people who have tried Asterisk or other Codec related packages on
> AMD and watched the machine puke up its guts.

Yes, Intel has been _very_successful_ at _purposely_ re-using opcodes
from the 386/486.  Thus, they have rendered the Athlon incompatible
until a new version comes out with new microcode at the expense of
386/486 code compatibility.

All for what?  Maybe 2-3% performance increases?  Sometimes they are
_not_!  And they are "decreases"!  Intel is inventing new SIMD
instructions out of _pure_marketing_necessity_!

> Stick in an Intel CPU and everything is peachy.

Not!

> If AMD claims to sell x86 compatible CPUs they should make one.

Athlons are _more_ x86 compatible than P4s now!  The problem is that AMD
is at the mercy of Intel's marketing campaign that sells these SIMD
instructions _purposely_incompatible_ with existing 386/486 code!

As a Linux user, you should _understand_ what the problem is here!

> Laughs, and then files for bankruptcy.

You think this is funny?  If AMD goes, Intel's pricing will skyrocket
dude!  Man oh man, you're sick!

> I certainly never thought K6 =3D P4, or even PPro.  I WANT TO DRAW
> BORDERS AROUND THE WINDOWS.

And you have thrown a Pentium in the _same_ system and saw no issues?

> Yes, I'll bet Intel is frightened.

Yes, they are losing major science/engineering acounts that is costing
them a significant chuck of their revenue.

> Understand?  I've never read an announcement from Cray.  I don't care abo=
ut
> Cray.  I don't own one, use one, and probably never will.

Nevermind dude.

> And I originally was an angora goat rancher, then an auto-mechanic, then =
an
> electrical technician.  Computer guys get better benefits, a clean (excep=
t my
> desk) work environment, and climate control!  But I don't pay any attenti=
on to
> (a)what the textile composite of my clothes is, (b)what kind/size of engi=
ne my
> car has, or (c) what ICs my home stereo uses.  I want them to (a) hide my=
 pasty
> Finnish skin, (b) go forward when I press the accelerator, and (c) drown =
out
> Celine Dion when my next door neighbors start spinnin' the divas.

And when everything costs 3x more and runs slower than crap, you won't
care either, eh?

The "clueless consumer" is what created Microsoft's monopoly.  Watch AMD
die and see what happens with Intel.

AMD continues to drive x86 innovation, Intel continues to drive x86
marketing.  I'm not saying buy AMD instead of Intel because of that.=20
Just don't not buy AMD because of ass-u-mptions that are _not_ true!

And that's what this whole thread was about!  It was about your
_ignorant_ ass-u-mptions of AMD that were _not_ true.  Buy Intel if you
like, even I do for some things.  But don't spread BS about AMD that
simply isn't true.

So go ahead and soak in your ignorance.  Spread falsehoods about AMD,
and don't give a rats @$$ about technical accuracy.  That's the #1 thing
that pisses me off.  Which is what I don't like about Microsoft either.


--=20
Bryan J. Smith, E.I.            Contact Info:  http://thebs.org
A+/i-Net+/Linux+/Network+/Server+ CCNA CIWA CNA SCSA/SCWSE/SCNA
---------------------------------------------------------------
The more government chooses for you, the less freedom you have.

--=-Ze7amJygv3jPEZlMILFc
Content-Type: application/pgp-signature; name=signature.asc
Content-Description: This is a digitally signed message part

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQA95+CZDjEszaVrzmQRAsoRAKDD1DHVgZpSJTcf+HVj2oWLgvO/7ACg1ZPc
1qBTfoa0MtYdXgdCnbLMjp8=
=TJfv
-----END PGP SIGNATURE-----

--=-Ze7amJygv3jPEZlMILFc--