Re: AMD's v. Intel

["Bryan J. Smith" <b.j.smith@ieee.org>]

On Tue, 2005-07-12 at 11:15 -0500, dsavage@peaknet.net wrote:
> For those with time for a fascinating read, here's the original text of
> AMD's lawsuit against Intel alleging anti-trust violations:
> http://www.amd.com/us-en/assets/content_type/DownloadableAssets/AMD-Intel_Full_Complaint.pdf

#48 and #64 are clearly a _big_deal_ in this case.  It's one thing to
claim a competitor is preventing you from attaining potential sales.
It's completely another when an OEM, like HP, has people screaming for
products, which is the case with AMD Opterons.  HP co-designed the
Itanium with Intel, with both PA-RISC and x86 compatibility.  What has
HP switched to?  Other than for VMS and limited Tru64, it's Opteron!  Go
try to order some DL145, DL385 or DL585 systems and check the back logs.

It's not because AMD doesn't have stock either.  And that's just the
server-end.  On the desktop, HP has had _massive_success_ with AMD
sales.  And Intel is pissed, and _everyone_ knows it.

#106 is a name from the past, Vobis.  Vobis went from 100% DR-DOS to
100% MS-DOS in less than a year due to a man at Microsoft by the name of
Jochaim Kempin.  Vobis didn't want to make the same mistake again.

#110+ is definitely a major issue.  Intel is seriously affecting the
industry through the standards process with ADT.  Not surprising, where
Intel's partner in Rambus failed to overrun JEDEC, Intel might succeed
itself.

#122-126 is a _real_problem_ for AMD.  Because of the real-time, out-of-
order and register renaming capability inherit to _all_ Athlon onward
processors, the compile-time optimizations for the P4 give the _same_
boost to Athlon processors too.  So when the first generation of P4
optimized Intel compilers came out, Athlon was right there, getting the
same boost!  For the first time, Intel was having to deal with a
competitor that has 1 more ALU pipe (with far shorter, more optimized
pipes), as well as 1 more FPU pipe (the 3 pipes being capable of 2
complex + 1 ADD/MULT versus their 1 complex _or_ 2 ADD -- hence why the
SSE pipes are typically leveraged).

So now Intel has added CPU detection software to _purposely_ run
alternate code on Athlon processors.  Let me say that again, the code
that runs on Athlon processors would be _crippled_ even on a P4!  In
other words, it's no longer about the code being Intel optimized which
runs poorer on AMD.  It's about the loader detecting if it's an AMD, and
then throwing it garbage, garbage that would also run poorly on a P4,
but doesn't because it loader detects and Intel chip and runs a
completely different set of code.  The FPU in all Athlon processors
_roast_ the P3, let alone the P4, and even SSE is "lossy math" on the
P3/P4 (whereas AMD leverages spare FPU cycles -- it can do 3x MULT, the
i686 FPU in the P3/P4 can only do 1x MULT, and uses its lossy SSE pipes
to add 1 and 2 more, respectively to the P3 and P4).

I didn't agree with most of the MS trial on either side.
But in AMD v. Intel, sorry Intel, you're gone too far.

I think HP is a perfect example, customers wanting AMD, willing to pay
for it, even HP saying Opteron is superior to both Xeon and Itanium2,
and you won't let them have it, even if HP doesn't need your money.

Worse are the compilers -- you can't win by optimizing for even your own
products because even AMD bests you there.  So much so that even the
Fedora Core project optimizes for P4 because Athlon benefits too.  So
you have your own compilers chuck out _junk_ and the loader executes it
if it detects a non-Intel.  That's just blatant.


-- 
Bryan J. Smith                                     b.j.smith@ieee.org 
--------------------------------------------------------------------- 
It is mathematically impossible for someone who makes more than you
to be anything but richer than you.  Any tax rate that penalizes them
will also penalize you similarly (to those below you, and then below
them).  Linear algebra, let alone differential calculus or even ele-
mentary concepts of limits, is mutually exclusive with US journalism.
So forget even attempting to explain how tax cuts work.  ;->



-
To unsubscribe, send email to majordomo@silug.org with
"unsubscribe silug-discuss" in the body.