Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Lucky that Opteron runs 32 bit software so well, while giving purchasers that warm fuzzy feeling that
comes from knowing they will be able to run 64 bit software if it becomes the wave of the future.
Just like 3Dnow gave Athlon users the warm and fuzzy that they would be able to run SIMD-intensive
code of the future? Curious how AMD had to add SSE support to Athlon and SSE2 support to K8.
Too bad there are probably legal roadblocks that prevent it from adding IPF support to K9 and allow
AMD users to run the 64 bit software of the future. Instead they will stuck in the 64now! ghetto. :-P
Ooops, Freudian slip.
I thought I had told you not to bother responding to my posts unless you could behave. Don't
waste my time because I won't stoop to your level again.
Having justed looked over your track record of 29 posts to refresh my memory I must say
this comment makes me LOL.
soon we might see Sun carrying Itanium 2 servers.
As long as Sun keeps its why-Itanium-will-fail web page around I give this about
a zero chance. Maybe you should benchmark that page and look at it regularly
as an early leading indicator. ;^)
Anyone want to give odds whether there is an Xscale in that picture phone?
YB, Intel's own published research papers show that 90nm SRAM is twice as dense
as 130nm SRAM. By that, I think you can assume that Dothan at 2MB L2 will be about
the same size as Banias with 1MB of L2, or most likely a little smaller.
If you look at the Banias die the L2 occupies about 40% of the area. Doubling the L2
capacity while halving the cell size will slightly reduce the L2 in absolute size (larger
absolute capacity raises array cell efficiency - i.e. larger memories have a smaller
relative area fraction for peripheral circuitry). The CPU core region will probably shrink
in area by about 35 to 40%. So Dothan will be roughly 62 to 65 mm2 even with 2MB
L2, assuming no other major changes. (Banias is 83 mm2)
Besides that, Intel has said to expect 50% higher frequency out of Dothan. That's
would give an EOL frequency of about 2.5GHz, with about 5% greater IPC.
The top end frequency is about right but I doubt IPC will go up 5% without improvements
to the CPU itself. The larger L2 will barely cover the effect of CPU/memory performance
scaling imbalance. Intel might up the FSB but that would raise power disproportionate
to the associated performance increase.
Cowabunga!
Farmboy eh? Please, don't ask, don't tell.
At least the rumor mill is giving Dell a break for a change.
So, no, they don't traditionally
sell for whatever price the market will bear, that was not their business model.
ROFL!
Are you trying to bring in the idea of price/demand elasticity for AMD's products?
That is hardly a good tactic because that highlights the fact that not only does
the market value Intel processors about 2.5x higher than AMD processors it
is willing to buy more than 5x more Intel processors at that relative valuation
even while AMD manufacturing capacity goes underutilized. Ouch!
Haven't you learned yet that the Intel products sell at the price for which the market will pay?
Why would he accept that if he won't accept that AMD products also sell at the priice for which
the market will pay? Accepting both, along with the most recent ASP figures in AMD and Intel
financial disclosures, would mean facing the cold hard fact that the market votes with its wallet
with the opinion that Intel uPs are roughly 2.5x more valuable than AMD uPs on average. :-P
Not doing so well on your short position hmmm?
Andy Glew is at AMD now?
Missed that. Wonder what happened? Itanium didn't suit him?
*** NEWS Flash for the clueless ***
Intel still has multiple generations of new x86 processor cores in development.
Andy never worked on IPF as far as he's said. Judging by his various comments
on comp.arch IMO Andy left Intel for AMD in attempt to regain a feeling of being
a big fish even it meant moving to a small pond. He was prominent in the early
1990's development of the Pentium Pro but doesn't seem to have contributed
much since. For whatever reason he wasn't part of the core group that designed
the Pentium 4 microarchitecture, a group he apparently has some philosophical
differences with. He seems like a bright guy so I hope he finds his skills employed
more to his satisfaction at AMD. However I would keep in mind his circumstances
when considering anything he might say about Intel and its current product
directions.
Intel has spent hundreds of millions of dollars on compilers, tweaking code to take advantage o
loopholes in the SPEC suite design rules during the past few years.
You are just bitter that the best x86 compiler development group in the world will never
use any of the 64 bit bells and whistles that AMD has spent years cramming into Hammer.
And worse yet, they still can make it run faster than all the second class compilers that do.
Intel now expects revenue to land somewhere between $6.6 billion and $6.8 billion for the period,
which end June 28. Analysts had previously been told to look out for revenue between $6.4 and &7.0
billion, but Intel is allowed to change its mind.
Huh?
How can narrowing a range of uncertainty by bringing both sides closer together by moving
each inwards equally be considered changing one's mind?
The stereotype that most journalists end up in their profession because it is one of the few that
requires zero math skills is once again reinforced.
a field of four - Intel, AMD, Transmeta and Via
Transmeta is in the x86 processor business?
By that loose a criterion so was DEC when it offered FX!32 for Alpha/NT.
Intel will launch new versions of the chip, and the king of processors says its plan for more Itaniums
is unstoppable. If this is true, Intel will soon launch Itaniums that use an improved architecture.
Code optimized for the McKinley I2 which has been shipping since last summer will run
full speed on the this year's Madison I2 6M, next year's Madison I2 9M, and 2004's Montecito
since they all use the same CPU microarchitecture.
But the problem with compiler-based optimisation is that software runs best when it is
optimised differently for each version of the chip's architecture.
And this is somehow different for IPF than it is for POWER3 vs POWER4, EV5 vs EV6, PIII vs P4,
and K7 vs K8? I bet you get a far greater performance increase recompiling a PIII app for P4
than you do recompiling a Merced app for McKinley. What an idiot.
The Epic approach seems to signal server software upgrades every year or two.
Only if you are blindly ignorant of the facts.
After some 10 years of development, Intel and HP have made a costly chip that seems suited to a few niches.
Yeah, commercial servers and technical computing.
BTW, the McKinley design project started in 1997 and the chip taped out at the end of 2000. The chip was
officially introduced in mid 2002. That is roughly 5 years, not 10 years and compares well to any number
of recent server MPUs I could mention. IT Weak must pay this guy by the mistake, not a fixed fee.
It's hard to say whether HP is disappointed by Itanium sales, but it seems odd that last year's release
of HP-UX supported only Itanium chips, while this year's version supports both the Itanium and its
predecessor - HP's PA-Risc chip.
Huh? HP-UX has supported HPPA since what, the 1980s? HP had two separate branches of HP-UX,
one for HPPA and one for IPF, and recently merged them. This guy is unreal.
It looks more like a rack mounted cluster of 4 way servers than a single
64 processor server.
BTW, if this exists why did they have to put a sloppily photoshop'd fake
image of it on their website? Visual Technology indeed! I guess they will
assemble the first one when (if?) they get their first order and take a snap
of the real thing then.
Intel is feeling the pressure from AMD.
LOL, that must be why Intel and AMD share prices did the triple cross-over
thing today.
Intel probably feels more pressure when Barrett eats lunch at Taco Bell than
from anything AMD does.
I originally stated something to the effect that there is a PIII core that runs faster (not clock speed,
just application performance in the avg joe mindframe) than the P4 at the initial speed grades (1.3 to
1.5 GHz).
The PIII shipped as fast as 1.4 GHz. But all PIII faster than 1.0 GHz were Tualatin PIIIs, i.e.
manufactured in a 130 nm process. The 1.3 GHz P4 was made in 180 nm process and
no 180 nm PIII ran faster than 1.0 GHz (there was a 1.133 GHz but it was recalled).
If you want to compare without regard to process you can show just about anything. At the
end of the month you will see 130 nm IPF processors that clock faster than a 1.3 GHz P4.
I originally stated something to the effect that there is a PIII core that runs faster (not clock speed,
just application performance in the avg joe mindframe) than the P4 at the initial speed grades (1.3 to
1.5 GHz).
The PIII shipped as fast as 1.4 GHz. But all PIII faster than 1.0 GHz were Tualatin PIIIs, i.e.
manufactured in a 130 nm process. The 1.3 GHz P4 was made in 180 nm process and
no 180 nm PIII ran faster than 1.0 GHz (there was a 1.133 GHz PIII but it was recalled).
If you want to compare without regard to process you can show just about anything. At the
end of the month you will see 130 nm IPF processors that clock faster than a 1.3 GHz P4.
Anyway, HP's customers at least have a choice now. Which is a good thing. Dell's customers dont have that.
And look how oppressing the masses has cost Dell so dearly. ROFLMAO!
I sure hope HP will ask a couple of Dell's big customers if they will or not.
That shouldn't be too hard. I am sure they already have the names and phone
numbers of many of them from when they used to be HP or Compaq customers.
Yes, they have. Any more stupid questions?
Sorry to be so blunt, but asking if Dan3 has any more stupid
questions is an even stupider question than the one posed.
I wouldn't have said the above except the bee's give no quarter to the descenter. Just as well though.
Are you calling AMD supporters smelly? And what's this about giving them a
quarter? If they want to buy an XP 2500+ they can use their own money.
Uh, people, pay attention here.
Uh, we understand what was said in the original story and how edgarcayce misunderstood
it. What is being commented on is how ludicrous and unlikely edgarcayce's intrepretation
was with even a tiny amount of critical thinking applied.
wbmw-As of 9/00 AMD has 10Mil Patents...
You don't seriously believe that do you?
Newer benchmarks always tend to favor Intel's processors over AMD's.
Newer benchmarks also tend to favor Intel's new processors over its older processors.
Considering that Intel often changes the code optimization rules from uP generation to
uP generation this is no big surprise. Good coding practices for P6 based products
differed from Pentium just as good coding for Netburst based products often differs
somewhat from P6.
If processors reach the commoditization of soft drinks, both INTC and AMD would
be in a world of hurt, and only VIA would be profitable in this business.
I disagree. In such a situation branding would be all inportant. Coca Cola and Intel/Pentium
are brands worth tens of billions of dollars and have global reach. AMD would be in the
position of perhaps Dr. Pepper, or maybe even local bottling plant's generic cola.
Would you agree that it means Intel has violated the terms of its consent agreement with the FTC
No, I am not privy to all the details but since the license to Alpha IP that Intel obtained
was not exclusive the consent degree was not ostensibly violated. All previously license
holders are still free to design, build, and sell Alpha processors to whatever extent they
could before. BTW, it is up to the FTC, not you or I, to decide whether the agreement
was violated or not and if so take appropriate action.
and that it's a criminal organization
After all your nastiness, that's not really good enough.
You lecturing others on etiquette is like Jeffry Dahmer promoting vegetarianism.
If I can show you a link that shows Intel paid Compaq for that technology, what will you do?
I will freely admit that they bought it.
Intel didn't buy anything
Compaq Visual Fortran engineering team has joined the Intel Compiler team
buy v.t. to obtain by payment; to purchase; to pay a price for;
Show me an URL describing what Intel paid for the compiler team or admit you
are as full of excrement as ever.
Chipguy replied: Intel didn't buy anything.... AFAIK the EV79 team is still with HP/Compaq.
WBMW would be typing LIER at this point, but we'll be charitable and just assume you're completely
ignorant about the industry.
No he wouldn't for two reasons. First of all he is literate and would probably be able to spell a
four letter word like liar properly. Second of all there isn't a single mistatement of truth in my
comments about the fate of the Alpha design team. Knowledge of which comes directly from
people involved in the move.
Dan3, I burned you on your filthy lies about Alpha in the past and I challenge you to show anything
I said in that post was incorrect. Go ahead and just try to show you weren't just engaging in your
trademark combination baseless ad hom attack and misrepresentation.
Here's what Intel says on Intel's web site:
The Compaq Visual Fortran engineering team has joined the Intel Compiler team
I don't believe I made any claims about the compiler team one way or another in my post about
Intel hiring elements of the Alpha processor design team. AFAIK they went in the same wave as
the EV8 crew. Nice attempt to obfuscate
chipguy, have you ever done architecture work??
Yes.
I'm as disappointed as you are on the low Mhz situation.
But, it is misleading to say that opteron is not competitive.
I agree. I think the Opteron is a competitive server chip. I was responding
to a series of obnoxious chest pounding by AMD partisans on how it did
on a collection of benchmarks. I thought it was appropriate to bring these
individuals back to earth by reminding them that AMD blew the big one,
pure single thread CPU performance, the one which AMD VP Weber had
explicitly promised that Opteron would take the crown for when it was
released. Will it affect Opteron's success? I don't think so. The lack of
visible first or second tier OEM support and the financial fragility of
AMD are much bigger issues IMO.
I am not pretending about anything. I am poking holes in the thin
veneer of excuses that Weber/AMD apologists are tossing out to
explain away the quite clear inability of Opteron to deliver on the
promise of SPEC CPU 2k supremacy when it shipped made by
Weber last October.
Intel bought the entire Alpha group including their compiler group.
I thought slavery was outlawed in the U.S. more than 100 years ago.
Intel didn't buy anything. Compaq layed off the Alpha design team in stages
starting with the EV8 team. AFAIK the EV79 team is still with HP/Compaq.
Not one to miss an incredible opportunity to acquire A-1 talent en masse,
Intel made individual offers of employment to every affected member of the
ADT and nearly everyone personally accepted Intel's offer of employment.
I don't know if Intel committed to make offers to the remainder of the ADT
when their work at HP/Compaq is completed.
it is beaten by a bulk CMOS chip 2/3 of its size and half
the pin count.
That chip won't address more than 3G of memory.
Disregarding PAE for a moment, you think the P4 uses 31.5849625 bit addressing?
You are not exactly a "details person" are you?
Weber also lied about the benefit of 64 bits. AMD's 32 bit SPEC CPU submissions
beat the 64 bit SPEC submissions.
That has everything to do with compiler optimizations and nothing to do with Weber's veracity.
So Weber assumed ISVs would develop x86-64 applications in asm? I see
that you would rather paint him as an idiot than a liar.
I guess that takes your mind off AMD's failure to take the SPEC CPU 2k performance crown as Fred Weber promised
Failure, smailure. Q2 isn't over yet, is it?
So AMD released Opteron prematurely? They must have figured shipping slow
was better than shipping *even* later.
BTW, Weber also lied about the benefit of 64 bits. AMD's 32 bit SPEC CPU submissions beat the
64 bit SPEC submissions.
As you know, 64 bit mode is faster than 32 bit mode if you keep the compiler constant. Perhaps Fred wasn't expecting both the top Athlon compiler teams to get bought up in the mean time.
So what was Weber expecting when he made his claims? That Intel would support x86-64 in
its compilers?
To equate that with lying seems needlessly antagonistic.
Considering the post I was responding to I think that criticism is unwarranted, or at least misdirected.
Can I gloat a little bit, Chipguy, over performance of the Opteron serving web pages (SPECweb99_SSL) ?
I guess that takes your mind off AMD's failure. AMD's failure to take the SPEC CPU 2k performance
crown as Fred Weber promised at last year's Microprocessor Forum. Despite the expensive SOI
processing and on-chip memory controller it is beaten by a bulk CMOS chip 2/3 of its size and half
the pin count.
So chin up old boy, and keep focussed on serving web pages and 4P SPECrates.
BTW, Weber also lied about the benefit of 64 bits. AMD's 32 bit SPEC CPU submissions beat the
64 bit SPEC submissions. The fellow has a bright future in politics or used car sales. Or he can
stay at AMD and continue to lie on its behalf.
SPEC Validates AMD Opteron Processor's Claim as the World's Highest Performing Processor for 2P and 4P Servers
Yup, it really smokes those 2-4 way benchmarks. I guess that explains all the product announcements..."
Did you notice the careful wording - "highest performing", not fastest?
That's because they win on 2P and 4P *throughput*, not speed (i.e. SPEC CPU) as
a certain Fred Weber once promised. I guess when your orange turns out to be a
lemon it is best to make lemonade and hope people don't notice you once claimed
you'd have the best OJ.
Not that I expect the throughput victory to stand long with 130 nm IPF processors
imminent and 90 nm IA32 processors due later this year.