Microsoft slowly backs away from Itanium

[trooper11]

i saw this remark about ms itanium support that i found interesting.:

<A HREF="http://www.ptc.com/partners/hardware/current/itanium_letter.htm" target="_new">http://www.ptc.com/partners/hardware/current/itanium_letter.htm</A>

it does seem ms is dropping some support for itanium now and maybe more later if xeon takes off more

 

[sonoran]

Funny - I couldn't even find the word Microsoft on that web page.

 

[FUGGER]

False info, you are misreading the point of the announcemment of PTC

Xeon will have 64 bit enhancements somewhat similar to the A64. They plan to go with Xeon it seems as it will hit more customers than the I2.

<b>My sig is better than yours.<b>

 

[trooper11]

hmm true, i wonder why i kept thinking ms lol, sorry about that.

but it still shows now that intel is making the move to integrate itanium and xeon to the same platform, that developers are leaning more towards the xeon 64bit architecture.

 

[P4Man]

well as pointed out, this is not about MS pulling the plug, but PTC.

That still doesn't send a good message though, since Pro engineer and other PTC apps *have* been ported to IA64 (I wonder, using intel IPF development funds ?), and yet they are pulling support for it, effective *immediately* even though there is no 64 bit x86 replacement yet. That clearly shows Pro/E IA64 sales where close to zero, and PTC was not expecting it to pick up in any significant way. So much for IPF and the workstation market.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[slvr_phoenix]

You can lead a horse to water, but you can't make it drink. And in this case, why water it with Perrier when tap water will do?

<pre><b><font color=red>"Build a man a fire and he's warm for the rest of the evening.
Set a man on fire and he's warm for the rest of his life." - Steve Taylor</font color=red></b></pre><p>

 

[juin]

In graphic Application it too hard for itanium to compete as it use much less mature graphic driver.Even XP 64 see a large decrease in game due to driver even if is X86 for X86.For nvidia it got more sense to back up AMD 64 that IA64 in any case itanium sale are growing fast and it will enter mainframe soon.

i need to change useur name.

 

[P4Man]

Intel can't make a profit with Itanium being restrained to the high end server market they need bigger volumes. Workstation market is an obvious candidate, especially since it tends to love FP performance and more often than not, requires 64 bit. If intel can't crack this one, it doesn't bode well at all.

As for fast "growing sales".. 100% increase over a half dozen is still only a dozen. After almost 5 years, its sales are still almost insignificant. That is not quite impressive for a platform that was poised to replace pretty much the entire risc market, and for which PA Risc, Alpha and Mips where sacrified.

And mainframe ? LOL, I had not yet heard IBM was porting z/OS to it. Maybe you meant the Tandem/Nonstop market.. thats great, that may sell another tray or two of chips per year, I'm sure that will make intel rich. And yes, I know how much such systems and their support cost, but that only helps HP/IBM/Unisys/Bull/.., not intel which only gets a tiny fraction of that revenue. A few thousand dollars per chip in a market that sells maybe a few hundred systems per year... that is not even pocket change.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[FUGGER]

Itanium does not have to make a profit, its future is already set and it has little to do with consumers. You are missing the bigger picture.

Goverment, medical, and space applications alone ensure the I2 growth. Three segments you are completely obivious too. You are limited to what you read here it seems.

I2 is not aimed at the consumers or small business like the Xeon is, the Xeon will hit more customers for PTC and it was a good business choice to move to that platform and not the A64 platform.

<b>My sig is better than yours.<b>

 

[P4Man]

>Itanium does not have to make a profit, its future is
>already set and it has little to do with consumers. You are
>missing the bigger picture.

Oh no, I'm not. Of course Itanium should pay for it selve, or it will be scrapped. Intel currently pays around $4-$500M for ongoing R&D in cpu development, compilers and chipsets while last year, it got some $100-150M in revenue. That is a $350M loss on top of an initial investment of a ~$5B. You really think intel's shareholders will let intel continue throwing away their money if it doesnt offer a promise of future profitability ? You think they aren't pissed off enough yet that those billions are likely to be completely unrecoverable now Itaniums future is most likely limited to the high end, low volume niche markets it is currently competing in ?

>Goverment, medical, and space applications alone ensure the
>I2 growth

Growth is such a meaningless term at its current peanut shipment levels. It needs volume, those $500M recurring costs need to be paid for somehow, and selling thousand chips per year in the tandem/nonstop/vms market is NOT going to make that happen and the fact those <b>systems</b> may represent a billion dollar market is no help for intel.

>Goverment, medical, and space applications alone ensure the
>I2 growth. Three segments you are completely obivious too.
>You are limited to what you read here it seems.

Completely irrelevant. You are probably confusing system sales and associated revenue with intels cpu sales. Sure, HP, SGI, Bull, etc could make a nice buck selling only a handfull of high end IPF machines in such markets, much like they do or did with PA Risc systems, MIPS, etc, but the percentage of that revenue that intel captures on for instance a 64 cpu Altix system and support contracts, is marginal. Maybe 3%. You think IPF is a charity program of intel for its oems ?

>I2 is not aimed at the consumers or small business like the
>Xeon is

No kidding einstein, that is not the point. But intel needs to find markets that will allow it sell at least 500k or a 1m chips per year, depending on ASP's, or IPF will just bleed more money; "Goverment, medical, and space applications" just do not offer such volumes, where the technical workstation market would be a better bet already. Why is it you think IBM is pushing Power chips into Macintoshes and even into consoles these days ? And consider IBM, unlike intel, can absorb its losses on Power development since these costs allow it to sell high revenuee high margin systems. intel can't it only makes money selling the chips.

Its very simple: HP dumped PA Risc development because it was getting too expensive, in spite of the fact PA Risc chips are what enabled it to sell billions of lucrative high end gear, software and service contracts. So they made a deal with Intel which took over the cost, and designed IPF which was poised both to replace PA Risc and Alpha as well as become a high volume, industry standard product over time. Thats the only way such a deal makes sense. As it is, it looks like IPF will basically replace PARisc and alpha, and nothing else, so HP managed to outsource that development at almost no cost. I'm sure some people at HP are laughing their asses off right now.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Xeon]

You really think Intel’s shareholders will let Intel continue throwing away their money if it doesn’t offer a promise of future profitability?

For the last 35+ years Intel has promised future profitably from technologies. You really think the investors didn’t see the Willy? Intel has something AMD does not Investor trust. Never in the history of Intel has their investors tried to sue them *cough* AMD.

Growth is such a meaningless term at its current peanut shipment levels.

So the purchase of the I2's for the FBI's carnivore project, I2's for NASA's Mars landing simulations, I2's for the Pentagons missile defense programs and nuclear weapon simulations or perhaps the future stealth air superiority fighter the raptor, I2's for Wall Street, or I2's for the human genome project, or maybe NVIDIA’s drivers churning systems that run on I2's and A64's?

Where does the scope begin and you’re European short sightedness end?

Its very simple: HP dumped PA Risc development because it was getting too expensive,

Sorry Itanium development was well underway the technologies from the alpha team were purchased by Intel and everything there in.

See with you it’s quite obvious that you’re short sited. Over the last 5 years Intel has been consuming companies that make competitive technologies to the Itanium line, most of which are patents that incur possible "sharing". If Intel wasn’t planning for the future they would have dropped the technologies quite quickly, yet they continue purchasing companies up, like Elbrus and UniPro for instance.

I'm sure some people at HP are laughing their asses off right now.

If they are they should be fired, disobedience in the workplace is unacceptable.

Xeon

<font color=orange>Scratch Here To Reveal Prize</font color=orange>
<font color=white>In the case of over scratching contact THG for property damage claims.</font color=white>

 

[P4Man]

>You really think the investors didn’t see the Willy?

Willy made intel a bunch of money, it wasn't an unrecoverable cost of a few billion dollar.

> Never in the history of Intel has their investors tried to
>sue them

They don't have to sue them for pressuring the intel board. Intel is owned by its shareholders, they have the last word, period.

>So the purchase of the I2's for the FBI's carnivore
>project, I2's for NASA's Mars landing simulations, <snip> >blah blah

That's a great list. very impressive (aside from the human genome project where intel actually donated the chips) and I'm sure you could make a looong list. To the point where you have listed each and every sale that contributed to intels 100k units shipped last year. It doesnt matter wether those chips are used to power game consoles, pornowebsites or space stations, what matters is how many they sell, and at what price. So far, they sell peanuts. ($150M in revenue, is peanuts for a company like intel, and is peanuts compared to the costs of developping the chip).

>Sorry Itanium development was well underway the
>technologies from the alpha team were purchased by Intel
>and everything there in.

So what are you saying ? HP dumped PA Risc development, because they found it too expensive. Are you denying that or contradicting that somehow ?

>See with you it’s quite obvious that you’re short sighted

No I'm not. Its quite simple really, the market Itanium currently competes in, just doesn't offer the volume required to pay the R&D bills if you're only generating revenue from the cpu's like Intel, and unlike anyone else. Seriously, if HP saw no financial reason to continue developping its own high end CPU's which enable it sell a few billion worth of servers, HP-UX licences, software, consultancy, support contracts,.., you seriously think intel could make money selling just the chips if it can not outgrow the PA Risc market *significantly* ?

intel would need around 50% of the high end unix and HPC market if it wanted to just break even on its investments, it simply doesnt make economic sense at their current 5% or so. Especially when you consider the market this market is shrinking year after year after year.

Anyone with a clue understands IPF made sense for both Intel and HP only if it could leverage the investment by selling those chips in a much bigger market than just PA Risc superdomes and Altix servers. IPF only made financial sense to intel as a long term replacement of x86 when and where 64 bit support was required. With the arrival of EM64T, IPF's chances of ever achieving that, are getting close to zero.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Xeon]

Willy made Intel a bunch of money, it wasn't an unrecoverable cost of a few billion dollar.

No but you missed the point, the promise to make money on either core was the same. But from a technology acceptance point IA-64 is too new people don’t want to leave behind everything that’s built around IA-32, remember Intel is aware of this.

For the long term I see the IA-64 technology evolving significantly better than IA-32 IPC cap and clock speeds vs. micron size leakage, yada yada.

Intel has taken this into account and rest assured they will have the final say in IA selection for mainstream users.

They don't have to sue them for pressuring the Intel board. Intel is owned by its shareholders, they have the last word, period.

As is every other public company here in North America what’s your point? Perhaps coining the idea that the Board has controlling shares in the company? Which is impossible with people as rich as Bill Gates popping up all over.

That's a great list. very impressive (aside from the human genome project where intel actually donated the chips) and I'm sure you could make a looong list. To the point where you have listed each and every sale that contributed to intels 100k units shipped last year. It doesnt matter wether those chips are used to power game consoles, pornowebsites or space stations, what matters is how many they sell, and at what price. So far, they sell peanuts. ($150M in revenue, is peanuts for a company like intel, and is peanuts compared to the costs of developping the chip).

You again missed the point, the technology is gaining acceptance. Where it lands people notice that’s the kind of effect that they are looking for, it spurs talk about power, speed, resources the whole nine yards. When you convince the Top 100 companies in the world with a product priced and designed for them don’t think fortune won’t write about it.

Once the movers and shakers get on the industries gaming/SOHO software/Specialized computing ect. Will build software to meet the new immerging technologies.

So what are you saying ?

No HP saw they were outgunned with Intel’s solution.

No I'm not.

Yes you are, and we will discuss this later.

Its quite simple really, the market Itanium currently competes in, just doesn't offer the volume required to pay the R&D bills if you're only generating revenue from the cpu's like Intel, and unlike anyone else.

It doesn’t need to it’s the market penetration they are looking at. When something is damned good don’t think people won’t talk about it.

Seriously, if HP saw no financial reason to continue developping its own high end CPU's which enable it sell a few billion worth of servers, HP-UX licences, software, consultancy, support contracts,.., you seriously think intel could make money selling just the chips if it can not outgrow the PA Risc market *significantly* ?

Lack of foresight as always they want the technology to come out of the gates unchallenged. No patent infringements, good compilers, strong chipset support, strong OEM support. Add that all up and when Intel brings IA64 to a desktop solution there won’t be competition for 64bit 100% support home user solutions.

50%

Note even close enough people that buy that type of hardware don’t upgrade for quite some time, sometimes in excess of 10 years.

IPF only made financial sense to intel as a long term replacement of x86 when and where 64 bit support was required. With the arrival of EM64T, IPF's chances of ever achieving that, are getting close to zero.

Right on the fiscal and long term replacement of x86. But EM64T is nothing, as long as somewhere in there the technologies are compatible with IA64 it will be a smooth transition.

Intel likes transition that’s how they have always been.

Xeon

<font color=orange>Scratch Here To Reveal Prize</font color=orange>
<font color=white>In the case of over scratching contact THG for property damage claims.</font color=white>

 

[Kelledin]

No but you missed the point, the promise to make money on either core was the same.

Except that the IA64 promise is slipping away.

But from a technology acceptance point IA-64 is too new people don't want to leave behind everything that's built around IA-32, remember Intel is aware of this.

With the advent of 64-bit x86 extensions, people have even less reason to go to IA64.

For the long term I see the IA-64 technology evolving significantly better than IA-32 IPC cap and clock speeds vs. micron size leakage, yada yada.

The problem with that prediction is that research, development, and refinement efforts tend to focus on the profit-generating volume part. Otherwise x86 would never be where it is now.

Intel has taken this into account and rest assured they will have the final say in IA selection for mainstream users.

No, the mainstream users will have the final say in IA selection for mainstream users. That's the way a competitive market works, and Intel doesn't have quite enough monopoly power to stop that.

You again missed the point, the technology is gaining acceptance.

Except that AMD64 is gaining acceptance quite a bit faster.

Where it lands people notice that's the kind of effect that they are looking for, it spurs talk about power, speed, resources the whole nine yards. When you convince the Top 100 companies in the world with a product priced and designed for them don't think fortune won't write about it.

Except that in many cases, AMD64 parts perform better. IA64 is primarily good at floating-point and RAS features (like lockstep, ECC cache, et al), and in some cases it can have better SMP scaling. Ironically, none of that actually has to do with its "64-bitness."

No HP saw they were outgunned with Intel's solution.

That's not saying much; HP isn't that great at processor design. PA-RISC wasn't that great, and Alpha wasn't even HP's baby. Alpha was DEC's baby, then Compaq's neglected foster child.

IA64 was just a convenient means for HP to foist CPU development off on someone else while still building servers and operating systems around Intel's baby.

It doesn't need to it's the market penetration they are looking at. When something is damned good don't think people won't talk about it.

When AMD64 matches so much of what IA64 does, don't expect the world to beat a path to Itanium's door.

Right on the fiscal and long term replacement of x86. But EM64T is nothing, as long as somewhere in there the technologies are compatible with IA64 it will be a smooth transition.

Intel likes transition that's how they have always been.

How well can Intel pull off this kind of transition? One word (or number): 80186

<i>Look! Up in the sky! It's a bird! It's a plane! It's...

...an asthmatic werehamster?

<LHGPooBaa> Well, @#!& on me.</i>

 

[P4Man]

No offense, but it seems you are utterly clueless about the markets you are talking about..

>Intel has taken this into account and rest assured they
>will have the final say in IA selection for mainstream
>users.

So what are you saying.. Intel will indeed drive IA64 to the desktop ?

>As is every other public company here in North America
>what’s your point?

Point is intel can't continue to waste money on IPF forever (even if it can afford it) if it can not demonstrate a likely future ROI to its shareholders. Shareholders rarely have an horizon that exceeds 3-5 years btw.

> the technology is gaining acceptance.

Yes, at the current rate it will be where Alpha was 5 years ago by the end of the decade. Very bright future.. You still don't seem to get it. IPF is sold in a merchant model, intel doesnt sell systems.

-Power is developped and fabbed by IBM, and IBM makes tons of money on its (very) high margin I-series and P-series (and some selling POwer chips by the millions to Apple and soon microsoft and Sony). Therefore, IBM can afford to dump billions into Power development.

-PA Risc was/is developped by HP, and enabled it to sell truckloads of multimillion superdomes contracts, HP-UX licences, Non stop systems, etc, so you might think HP could afford to dump some money on cpu development, but they could no longer warrant the expense ! Hence, they handed their VLIW design to intel, killed PA Risc as well as Alpha and bought their chips from intel.

-UltraSparc was developped by Sun, enabled it to make a fortune selling high end gear and solaris solutions, but guess what.. they no longer could warrant the expenses, they killed US-V and will use Fujitsu Sparc64 cpu's instead.

-Intel developped IPF (together with HP), and sells ONLY cpu's (and some chipsets, big deal). This is a totally insignificant portion of the revenue in these markets. $1-4.000 per cpu may sound like a lot, but its *nothing* on a typical Altix/Superdome contract, yet intel is spending billions developping this chip.

Now, if HP and SUN couldnt warrant developping highend cpu's anymore, IBM only because of its volume with their Apple/Xbox/PS3 sales, even though these companies are reaping in revenue maybe 30-50x the cost of the cpu's, how do you expect intel to make a living on the most expensive (to develop) and least rewarding (revenue) part of this entire market ? The market just isnt big enough for a merchant model chip, no matter how good or how well accepted it is. It is what killed Alpha, it is what will kill IPF if intel can not sell IPF chips in higher volume, or does not manage to capture the majority of the high end Risc/Unix/HPC market. Neither of them looks likely.

>Once the movers and shakers get on the industries
>gaming/SOHO software/Specialized computing ect. Will build
>software to meet the new immerging technologies.

Oh gimme a break. Alpha has been there done that, went nowhere. Power, Mips and Alpha where very well entrenched in their high end markets, outperformed x86 by HUGE margins and tried to grow into the x86 market, with Windows support and everything, yet they failed miserably. Instead, ironically, x86 ate a very large part of their original market (workstation, entry level server, HPC) to the point where all three architures are either dead or dying now. IPF *could* have been different, if Intel had forced the market to switch from x86 to IA64, by not releasing 64 bit x86 chips, that was their last hope. Opteron killed it, plain simple. Now even 100% performance leads is not going to kill x86, IPF can never compete with the economy of scale and the incredible software support for x86.

>No HP saw they were outgunned with Intel’s solution

Do some reading or don't post on things you know nothing about. VLIW was a HP concept, they handed it over to intel asking if they didn't want to develop and fab the chip. intel turned it into the way overdue Merced dud, and HP took over the development of McKinley again to produce at least a somewhat respectable chip. IPF/VLIW/Epic originated at HP, not intel. And HP dumped PA risc development for no other reasons as cost cutting.

>It doesn’t need to it’s the market penetration

Please enlighten me, what is the difference between market "penetration" and market share ? And why on earth would penetration be more important ?

>they are looking at.

Indeed. But hardly buying it.

>When something is damned good don’t think people won’t talk
>about it.

LOL, get a clue. This market doesnt sell on hearsay.

> Add that all up and when Intel brings IA64 to a desktop
>solution there won’t be competition for 64bit 100% support
>home user solutions.

ROFL !!! They are still nearly nowhere in their primary market (highend unix), they are completely nowhere in the workstation market, but you think intel will manage to conquer the x86 desktop market with IPF now ? LMAO ! Not even intel has any hopes left for this, or they would have tried by now, and definately NOT have released EM64T. No way in hell IPF is going to play any role on the desktop for the next ten years, which could as well be forever in this market.


>Note even close enough people that buy that type of
>hardware don’t upgrade for quite some time, sometimes in
>excess of 10 years.

Waaaaaaaaaahahaha.. you're hilarious. So, Intel doesn't need 50% of the market because those "people" don't upgrade often ? Now there is some logic. Get a clue, intel needs to SELL chips, and competes with IBM, Sun, etc in doing so. Only so many chips/systems are sold each year, if intel wants to make money it needs a part of the pie. By my math, around HALF the pie, if not, it will cost intel more than it would generate revenue. long lifecycles has nothing to do with this. If anything, IBM/Sun/HP take advantage of this since even sales they achieved 10 years ago, still generate revenue for them in support/service/consulting, but not for intel.

> But EM64T is nothing, as long as somewhere in there the
>technologies are compatible with IA64 it will be a smooth
>transition.

Well they are completely, 100% incompatible, and there will not be a transistion. IPF will mainly serve as a cheap PA Risc replacement for HP, enabling Intel *maybe* 20-30% of the risc market if HP customers don't switch to IBM Power, Sun or x86 instead, but in a market that will shrink further and further because of x86 commodity chips. 10 years from here, it may be as small as the IBM mainframe market today, dinosaurs from the past, which still generates nice profits for IBM, but not because of the few dozen cpu's per quarter that are involved. Intel is scewed with IPF.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[P4Man]

>That's not saying much; HP isn't that great at processor
>design. PA-RISC wasn't that great

Hmm.. I kinda liked my old PA Risc workstation to tell you the truth, and I think they had a damn nice ISA. Performance was also right up there with rest, and even though its nearly dead now, even the recently released PA 8800 dual core chip still kicks some major butt in spite of being deprived of any serious development funds. It really wasn't a bad chip, much better than UltraSparcs or Mips, and not far behind power or alpha.

As for HP's design team.. Intel messed up Merced big time, and HP actually saved it by designing McKinley (which evolved into madison). So if Itanium is quite a respectfully performer today, its mostly because of HP's design team, and intels fab engineers. We have yet to see how well Intel's own design will be...

IMHO, intel is good at a lot of things, like process engineering (world class without a doubt), chipset design or compiler development; but designing cpu core's is not its greatest strength IMHO. Even AMD much smaller team is a lot better at it (but not quite as good in process engineering). Maybe the ex-alpha guys can bring us some magic again, but intel's own teams really aint that great IMHO.

In 25 years they made just one really nice design, and that is the PPo. Everything else ranged from complete disasters (80186, iAPX 432, i960, Tejas) to piss poor (286, merced, willamette, prescott) to mediocre at best (3/486, pentium, northwood*,). Not a fabulous trackrecord...

*northwood was a mediocre design IMHO, saved by excellent fabbing, chipsets and compiler support. If Athlon had enjoyed similar benefits, it would have been a much better chip.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Zeekfu]

Yea, what he said.

 

[Zeekfu]

....but wait. That sounded good too.

 

[Zeekfu]

...still more interesting points...

 

[Zeekfu]

......wait.....what were we talking about????????

 

[endyen]

We were talking about what xeon said.
"When something is damned good don’t think people won’t talk about it."
Just remove the word "good".

 

[Zeekfu]

My prediction is that after the 4th quarter...

 

[Zeekfu]

I'll have a whole dollar...

 

[endyen]

And you are doing it your 2 cents worth at a time.

 

[Zeekfu]

he he:>)

<P ID="edit"><FONT SIZE=-1><EM>Edited by zeekfu on 06/15/04 10:12 PM.</EM></FONT></P>