An upcoming Microsoft software product for clustered servers won’t run on Intel’s high-end Itanium 2 chip, according to a report first published on News.com. Instead, it will be optimized for a more mainstream type of server chip from Intel and rival AMD.
Microsoft Nixes Intel’s Itanium Chip
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
AMD’s work is finally paying off.
You can say that again, also Dell is again — let’s see if they actually do it this time instead of talking about it — talking about selling AMD CPUs with their server systems!
mv * > /dev/null
It’s not like people actually use Microsoft on clusters anyway, and if they are too lazy to optimize their software to run on Itanium 2 that’s their problem. Opteron is easy to port to because of the 32-bit compatability. They can do it slowly, unlike Itanium 2 which would be a significant rewrite because it lacks any decent 32-bit mode.
Microsoft conceeds the Itanium 2 cluster market to Linux-powered Altix clusters?
Don’t focus on the wrong aspect of the post. As I stated, the reason for doing so is the speed at which it can be done to port to Opteron compared to the amount of work needed to get it on Itanium 2.
I may be on crack, but did Intel miss an opportunity with their IA-64 platform, by refusing to create versions of the chip for everyone, including consumers? Apple successfully migrated their customers from the Motorola 68k series CPU’s to the faster and more modern RISC PowerPC line. Sure, their might have been some initial headaches, but they got it done.
I understand that Itanium runs x86 code pretty slowly, but wasn’t there a way for Intel to provide “good enough” performance for that window of time where new applications would be written that would take advantage of the new architecture?
Maybe Intel wouldn’t be facing the big problems they are having feding off the Athlon 64’s on the desktop, if they had been willing to be a little more courageous and willing to try something a little more radical.
As it stands, Itanium looks like it will be a very expensive CPU (cost for Intel to develop) for a very small niche.
AMD is now trading in the $21 range, up from $18 yesterday. (Disclosure: I own several shares of AMD.)
Apple is also trading a bit higher today, too.
(Disclosure: I own several shares of Apple.)
Who, exactly, is making itainium computers? I mean, I know that SGI is, but SGI does not sell enough units to make it worthwile for Intel. I wonder what this means for their Intel/Linux dreams. (And boy oh boy oh boy do I ever think SGI made the wrong choice going with the itanium. They should have kept designing their own chips or gone with the Power 5).
(And boy oh boy oh boy do I ever think SGI made the wrong choice going with the itanium. They should have kept designing their own chips or gone with the Power 5).
Apparently the performance of the Itanium2 for these type of machines is quite impressive. The only way the use of the Itanium could hurt SGI is if Intel pulls the plug on the chip. Otherwise their new Itanium2 systems seem to be very capable from what I have read.
Well I believe they did miss their opportunity to attract consumers to the Intel Itanium line. One factor was the cost and the other was delaying release because of waiting on Microsoft. Sorry but unless I’m mistaken Microsoft doesn’t own Intel and Microsoft’s Windows XP 64 is still in Beta testing. Even Longhorn won’t be released till late 2006. Let’s not forget that there are other distributions such as Novell’s SuSE Linux and Red Hat’s Enterprise Linux that already offer 64-bit software. Both Linux distributors support Intel and AMD processors. Consumers whether home or business based do not want to hear from a chip manufacturer that they are waiting to release technology when Microsoft catches up. That’s just a poor excuse and an indication the chip manufacturer doesn’t have a real understanding of what software is being used and where in particular markets.
Sorry but unless I’m mistaken Microsoft doesn’t own Intel and Microsoft’s Windows XP 64 is still in Beta testing.
The x86-64 version of Windows XP is still in BETA, but an IA-64 (Itanium) version of XP has been out for quite some time. Nobody uses it though, at least not for workstation/deskstop stuff.
http://www.microsoft.com/windowsxp/64bit/evaluation/overviews/2003/…
Oh, I’m not disputing that the performance of SGI’s itanium machines is extremely impressive. (I read the article here a few weeks back and … wow!) SGI’s machines are also blood curdlingly expensive, too.
But, if no other major computer manufacturer is making steady use of the the chip, is SGI’s business enough to keep Intel making it? I doubt it.
SGI’s sales have declined and the company has hemoraged money the past 5 years. Itanium & Linux were to be the big things saving them (in terms of making their machines more interoperable with other servers and cheaper) but if the itaninum goes, what then?
considering microsoft had Dave Cutler working with AMD on that processor design.
I imagine other shops that sell Linux servers on the high end will probably do okay with Itanium as well, if the performance continues to improve. Linux isnt’t as big of a problem when moving to the new IA-64 ISA because everything can be recompiled for Itanium in short order.
There _is_ logic in what you say however. I wonder how worried SGI is, now that Intel has their AMD64 clone, that Itanium may eventually be scrapped.
I for one am just thankful that, for SGI, the Windows NT, and Fahrenheit API days are over. I for one am pulling for them. They have donated some cool code to FOSS.
I just wonder what might have been if IA-64 was for the masses. An all new architecture to play with… and no BIOS… *sigh*
I wonder how worried SGI is, now that Intel has their AMD64 clone, that Itanium may eventually be scrapped.
Kind of off topic, but for the time being, Intel’s rip of AMD64 is seriously inferior technology. Sure it’ll definately improve over time, but right now, it just doesn’t compare.
AMD64 (both Athlon 64 and Opteron) have integrated DDR memory controlers, and NX support as only very minor examples, not to mention the fact that they were designed from the beginning to be dual core processors, and have no need to mess around with scheduler inefficient SMT technology, unlike Intel’s NetBurst.
It’s truely a shame that more big companies aren’t backing AMD more forcefully, because AMD64 is just a fantastic architecture.
Kind of off topic, but for the time being, Intel’s rip of AMD64 is seriously inferior technology. Sure it’ll definately improve over time, but right now, it just doesn’t compare.
AMD64 (both Athlon 64 and Opteron) have integrated DDR memory controlers, and NX support as only very minor examples, not to mention the fact that they were designed from the beginning to be dual core processors, and have no need to mess around with scheduler inefficient SMT technology, unlike Intel’s NetBurst.
It’s truely a shame that more big companies aren’t backing AMD more forcefully, because AMD64 is just a fantastic architecture.
My next upgrade here at home will almost certainly be an AMD64 based system. They do rock. I wonder if they can keep up with Itanium at the very high end however.
Intel is going to get stomped good in the low to mid range market by the looks of things.
I also read that AMD’s dual core is very compelling, whereas for the Intel P4/Xeon, its more of a hack. Think I read that on Ars Technica.
During the 68K to PowerPC transition, PowerPCs ran 68K code faster than 68Ks, so it was a no-brainer. Itanium runs x86 code much slower than real x86s, so no transition is possible. Customers will not accept a short-term slowdown in exchange for a long-term speedup. There was no opportunity to move the entire x86 market to Itanium.
“Microsoft conceeds the Itanium 2 cluster market to Linux-powered Altix clusters?”
Not so sure about that. It may prove to be a death sentence for the entire processor line via Intels withdrawl of the product. We’re going to need to wait for an official response from Intel before we can do more than guess.
If u read intels documents.. u were never ment to run 32 bit code on the itanium. It was never ment as a bridging chip.. it was mealy like the Alpha running x86 code.. it would do it slowly but still useful for some clients. Obviously your ment to compile all your code for it.
Intanium isnt that bad.. it just goes to prove x86 compatibility is evertying these days. And the AMD opteron goes to prove x86 instruction set dosent hinder performance in any way.
Btw my Amd64 bit processor rocks.. oh and so does 64 bit windows beta.
I assure you that SGI is not in any way worried about the P4 64-bit offering (consumer/workstation). The processor is, again, targeted at a different market segment than Itanium (HPC).
Intanium isnt that bad.. it just goes to prove x86 compatibility is evertying these days. And the AMD opteron goes to prove x86 instruction set dosent hinder performance in any way.
The AMD Opteron goes to prove that you can’t just ditch your entire user base and go to a different platform. This is essentially what Intel did. Itanium is still a better architecture for a great number of tasks and is a much better architecture to x86, but you can’t whole heartedly say we’re going to change and expect everyone to follow.
AMD while a great processor still has the limitations of the x86 platform.It benefits however, by being backward compatible and therefore hasn’t got to worry about whether the software will run or not.
It’s a good interim processor to move to full 64-bit compliance.
Umm! Think not! Apple PPC 601-603 systems could only emulate 68K code at rough speed of on 68020 @ 16-25 mhz. Last series of 68K machines (Quadra 840AV,Quadra 630 etc.) were much faster at runing non-PPC code than early PowerMacs.
Cheers
I really want Intels Itanuim chicp to work out for them…..I fel bad they are getting such a butt wooping from AMD (dont get me wrong I love AMD) but ya..
Could this be the beginning of the end of Wintel and the beginning of WinAMD?
Bwah. Couldn’t resist.
would os x ported to amd64, running exclusively in 64-bit mode do well?
This is really a non-issue here. Microsoft is not a player in the clustered, middle, and high end server markets to begin with. They are a low to middle desktop, and group server player. They merely hope to break into the market already dominated by Unix and it’s children. Itanium is really not selling well to begin with, so MS sees no profit in making a port for Itanium based servers. Don’t blame them when they don’t even have a significant presence in that market anyway.
I may be on crack, but did Intel miss an opportunity with their IA-64 platform, by refusing to create versions of the chip for everyone, including consumers?
No, you’re simply missing the point that the Itanium wasn’t _meant_ to compete in the same product space that the Pentium, Xeon, Athlon, Opteron, etc do. At least not for a while.
I understand that Itanium runs x86 code pretty slowly, but wasn’t there a way for Intel to provide “good enough” performance for that window of time where new applications would be written that would take advantage of the new architecture?
They did. _For what it’s mean to do_, the x86 emulation is “good enough”.
Maybe Intel wouldn’t be facing the big problems they are having feding off the Athlon 64’s on the desktop, if they had been willing to be a little more courageous and willing to try something a little more radical.
Say what ? The “problems” (what problems ?) Intel are having is because they tried (assuming the Itanium was meant to be a consumer chip from the start) to be _too_ radical and create a completely new architecture.
As it stands, Itanium looks like it will be a very expensive CPU (cost for Intel to develop) for a very small niche.
Ah, but the margins in that niche are pretty big. Intel aren’t going broke anytime soon.
“Who, exactly, is making itainium computers?”
SGI and HP are Intel’s current primary customers for the Itanium line of CPUs. HP hasn’t been selling very many of them either. I ran across a quote somewhere that out of so many Itanium based systems HP supposedly sold, only about 1 in 4 were actually bought. The rest HP gave away to developers to write applications for them. Intel apparently wasn’t impressed and told HP to stop doing so. Whether it’s true or not I don’t know, just another Internet rumor that I ran across that I’ve never bothered to follow up on because I’m personally not really that interested in the sales figures on Itanium. *grin*
Incidently, that’s the same tactic IBM used with universities in the 80’s. They gave away hardware to the poorer universities to gain steady service contracts and getting computer savvy students experience on their hardware hoping to generate more corporate sales down the line. It unfortunately doesn’t seem to be working so well for HP and Itanium tho. Itanium is still in direct competition with the older x86 architecture. Over $1k for a single ENTRY level Itanium 2 CPU when you can buy serveral nodes for a x86 32 bit cluster for that… the cluster will generally win in many situations on price/performance if the application is designed for clustering. Easier to create a cluster aware application from already existing source, than to port an application to a totally new and non-compatible architechture.
Say what ? The “problems” (what problems ?) Intel are having is because they tried (assuming the Itanium was meant to be a consumer chip from the start) to be _too_ radical and create a completely new architecture.
I meant the problems Intel is having trying to compete against AMD with their crappy NetBurst IA-32 design (the Pentium 4). Intel is caught now because:
A.) They designed that monstrousity for clock speed and nothing else, so bascially the whole chip is nothing but a marketing gimmick. They don’t like to talk about the fact, that clock for clock, the NetBurst (P4) sucks.
B.) AMD designed the Opteron to be dual core from the get-go, whereas Intel’s upcoming dual core designs are a hack, and will offer nowhere near the same performance gains.
Perhaps if they had worked on making an IA=64 chip for consumers, they would have been able to compete squarely against Opteron from more of a position of strength. Otherwise they are in for a rough ride in the next little while.
MS has little to no share of the HPTC market. So few codes run on it regardless. Most HPTC codes are designed to be ported or have been ported 10 times already.
MS a small time player in HPTC is saying it won’t port to Itanium another small time player in HPTC. Who cares?
The truth about Itanic is that it is a good chip and intel has destroyed it.
MS has no idea about HPTC and any OS that requires a console is hopeless for cluster farms. Trust me it requires a console. former linux cluster admin turned windows administrator…. Thank god for surf….
Unless Microsoft really wants to compete with high-end Unix. Itanium 2 boxes are hot and loud – best suited for the server room. I have 6 of them right now, running Red Hat and HP/UX and while they make great servers I could not see one on my desktop.
Still, for most applications I do not see what an Itanium 2 will give you. Zeons are much cheaper and are fast as hell.
It’s not like people actually use Microsoft on clusters anyway, and if they are too lazy to optimize their software to run on Itanium 2 that’s their problem. Opteron is easy to port to because of the 32-bit compatability. They can do it slowly, unlike Itanium 2 which would be a significant rewrite because it lacks any decent 32-bit mode.
I would think porting an OS would be harder than an App. MS has had a version of XP 64-bit for Itanium for a long time. So that means the writing is on the world for the Itanium.
Even HP, perhaps the strongest proponent of Itanium, cancelled thier Itanium workstation line inlieu of AMD64 and EMT64.
I think the Microsoft move is another indicator that Itanium will be relegated to a very very niche market. Any CPU without volume will be expensive, More so if there is a lack of software to run on it.
No, you’re simply missing the point that the Itanium wasn’t _meant_ to compete in the same product space that the Pentium, Xeon, Athlon, Opteron, etc do. At least not for a while.
Tell that to Intel. They developed Itanium to replace their 32-bit x86 cpus. Now since that didn’t work out they followed AMD’s route.
They are still trying to unify thier Xeon and Itanium chipsets aren’t they?
Ah, but the margins in that niche are pretty big. Intel aren’t going broke anytime soon.
For vendors who make the Boxes yes. But not for Intel. Itanium is expensive to produce with yields many times less per wafer than xeons. Yield dictates price and I don’t think Intel can jack the price up for a low demand CPU. So they are heavily discounting it to increase sales. They were even giving them away at first. The margins are slim if not negative on the Itaniums.
Especially with HP cancelling thier workstation line to support AMD/EMT64. And now MS, I would think Intel is in pain.
Good for AMD!
I meant the problems Intel is having trying to compete against AMD with their crappy NetBurst IA-32 design (the Pentium 4).
Well, until quite recently they’ve been doing pretty well. Somehow I doubt intel is shaking in their boots
.
AMD have a lot, lot less marketshare than intel.
They designed that monstrousity for clock speed and nothing else, so bascially the whole chip is nothing but a marketing gimmick. They don’t like to talk about the fact, that clock for clock, the NetBurst (P4) sucks.
Clock for clock is an almost completely irrelevant metric. If chip A has 1.5x the performance “per clock” than chip B, but chip B can scale to 3x the clockspeed, chip B is faster, period.
AMD designed the Opteron to be dual core from the get-go, whereas Intel’s upcoming dual core designs are a hack, and will offer nowhere near the same performance gains.
If intel can get their chips performing at the same level, _how_ they do it isn’t going to matter to anyone important.
Perhaps if they had worked on making an IA=64 chip for consumers, they would have been able to compete squarely against Opteron from more of a position of strength.
The IA64 would never have competed with the Opteron, “consumer version” or not. The Itanic’s weakness is its x86 compatibility.
Otherwise they are in for a rough ride in the next little while.
Like I said, I doubt intel are going to go broke anytime soon.
…To finally move on, and throw away some of the legacy crap.
Clearly the x86 instruction set is now many times larger than it needs to be, and yet we’re going to end up extending it for another generation now. It was a brave move for Intel to try and do something better; shame it doesn’t seem to be working out.