Microsoft Phasing Out Support for Intel Itanium

You don't really believe Microsoft, do you?

The reasons for phasing it out are due to the poor performance of the Microsoft OS and SQL Server. Itanium chips running HP/UX have roughly twice the performance of the same machine running Windows and SQL Server.

Since that's pretty much all Microsoft sells to in that market, it's a useless product.

Also, why would core counts going higher be related to this? By that measure, x86 has always been ahead of Itanium. The Tukwila actually makes the Itanium more competitive than it has been for a while.

Microsoft is talking their normal nonsense. Their product sucks and can't compete, so they had to leave the market. It's not the first time that's happened.

TA152H: Actually, it's the performance of Itanium that sucks, not the other way around. Microsoft said it sucks, Linus Torvalds(creator of Linux) said it sucks, who else do you need to hear it from?

Also, MS had only a tiny sliver of marketshare on Itanium platforms: most ran HPUX (HP's Unix distro) if I remember right. Little loss for MS to drop the platform.

Itanium was Intel's bad idea to take over the server processor marketplace. It was proprietary as hell and they had no plans on licensing its instruction set to any other manufactures (AMD). Thankfully it didn't work out, it sucked performance wide and required everything to be compiled for it naitivly.

Itanium is actually a great idea it's just phasing out the x86 instructions is downright tricky. I'm surprised why x86 manufactors haven't introduced new cores with only x86 instructions. This would be the easiest way to solve this problem in the long run.

err I mean cores with x64 instructions....

Who needs EFI of 64bit? -sarcasm

The area Itanium sucked in was x86 because it was not naitive. It was emulated instead which cut 20% performance. Microsoft is mainly a x86 OS so it would make sense.

Itanium will go forward in the pure x64 market.

Itanium does not have any features most x86 CPUs have (branch prediction, speculative execution etc.) and relies on the compiler to put this into the executable. Works great in theory, fails (failed) practically because the compiler is too hard to develop. It's not the hardware that's bad...

OTOH, it would be great if MS used the resources to support ARM. Having windows 7 on ARM wouldn't be so bad.

And so the beginning of the end for the Itanic

Watchman: ‘Captain, ice berg ahead!’
Captain: ‘Don’t worry, we’ll have plenty of time to steer around it’

Fare well Itanic, you probably won’t be missed.

This was supposed to be the next biggest thing, what a flop....

Sounds like the typical phasing out of old tehcnology that will be replaced by new technology.

Ahhh... Itanium looked so good on paper.
Kinda like a real GPU from Intel...

Intel invented the x86 architecture,
AMD kick started the x64 architecture.

Fixed.

The only "true" 64bit CPU from Intel/AMD was the Itanium.

this is tech news I like to see Marcus

Linus Torvalds isn't always right, despite your apparent worship. He and another fellow have argued back and forth on the Itanium. Linus believes even a bad instruction set is OK, because over time it gets understood and works OK. Volume and time makes up for a lot of the problems, according to him.

The problem with Itanium is it's always way behind in manufacturing technology. It just moved to 65nm, whereas Intel has been using 32nm for x86 for months.

But, your point is irrelevant. Itanium has always been slower, for reasons mentioned above, but Microsoft supported it. Tukwila closes the gap, it doesn't make it bigger. But, the problem is, HP/UX and Oracle are roughly twice as fast on the Itanium as Windows and SQL Server. So no one wants it. Well, not enough people want it to make it feasible for Microsoft to keep developing it.

64-bit is generally slower than 32-bit, and almost completely unnecessary for many apps. They are also lousy like x86, so it's not like you could get rid of the decoders and have a nice RISC instruction set. You'd still have the problems, with little benefit. Also, did you forget BIOS screens are 16-bit (real mode)?

The problem here is that Itanium costs 8 times as much and uses more than twice the power. Using x64, a company can cut costs by 75%, cut power by 10-15%, cut cooling by 10-15%, and get the same performance.

There are just guys out there who don't want to change because then they wouldn't "be needed". They want to secure their jobs rather than save the company money. These same type of guys are resisting the death of both OS400 and many Unix applications.

Wintel/x64 can keep up in many of these arenas, although not all. Microsoft and Intel are catching up in the areas where they don't currently keep up. All the while, there are guys out there resisting the change because they wouldn't know how to take care of the new systems (even though it is much easier to learn, thus the labor market for Windows Server admins is wider and cheaper) and they'd lose their exclusivity.

People who hold back change just to keep from having to learn a new job make me sick. The wave is coming. There is no getting out of the way. Adapt or die.

no it certainly isn't. 64-bit software has access to twice as many registers and is executed faster on AMD processors. Most floating point instructions execute from 1 to 5 clock cycles faster on AMD processors, and SSE/MMX instructions are 2 clock cycles faster almost all the way across the board. Granted, Intel chips don't have this advantage, but there is no penalty for running in 64-bit mode. The register advantage alone is enough to give it a 3-5% advantage.

darnit, quoted the wrong post. That was supposed to be a counter to the post about x64 being slower than x86.

Another nail in the HP/HPUX "Integrity Server" coffin.

That would make a nice wallpaper if it was high res...

and Open VMS

The Itanium always had a better, forward thinking architecture. Problem was, Intel wanted to keep it to themselves while AMD figured it would be cheaper for them to just ad 64-bit extensions to existing chips which saved software devs a lot of time. Also, the Itanium was architecture was alot different from the x86 chips Intel traditionally supported so the initial compilers sucked. It took a while for them to develop a compiler to properly take advantage of the architecture, but by that time the chips themselves were behind.

So yes, I like the ideas behind the Itanium Architecture. The reality that x86 designs are rather inefficient is hidden by the fact that AMD and Intel keep pushing out better, faster chips with more cores. Sadly though, the architecture of the Itanium was too far ahead of the software, and AMD took advantage of that with it's Athlon 64 based Opterons making Intel follow suit with 64-bit Xeons and thus pushing Itanium to the sidelines.

in theory it's a good idea, however in practice, most software is still 32 bit even in the server space, and the Itanium is forced to emulate these things, where a cheaper x86-64 cpu can do the 32-bit programs natively..it's hardware that software just isn't ready for

Meh, Itanium was practically dead when it hit the market. Finally Microsoft woke up and just cut the development cost.

I was never a fan of Itanium, IBM always had the forward thinking chips with its POWER and POWERPC line of processors. If we were to shift to a new processor technology, I'd rather us use IBM designs.

Either way, you're wrong. You're forgetting that you put pressure on the memory and caches, and get fewer hits because of it. Consequently, you wait more on memory, so 64-bit is going to be slower in many cases.

I'm guessing from this post you don't understand mission critical type of applications, and are really talking about mainstream stuff. Your post makes sense only within that context.

Nehalem-Ex comes a bit closer, but it doesn't have all the reliability features that Itanium based systems do. So, you need something to stay up, and give you the correct results, as close to all the time as possible, x86 can't do it. It's not intended for that, and there are inherent compromises, some of which reduce performance and increase power use because of the way the circuits are designed. x86 isn't willing to lower performance and increase power use to do it. It's a different market.

LOL... and intel never fails or screws up....

Intel and HP put so much money into this chip, and even recently in 2010, they were talking about its future as a high-end server class chip. Well... maybe it has a future, running MS-OS wasn't its strong suit anyway.

When AMD started shipping 64bit CPUs about... 6~7 years ago, it quickly blew away Itanium in sales units sold. The AMD Opterons were (A) Faster (B) easier to program (C) requires no custom hardware (D) costs a lot less.

It was years before intel admits defeat and soon add 64bit abilities to thier XEON and of course consumer CPUs. So yeah, intel is using AMD-code for 64bit.

Whole AMD is having severe problems with their CPUs in the server market... they are a lot healthier than HP with Itaniums.