Intel Removes Eggs from Itanium Basket
It was inevitable, but it's official now: After 10 years and untold billions of dollars spent on IA-64, Intel is going x86-64.
Intel announces 64-bit "Nocona" Xeon
It took them longer than I thought it would to give up on IA-64, but they essentially have. They'll still sell Itaniums at least for a while, but it'll be a niche product until it dies. Now the course of the x86 world is more or less set on x86-64 (no word yet on whether Intel is borrowing AMD's x86-64 ISA or rolling their own).
What does this mean for us? It means that the period of fragmentation and uncertainty in the x86 world over where to go next has ended, and Intel and AMD have gone with solutions that preserve the momentum of the x86 legacy, for better and for worse. So a window is closing, and IBM has to make sure they gain as much as possible before the x86 train starts rolling again. I think they can, and I also think IBM expected this: Intel didn't get where it is by tilting at windmills and chasing fantasies.
Intel announces 64-bit "Nocona" Xeon
It took them longer than I thought it would to give up on IA-64, but they essentially have. They'll still sell Itaniums at least for a while, but it'll be a niche product until it dies. Now the course of the x86 world is more or less set on x86-64 (no word yet on whether Intel is borrowing AMD's x86-64 ISA or rolling their own).
What does this mean for us? It means that the period of fragmentation and uncertainty in the x86 world over where to go next has ended, and Intel and AMD have gone with solutions that preserve the momentum of the x86 legacy, for better and for worse. So a window is closing, and IBM has to make sure they gain as much as possible before the x86 train starts rolling again. I think they can, and I also think IBM expected this: Intel didn't get where it is by tilting at windmills and chasing fantasies.
Comments
Then new Xeons are for a different market then Itanium. Xeons are for workstations and entry-level servers... and the Itanium is built for mid-to-enterprise-level servers... They are different markets.
It looks like there are only going to be minor differences between AMD and Intel's chips and Microsoft is already on their way to supporting both sets of extensions.
I can't imagine PPC picking up any more market-share from x86 now that a standard has been agreed upon - regardless of how lame that standard is once again.
Intel may have announced it, but they've got a lot of work to do before they go mainstream with x86-64, and recently Intel has taken the number one spot when it comes to creating the least accurate product release road maps.
Itanium sounds like it's been put on the slow track. Intel finally woke up from their dellusions of grandeur. The technology to kick butt in home computers is not quite the same that's needed to kick but in Enteprise.
Edit:
DMB can't spell.
Originally posted by \\/\\/ickes
Intel is not going to drop the IA-64 chips...
Then new Xeons are for a different market then Itanium. Xeons are for workstations and entry-level servers... and the Itanium is built for mid-to-enterprise-level servers... They are different markets.
That may be what Intel marketing is hoping, but I don't see that.
First, IA-64 was not originally intended for that. It was the future of the Intel platform, first in servers and then everywhere. Intel saw how widely the x86 ISA was cloned, and they wanted the whole market for themselves the next time around. So they deliberately designed IA-64 to be difficult to clone - which also means that it's difficult to evolve and update, which means it's only viable if it's generating a lot of profit. It's a strategy that depends on market domination. Itanium was the runt of the 64-bit market when it was Intel's only 64-bit offering. What fraction of its current dribble of sales will it enjoy with the 64-bit Xeon encroaching on it?
Nothing prevents a company from stringing a whole lot of Xeons together to make a big server. There are already companies offering these machines, with the 32-bit Xeons. The 64-bit Xeon will be able to slip into those machines and support their entire legacy while far more efficiently handling large tasks. The Itanium will... well, it'll get crushed between the new Xeon and the POWER series.
The only way Intel could forestall this would be to cripple the Xeon, since there's nothing intrinsic to IA-64 that makes it faster. But with the 970 breathing down the Xeon's neck, does Intel really want to gimp anything?>
They have a small problem here: Intel has zero credibility in big computing. IBM has decades of cred.
I will point out how if you were to downclock any Xeon to the speeds of an Itanium then the Itanium would kick some serious ass... I think Intel might be able turn things around when it comes to bumping up the clock speed of the chip and keep it for high-end server use.
What we've come to understand as the "standard" superscalar chip roadmap is pretty much stuck in an uncreative series of fabrication advances. Whatever the Cell is, I applaud it, because we do know that it will probably be different architecturally than what we're used to, and most likely one step further in computer history.
Originally posted by Splinemodel
I wish Intel just said: "F--- you, you're going to make Windows run well in IA-64." It's time for the world of Electrical engineering to retake the dominant position in the computer industry. It's dangerous and unpreductive for computer scientists to expect that computers will just get faster and faster and be able to run the exact same code.
First, any computer scientist worth the name is uninterested in that goal. Businesses, on the other hand, are keenly interested. So are end users.
Second, while I have no problem at all with ditching ugly legacy and advancing the state of the art, I'm glad IA-64 is dead. It was needlessly complicated and hobbled by anticompetitive design criteria. Now, if you want to mourn a next-generation architecture, consider Alpha. The world is poorer for the demise of that architecture. At the hands of Intel, I might add.
Whatever the Cell is, I applaud it, because we do know that it will probably be different architecturally than what we're used to, and most likely one step further in computer history.
Amen. The languages popular in commercial applications - simply because they've been commodified by businesses eager enough to avoid training costs that they require C or C++ - assume an architecture that's pushing 40. A significantly parallel architecture would force a transition to cleanly factored, truly object-oriented code, and almost certainly a change to one or more languages that elegantly express those traits, instead of being built on C. Believe me, computer scientists have been waiting for this day for a long, long time. Actor languages (OO languages designed for highly parallel architectures) have been kicking around for two decades now.
Another thing that comes to my mind is the way empires fall.
Originally posted by Tuttle
I think this is pretty bad news for IBM and Apple.
It looks like there are only going to be minor differences between AMD and Intel's chips and Microsoft is already on their way to supporting both sets of extensions.
I can't imagine PPC picking up any more market-share from x86 now that a standard has been agreed upon - regardless of how lame that standard is once again.
I don't think so. Won't it take awhile for Intel's 64bit tech to come down to the Pentium 4, especially with them still finishing up from the Prescott?
BTW, though it gets lost in my wordy ramblings, I called this one about 3-4 years ago, before there was ever any product, just based on Intel's and AMD's design direction. Legacy rules in X86 land!