[NEWS] - Itanium's toasted, Intel almost officially admits

[CougTek]

"Our research suggests Itanium is in for a rough ride," Whittington writes in a recent research note. "Intel is now saying it will "go with the market" on 64-bit x86, thus is destined to unveil one when they think the market will ripen, which we judge as mid-2004 for volume delivery in 2005.

"This fateful step will necessarily consign Itanium to low volume, high end computing solutions as a mainstream, high volume x86 horse-race develops between Intel and long-time rival AMD that will push 64-bit x86 performance well into the low-mid range server territory for which low power Itanium was slated."

Na Na Na Naaa!
Na Na Na Naaa!
Hey Hey Hey!
Goodbye!

Glad to see the end of a bad design that should have never made it out of the design lab.

News source

 

[Mercutio]

I don't know about that, Coug. IA-64 was Intel's well-meaning effort to break with the 20-year legacy of x86. Its main fault was being uncompelling from a price/performance standpoint.

Of course, EVERYTHING is uncompelling from a price/performance standpoint, when compared to hordes of cheap x86 chips and in terms of raw performance, well, Intel cut its own throat with Alpha.

Itanium 2 chips are still technically faster than Opterons (at least in Dhrystone/Whetstone benchmarks). That's gotta mean something, right?

 

[sechs]

The fact of the matter is that going with Itanium is analgous to going with Sparc or PA-RISC. It's completely different, and so you have to mostly start over with your software. Why pay more to get less?

 

[Tannin]

I confess to still being utterly mystified by Itanium's design and market philosophy. Why, when you have a chip that given its development cost is by any rational standard an integer-performance dud, but an absolute humdinger when it comes to floating-point stuff ... why on earth would you position it as a server part?

 

[.Nut]

I confess to still being utterly mystified by Itanium's design and market philosophy. Why, when you have a chip that given its development cost is by any rational standard an integer-performance dud, but an absolute humdinger when it comes to floating-point stuff ... why on earth would you position it as a server part?

A server with these performance characteristics would best be deployed as a compute server (think Cray, SGI).

Something else the Itanium does better than anything else (currently) is encryption / decryption. This is, of course, just one of the many forms of compute serving.