Intel to spotlight new Itanium: 'Poulson'

(continued from previous page)

server designs, including Fujitsu, Silicon Graphics, Hitachi, NEC and Unisys.

The current Itanium processor is code-named Madison. A new version code-named Montecito is expected to boost performance significantly upon arrival at the end of 2005.

Montecito has dual processing engines, or cores, and each core is able to process two simultaneous instruction sequences, called threads. Madison has a single core that executes a single thread.

Tukwila take 2
Intel has been mum about most Tukwila details but has said it will arrive in 2007 and will have four cores. But it appears other elements of the chip have been changing to better compete against IBM's Power family of processors. A IBM Power5 system trounced an HP Itanium machine and all other rivals in a recent server speed test.

Tukwila appears to have fewer cores, which will execute individual operations more quickly, rather than multiple cores that can handle more jobs simultaneously but not as fast, Illuminata analyst Gordon Haff said based on a discussion last week with Rich Marcello, who leads HP's high-end server group.

"Tukwila has been redefined to have fatter cores...Intel is refocusing Tukwila on being very explicitly a competitor with Power," Haff said. "With Itanium now being very explicitly designed for the high-end back-end jobs, HP clearly thinks it makes sense to have fatter cores rather than maximum threading."

In recent years, Intel reoriented Itanium for higher-end servers, as opposed to the entire server market, using RISC, or reduced instruction set computing, processors. Today's mainstream RISC chips are Power from IBM and Sparc from Sun and Fujitsu.

IBM's Power5 has dual cores, each able to execute two threads.

Sun is working on a processor code-named Niagara, due in 2006 though being previewed this year. It has eight cores, each able to execute four instruction threads. That makes it more similar to the earlier Tukwila version, which began with four cores but which a source familiar with Intel's plans said eventually would reach 16 cores.

Intel declined to comment on whether it was revamping Tukwila. But the chipmaker has made similar moves in the past. Intel shifted Montecito to today's design in 2003, substituting it for an earlier model so the company could release a dual-core Itanium sooner.

[bobby_brady]

Itaniums remind me of Alpha

Intel can make it as fast as they want, doesn't matter if nobody is buying Itaniums.

[System Tyrant]

well

I won't say that Itanium is the best, but at some point I believe the x86 technology is going to become outdated nomatter what you do to it. I think at some point we are going to have to have a replacement. Maybe then (2070) Itanium will have a chance :).

[bobby_brady]

Itaniums remind me of Alpha

Intel can make it as fast as they want, doesn't matter if nobody is buying Itaniums.

[System Tyrant]

well

I won't say that Itanium is the best, but at some point I believe the x86 technology is going to become outdated nomatter what you do to it. I think at some point we are going to have to have a replacement. Maybe then (2070) Itanium will have a chance :).

[fred dunn]

Intel, wake up and smell the coffee...

Intels approach to Itanium is the same approach they took with the Pentium4 and their associated chipsets, either use slow SDRAM or use RAMBUS RDRAM because that's all we're supporting. Well that changed when nobody wanted RDRAM but they lost countless millions in the process.
Another instance is when AMD came out with the X86 64 bit extensions, Intels take was we are not going to do that because we we have the Itanium. Now they have implemented the 64 bit extensions into their chips but only after losing sales to AMD.
Itanium is not a bad processor but it is just not x86 compatible. When Intel finally integrates x86 microcode and compatibility (or at least very efficient virtual compatibility) Itanium will just be the Itanic.
When will Intel open their eyes and ears to the consuming public and develop for their consumers and not their research geeks? My bet is not soon.

Fred Dunn

[fred dunn]

Intel, wake up and smell the coffee...

Intels approach to Itanium is the same approach they took with the Pentium4 and their associated chipsets, either use slow SDRAM or use RAMBUS RDRAM because that's all we're supporting. Well that changed when nobody wanted RDRAM but they lost countless millions in the process.
Another instance is when AMD came out with the X86 64 bit extensions, Intels take was we are not going to do that because we we have the Itanium. Now they have implemented the 64 bit extensions into their chips but only after losing sales to AMD.
Itanium is not a bad processor but it is just not x86 compatible. When Intel finally integrates x86 microcode and compatibility (or at least very efficient virtual compatibility) Itanium will just be the Itanic.
When will Intel open their eyes and ears to the consuming public and develop for their consumers and not their research geeks? My bet is not soon.

Fred Dunn