Intel recently christened “the chip formerly known as Merced” with the newly coined name ITANIUM. (Yes, they really do want it in all caps. Tough.)
Remembering the blockbuster movie, I wonder if Itanic might not be a better name. After all, if ever a processor was over-designed, oversold, and had the potential to doom a chip manufacturer, Itanium is it.
Itanium is Intel’s next generation processor, one that they assume will replace the Pentium family in about six years time. When Itanium finally begins shipping, probably in 2000, it will be promoted primarily as a server CPU, not one for desktop or laptop computers.
Over the past several years, Intel has absorbed more and more RISC ideas into its x86 family of processors, but they still run the same old instruction with a few additions. This is in stark contrast to Apple’s move to a RISC processor in 1994 with the introduction of the PowerPC 601 processor and the Power Macintosh line of computers. Apple has been up and running with RISC for nearly six years; Intel doesn’t anticipate Itanium becoming a dominant desktop CPU for another six years.
Why is that?
Partly it’s because Intel is making two jumps forward at the same time: adopting VLIW RISC architecture and migrating from 32-bit to 64-bit processors. (Apple made the leap to RISC in 1994; the move to a 64-bit PowerPC processor will be more evolutionary than revolutionary.)
Besides, Intel completely dominates the personal computer market. Adding up all the PowerPC, Athlon, Alpha, and other systems probably still leaves Intel with 80-90% of the market.
Moving to an entirely new processor jeopardizes that.
So part of the Itanium game plan is building Pentium emulation into the processor. This is a bit like what Apple did to assure backward compatibility with 68K software on PowerPC Macs, but Apple left the emulation in software, allowing for future improvement. By building it intow the CPU, Itanium has to get it right the first time.
Worse yet, Microsoft it throwing up its hands in despair of getting Windows running on Itanium. That’s because the VLIW paradigm is very different – everything has to be optimized in software. Otherwise it can’t take advantage of all the pipelines and queues of the Itanium processor.
This puts the performance burden on the software developer. In the rest of the industry, it’s the processor maker who keeps pushing the envelope, but Intel doesn’t want to do things that way. Odd, since they are the hardware company, but it does effectively move the performance burden from Intel to software engineers, especially compiler builders.
In effect, Intel has created its own Titanic: a hulking monstrosity with lots of power, lots of pipelines, lots of registers, and lots of parallelism. If it works, it will be truly awesome (of course, by then we could have the G5). But if Microsoft can’t make it run Windows, Itanium meets its iceberg and becomes Itanic – the business crowd simply will not follow Intel if that means abandoning their beloved Windows.
Women and children first.
Update: Itanium first shipped in June 2001, and only a few thousand systems based on the original Merced version were sold due to a lack of software and poor performance compared to competing processors. Itanium has not been the runaway success Intel had hoped for. In fact, it’s a very small part of the server market.
- Merced Is RISC, David K. Every, MacKiDo. “…it is no more a revolution than a car company adding a few more colors, or increasing their horsepower.”
- Itanium: Intel Christens Its Newest 64-bit Chip, Alexander Wolfe, Byte. “Call it what you will – Merced, IA-64, or Itanium – Intel’s Itanium-class processors will not be aimed primarily at the desktop or even the workstation.”
- Chip Vendors Showcase Road Maps at Microprocessor Forum, Mark Hachman, Byte. “Intel’s architecture is based upon instruction-level parallelism, placing heavy burdens on the software compiler to translate software applications into optimized IA-64 instructions.”
Keywords: #itanium #itanic
Short link: http://goo.gl/PiFGTO