By 2007, Intel's Itanic processors will not only deliver a 50-100 per cent performance advantage over the chipmaker's Xeon line, but both server chip families will cost the same.
So said the chip giant's EMEA strategic marketing manager, Alan Priestley, today.
Essentially, the move signals not so much the demise of Itanic at the hands of the now 64-bit enabled Xeon, but a move to push OEMs in the over direction.
According to Priestley, while Itanium today costs 30-40 per cent more than Xeon, for which buyers get around 30-50 per cent more performance. End users will clearly pay a lot more for a given Itanium system than a Xeon one, but that's roughly the difference for the basic hardware, when processor numbers are matched, he said.
Complete Itanium systems typically costs significantly more that Xeon rigs - by a factor of more than ten, according to market watcher Gartner's numbers - and given the roles the two processor families are typically being put, that's unlikely to change significantly over time.
So Intel is essentially planning to subsidise Itanium-supporting server OEMs' margins. They'll be able to continue charging through the nose for Itanium-based hardware, but the chips and chipsets themselves won't cost them any more than their commodity Xeon products do.
The upshot for Intel: Xeon largely becomes the choice of customers who need to support legacy 32-bit apps and possibly non-mission critical 64-bit code, while Itanium is the standard platform for all new 64-bit roll-outs, from the low end to the high.
Such an outcome is predicated on Itanium spreading downmarket rather than Xeon expanding the other way. That's certainly the approach that AMD is taking, and it says something of the threat Intel perceives its arch-rival presents here that it needs to bring Itanium's heavy artillery to bear on the x86 market.
Priestley played down EM64T, Intel's recently released AMD64-like 64-bit extensions for x86. To him, the technology is no more revolutionary than the addition of SSE 3 or a faster frontside bus. He also claimed that 64-bit x86 apps won't really be ready until mid-2005. Intel expects to have its entire Xeon line supporting EM64T by that date, but if it'll take that long to clean up 32-bit code for 64-bit compilation then to get it revalidated, why ship 64-bit Xeons now?
Because AMD is a threat - customers may not yet be adopting Opteron wholesale, but if they use it to evaluate and develop 64-bit code, it makes it more likely they will subsequently use Opteron to run the finished stack. In short, EM64T is more about providing a reason not to buy AMD than a reason to buy Intel.
Long-term, that's not an durable strategy, so the better to draw a distinction between Intel server processors and their AMD alternatives, Intel will use Itanium rather than Xeon, with its 'better performance, same price' scheme. Hence this week's price cuts and the prospect of deeper ones as Itanium goes to 90nm and, later, 65nm.
Intel's strategy is also a tacit acceptance that Itanium has failed to hammer high-end Risc. Priestly cited Gartner numbers that suggest Itanium has a 3-5 per cent share of the high-end iron market. While he claimed Intel's forecasts anticipate "significant inroads" into the Risc market, he didn't offer up a timescale.
Certainly, Itanium has won the support of the big system sellers, with the exception of Sun, but none has yet switched exclusively to the Intel processor. And those who have said they intend to do just that appear to be taking rather longer to do so than they once forecast.
Intel has already expressed its intention to develop a common chipset platform for Itanium and Xeon, and that's really the first step toward price parity. ®
Intel Itanic prices chopped
Intel: common Xeon, Itanic chipset by 2007
Motorola plumps for HP Linux-on-Itanium boxes
HP unveils Unix roadmaps
Intel gets quiet about the competition
NASA splashes out on shiny supercomputer
Intel's CEO gets blunt about poor execution
Sun's Solaris shines on Itanium