'

Itanium slips further into trouble

With each delay and feature reduction, Itanium's future looks bleaker and bleaker. Faith in the future may not be good enough

Eschatology is rarely listed as a key competency on the CVs of the IT elite. The study of the end times, it has a sorry history that's been repeated again and again. Various sects and religions confidently predict The End Of The World and a bright new beginning: come the appointed hour and all that happens is that by no magic at all fervent believers are turned into silly puddings.

The pudding hour draws ever closer for Itanium's disciples. For the first time since the seismic upsets of 2004, when Intel indulged in very public Maoist self-criticism over its inability to deliver, there is blood on the fab floor as projects are cancelled or delayed. Some changes make sense: Whitefield was a four-core Xeon due in 2007: nothing wrong with that, but it's been replaced in the same timeframe by Tigerton — another four-core Xeon but with extra features aimed squarely at Opteron. Intel has been slow in the past to react to competition, so evidence to the contrary is welcome.

But it's hard to see Itanium's roadmap alternations in such a good light. For years now, the coming of Montecito has been held out as the resurrection of the brand. Running at more than 2GHz with a super-fast bus and advanced power management, this dual-core scamp would put Itanium right back on top of the heap when it came out in force at the beginning of next year.

That's not what's happening. The new power management stuff doesn't work which means the high clock speeds have had to be abandoned, and it'll stick with the old bus. Even with these hopes abandoned, the chip's still being pushed back to the second half of 2006. All those fine features will turn up, we're promised, just even later.

But later may be too late. The knock-on effect on Montvale (2007) and Tukwila (2008) puts the competition in an even better light; 2007 will see IBM's Power6 processor, which is rumoured to have all manner of interesting features such as x86 compatibility and mainframe-like failure resilience. Neither is in production, but with each Intel delay IBM's strategy looks more and more attractive. You can feel the buzz building over Power6: the only things buzzing around Itanium are bluebottles.

Intel may stick to its guns in predicting a bright new beginning for Itanium, but to the rest of us the end times look dark indeed.