Merced: Our industry expects...

Intel's Pentium III may have got a lukewarm welcome, but Merced -- its 64-bit high-end engine -- is expected to be a pivotal step in the chip behemoth's roadmap.

New tricks for old dogs?

Merced represents a totally new direction for the chip maker and some analysts claim IA-64 could surpass the fastest RISC architectures: "We'll get to the point where it competes seriously with the Alpha RISC," said Linley Gwennap, vice president of the US chip analyst, The Microprocessor Report.

That's good news for organisations working on demanding projects. The power of 64-bit architectures and the accompanying operating systems should make data warehousing, e-commerce and video-conferencing applications sing. "64-bit offers significant performance gains," said IDC senior analyst, Lars Rasmussen. "I can see Merced being used in heavy applications such as super fast rendering of 3D animation."

Merced could also significantly alter the workstation landscape. If Microsoft continues to be delayed with Windows 2000 -- which has some 32 million lines of codes -- more Unix vendors may port to Merced or its much vaunted successor, McKinley, which many experts believe is the real star on the Intel roadmap.

Currently there are seven operating systems being migrated to Merced including Linux. Sun Microsystems, Compaq, SGI and Hewlett Packard are all in the 64-bit arena and Sun's port to Solaris in particular is a big endorsement for Merced as the company's customers portfolio includes the likes of NCR, which has high-end banking, retail and telecoms customers.

And Microsoft is already disadvantaged as the industry gets savvy about the 64-bit architecture, warned Rasmussen. "While others are getting valuable experience in this field, Microsoft is still prototyping Windows 2000."

laying the foundations

Intel has also been investing heavily in software startups in an effort to ensure the success of IA-64 and to cement its role as a serious player in the corporate computing market. Poor software support at the launch of the MMX family taught Intel some valuable lessons and it's keen not to be caught out again.

But while the perception is good, it's worth remembering Merced is two years late. "It's very complex - a new architecture need new tools and compilers. I would not be surprised if we have to wait till 2001 for it," said Gwennap. Rasmussen supports this view: "We expect small volumes by mid-2000, but the dynamics of the workstation/server market are similar to the desktop market. You need volume."

Further down Intel's 64 roadmap is McKinley. If Merced fails to give Intel a lead in the market, this is hotly tipped as the one that could. Customers may end up just testing on Merced, and hanging on for McKinley. But as Gwennap points out: "Who is to say that McKinley will be on time?"