The product a company offers does not necessarily define its customers need. Railroads are the canonical example. As new technologies entered the market, trucking and air freight companies ate the railroads lunch as railroads discovered—surprise!—they were in the transportation, not the railroad, business. Intels Itanium 2 is headed down the same dead-end track, and Dell has decided to jump off now. Corporate buyers might do well to follow Dells example.
We can understand Intels logic. In the two decades past, people clamored for mass-market servers based on commodity PC technologies. Why not move those same economies to the enterprise sector with a new 64-bit design? If the only way to make it fast and make it cheap was to throw complexity problems over the wall to the writers of compilers and operating systems, what of it? If its good for Intel, it must be good for the industry. Right?
But while Intel was taking much too long to bring the Itanium to market, customers were finding other ways to achieve the goal of cost-effective and easily managed capacity. Virtual servers, carved on demand out of bulk-priced mainframe server cycles, combined with a new generation of network-based tools to provide the convenience of departmental servers without the nuisance of widely dispersed boxes tucked under too many desks. IBM and Sun are already shipping 64-bit machines to address the genuine enterprise requirements of enormous databases; AMD is on the verge of delivering on its promise of a smooth 64-bit migration path for the 32-bit x86 code that will still be with us and still be in need of continued price/performance improvement, for years to come.
Unfortunately, Hewlett-Packard and Compaq cannot be included among the 64-bit system vendors above, since those now-merging companies plan to move their users of PA-RISC and Alpha—good architectures both—to Itanium. That loss of diversity was, and is, one of our concerns about that merger. Intel must demonstrate to the IT community that its new chip, with new operating systems and new compilers taking on new challenges of compile-time concurrency analysis, is a solution to any problems other than Intels own.