By David Weisberg
Intel recently announced that Audi will migrate from RISC-based servers to its Itanium processor-based servers. The servers will be used optimize computer aided engineering (CAE) systems used to design and document in-car cooling and air conditioning.
When I read the announcement, a question came to mind: Is the Itanium simply too little too late?
It is hard to believe that Intel and Hewlett-Packard launched their joint development of the 64-bit Itanium nearly a decade ago. Until recently, I believed that this advanced processor was the wave of the future and would give technical computing a real boost in performance. Now I am not so sure.
Without boring readers with the details, Itanium implements a processor architecture that is internally extremely complex. More than that, it required advances in compiler technology to take advantage of these processor features. Intel and HP invested over $2 billion creating the Itanium and the compiler technology to support it. Originally, the intent was to be able to run current Intel-based code without recompiling it. That didn’t happen and as a result, there are relatively few programs available for use on Itanium systems.
Meanwhile, AMD has come out with the 64-bit Opteron that handles both 32-bit and 64-bit code. Computer manufacturers including HP are offering Opteron servers and several weeks ago, Intel announced that it would deliver 64-bit versions of the Pentium 4 and Xeon microprocessors.
The Itanium simply took too long to bring to market, was more complex than it needed to be and simply does not offer the performance advantage we expected. Intel will probably keep it in the product line for several more upgrade cycles but I doubt if it will ever see widespread use. Even the Intel press release that caught my attention simply mentions that Audi is installing Itanium servers with no mention of Itanium workstations.