Joe, Re: "And in no time, x86 chips (P4, K7, Hammer) will leave all RISC chips behind. Itanium is way back in performance, way high in cost, completely unproven. How can something so uncertain be the path of least resistance?"
Nobody knows how well x86 chips can do in servers better than Intel. Yet there is stark resistances in some segments, simply because they want the robustness and high availability of a 64-bit RISC CPU.
Now - don't get me wrong. I believe as much as you do that there is nothing inherent in the x86 ISA that prevents anyone from building the most reliable x86 class machine with the highest uptimes and best support. However, there is a big perception among these segments that a dedicated 64-bit CPU is the only way to go (64-bit extensions on top of x86 apparently isn't good enough).
The good news for Intel is that a lot of infrastructure support, and hardware / software enabling is going into IA-64. All the major OEMs are involved, and it won't take long before the support for IA-64 exceeds that of many popular RISC architectures.
I don't doubt that this will still require continuous push from Intel to maintain momentum, but I do think that as soon as robust systems start appearing, the results will speak for themselves. I think the worst mistake Intel can make is to sell Itanium short at this point. Their only chance is to push even harder, and gain respectability in the high end segments, where power consumption is less of a factor, customers demand performance at any cost, and nothing short of a true 64-bit ISA will do.
After that, they will do what Intel does best - proliferate their design to more and more segments, while adding value and features along the way. In actuality the beginning will be the hardest part, but once the engine starts, the momentum will win it the race.
wbmw |
