HWP is dependent on IA64, which could turn out to be a disaster for INTC and HWP (plus others).
Nothing exotic. These guys are betting the farm on a major architectural departure that's unprecedented in complexity for commercial devices (if you look at the system as a whole including the compiler). There are a lot of variables and it's far from a sure thing. It's late and slow and too hot. If it doesn't work, or the architecture burns up all the synthetic benchmarks but isn't fast enough for the commonest real-world cases, or it doesn't yield enough for a while, or whatever the glitch happens to be, then the farm is lost.
And even though they will make it up eventually (because I have great respect for Intel as an engineering company), this thing they have undertaken is so hard that any way you look at it, it represents a huge gamble.
To be fair, the converse is also true: IA64 could be a home run, there's no doubt. But with something that complex, if it didn't go through a lengthy baptism-of-fire period, it would be a first in my experience. And the longer it takes to get critical mass, the lower the liklihood that it ever will.
To some extent the same is true IMHO for the Ultra 3; except the Ultra 3 is an incremental improvement and thus depends on ordinary blocking-and-tackling logistics for its execution, whereas the IA64 is trying to put a new paradigm into play, and that's a lot harder.
You know Tony, I have no idea what's going to happen. I'm just very wary, after many years of experience, of people's tendency to talk about announced technology products as though they were real, long before they are. Because regardless of how much money you spend, some of them never become real. In that vein people talk about, and have been talking about, IA64 as though it's a rock-solid here-and-now threat to Sun, whereas to me it looks like a high-risk project just emerging from the vapor.
Regards,
--QS |
