stribe, I think the 432 is a better Intel antecedent for the Itanic than the i860. The 432 had a similar long delayed , much hyped gestation, ending with inadequate performance when it finally shipped. Of course, "inadequate" was an extremely gross understatement for what eventually shipped as the i432, that one had to be the dog of all dogs, and I don't think it ever made it into a shippable end system product. However lacking the Itanic may be relative to its contemporaries, it can't possibly be as bad as the 432 was, compared to the alternatives of the time.
The i860 was a funnier story, my understanding is that it was originally designed for graphics cards, but the marketeers got hold of it and sold it as something else. I imagine trying to use a contemporary graphics engine as a general purpose CPU would be similarly effective. I think the i860 may have been the beginning of various Intel compiler games, though. One of the cheezy little benchmarks of the day, dhrystones, was susceptible to this trick where a "strcpy" call could be replaced by an 8-byte double fp load/store, given a "clever" compiler with a -dhrystone switch. What made this extra funny was that the i860's exposed pipeline was basically impossible to compile for otherwise.
edit: on second thought, that last line makes me reconsider, maybe the i860 is actually a pretty good antecedent for the Itanic too, in its own way.
Cheers, Dan. |
