> FX!32 was good compared to what was available elsewhere (including for Itanium IIRC). I am not saying it had to compete with fully with x86 on consumer desktops from beginning, but it would have been more stable position, especially with known upcoming work on 64bit-clean NT5.
I meant not good compared with what was available elsewhere, i.e., actual x86 CPUs. It didn't emulate privileged modes, so none of your operating systems worked, it required Windows NT, and it only provided emulation for Win32 x86 binaries, and it did not emulate all facilities either. Independent benchmarks are not easy to come by, but even the few results published by DEC were not fantastic. A 500MHz Alpha (likely EV6) would achieve maybe little better performance than a 200MHz Pentium (P5, not PPro)! EV6 did 2.66x specint95/MHz that Pentium MMX 200 achived, and 2.5x the frequency. Natively it would havehave been 5.32x faster on those benchmarks, but it was a little over 1.x. This is not the 40-60% performance claimed of the trivial byte benchmark result, this is 20% performance at best.
The slip was the original development AFAIK. HP later restarted a failed process shrink I think, but I can't remember if they finally got that done in the end.
I meant not good compared with what was available elsewhere, i.e., actual x86 CPUs. It didn't emulate privileged modes, so none of your operating systems worked, it required Windows NT, and it only provided emulation for Win32 x86 binaries, and it did not emulate all facilities either. Independent benchmarks are not easy to come by, but even the few results published by DEC were not fantastic. A 500MHz Alpha (likely EV6) would achieve maybe little better performance than a 200MHz Pentium (P5, not PPro)! EV6 did 2.66x specint95/MHz that Pentium MMX 200 achived, and 2.5x the frequency. Natively it would havehave been 5.32x faster on those benchmarks, but it was a little over 1.x. This is not the 40-60% performance claimed of the trivial byte benchmark result, this is 20% performance at best.
https://www.researchgate.net/publication/3214997_FX32_a_prof...
The slip was the original development AFAIK. HP later restarted a failed process shrink I think, but I can't remember if they finally got that done in the end.