Remember that in Grove's time Intel was getting clobbered in benchmarks by the DEC Alpha chip and was struggling to keep up with the PowerPC consortium. It was through their relentless pursuit of clock-speed at any cost that they prevailed.
Unfortunately they slammed into the 4GHz barrier earlier and harder than the competition and went into a spiral of confusion for nearly half a decade. Adding more and more stages to their pipelines was the primary strategy for pushing higher and higher clockspeeds.
What saved them in the end was the very pragmatic and practical Pentium III platform, the one introduced a few years before Grove retired, not the outlandish and massively power hungry Pentium 4 architecture.
Grove did seem to keep the company focused on relentless incremental improvement. This is in stark contrast to the later Intel which showed unbelievable hubris with their VLIW effort thinking that people would spontaneously adopt a whole new architecture just because. Maybe they listened to their partner HP way too much as that company is quite clearly utterly clueless.
You mean the Pentium Pro platform? Pentium III is just Pentium Pro with off-package L2 cache and SSE instructions. The P6 microarchitecture kept Intel in business from 1998 (when the P5 line concluded) until the Nehalem architecture launched in 2008.
http://www.intel.com/content/www/us/en/processors/xeon/xeon-...
I wouldn't call it dead yet. It's still paying dividends.
Not to take anything away from Grove, but a lot of the focus on deep instruction pipelines can probably be traced to certain software partners being concerned about scaling across multi-core.
As for VLIW/Itanium, I expect some of the same motivations were in play--plus, a desire to reset and lock-out the various x86 clone makers.