If by 'a lot' you mean 4 iterations in 8 years, then yes. Williamette inn 2000, Northwood in 2002, Prescott in 2004, and Cedar Mill in 2006. In that time, IPC was pretty flat, and speeds only went from around 2ghz to 3.8ghz (most being 3.4 or below).
I call these the 'lost years' for Intel.
Pentium 4 went from 1.3 GHz to 3.8 GHz. Bus speeds went from 400 MT/s to 1066 MT/s.
Northwood was a die shrink of Willamette with doubled cache. It was pretty much identical otherwise. Prescott is, by comparison, quite distant from Willamette and Northwood. Prescott added support for (some models didn't enable all of these) SSE3, XD-bit, VT-x, EIST, and EM64T. It lengthened the pipeline significantly (which ultimately improved HyperThreading performance - it just wasn't noticeable back in 2004) and increased cache size. Cedar Mill eventually corrected the power consumption issues related to Intel's 90nm process, as well. Cedar Mill, of course, being a direct die shrink of Prescott.
The relationship between Willamette/Northwood and Prescott/Cedar Mill isn't too unlike the relationship between Pentium II and Pentium III. I'm surprised Intel didn't call Prescott "Pentium 5," honestly.
Intel's year-long cadence is what tick-tock brought. They've always used die shrinks of existing cores to test new process nodes. AMD has never had any predictable shrinkage/architecture process going on.