Re: It's interesting...
For decades, when it comes to CPU design technology, the prominent driver behind all scientific research has been performance
That's been the dominant driver, but it didn't prevent manufacturers from bringing CPUs designed for other goals to market in the '80s. For example, Intel had the i432, a capability architecture. IBM had the capability-like1 AS/400 CPUs, first the CISC IMPI CPU and then PowerAS which was a tweaked POWER design. The i432's primary design driver was the enhanced security of a capability architecture. The AS/400's design was guided by the "five principles" of "technology independence, object-based design, hardware integration, software integration, [and] single-level store".
Support for legacy software was the key economic driver for the AS/400, and that also gave us CPU families like the two Unisys ClearPath lines and IBM's 360-370-390-z line.
So it hasn't always been primarily about performance.
For general-purpose computing, though, it's hard to see how the economics could ever have favored anything other than a few performance metrics - operation throughput, price/performance, and power/performance.
1As Frank Solis describes in his book about the AS/400, the '400 architecture team discarded the S/38's true capability architecture (pointers carry access-right information) for the '400, largely to accommodate issues with transient changes in access rights. However, the '400 (and now System i) still uses pointers which refer to specific objects and cannot be altered by user-mode code.