Processor set comparison

The general idea

Microarchitecture refers to the level below the instruction set, i.e., details that can change without having to recompile programs. There are many different microarchitecture variationss – pipeline designs, approaches to achieving parallelism, etc.

The purpose of this project is to look at two designs, which can be two generations of the same instruction set architecture, or two competing designs of similar generations.

The challenge

There are many variations on microarchitectures, so you will need to take care to make usful comparisons (e.g., generation x of processor P was meant to improve on generation x-1, so let’s see if there really was an improvement) . You will need to consider issues like:

intended design goals vs. what was actually achieved
comparison of architecture and microarchitecture features across the two designs, to arrive at a prediction of which would perform better
analysis of published performance data (e.g., from SPEC) to compare against what you would expect from the two designs
analysis of factors outside the microarchitecture which may have influenced measured performance

It would be ideal if you could do actual performance measurement on your comparison designs, but this may not be practical, so make sure before pinning down the comparison models that there are published performance numbers.

Note: this is a different question to #11, instruction set comparison. The focus here is on the detailed design, whereas #11 is about examing code.