There's an in-depth video that compares the performance of Quake's software renderer on the Pentium MMX versus the AMD K6 at the micro-architecture level: https

There's an in-depth video that compares the performance of Quake's software renderer on the Pentium MMX versus the AMD K6 at the micro-architecture level: https://www.youtube.com/watch?v=DWVhIvZlytc https://www.youtube.com/watch?v=DWVhIvZlytc The most interesting insight in there is how the former manages to outperform the latter in this specific case due to a pipeline quirk in the FDIV instruction, despite the fact it is a previous generation, in-order design vs a next generation, out-of-order design.