The concept of cache misses and the importance of cache locality is nothing new. That's HPC 101.
Demonstrating that in the real world a specific data structure performs far better than the theoretically optimal data structure is not. That's novel, unless you can find any prior work that presents the exact same findings.
That's one of the main problems with ivory tower types: claiming that a novel poem is nothing new because all the words in it were already known and even published in a few dictionaries.
All the world was not x86. The DEC Alpha definitely had 3-level cache systems by the mid 90s, and I would not be surprised if other RISC systems did as well.
