TOP500 hasn't been a particularly useful measure of practical computing power in modern systems for many years because what it measures isn't a significant bottleneck in most real systems. It has become a measure of how much money someone is willing to spend for bragging rights. (HPCG is better in that it is a bit more bandwidth focused but still pretty narrow.)

Most companies with huge systems don't participate.

I wonder if there would have been an opportunity to generate some finer-grained benchmarks with something like BiCGStab+ILU (or maybe CG+incomplete cholesky). Instead of CG+Gauss Seidel. The pitch being, you might have made different memory vs compute trade-offs with designing your cluster, but you should be able to select a fill-in factor for the preconditioner to suit it.