I typically run a suite of workloads (see https://sbc-reviews.jeffgeerling.com https://sbc-reviews.jeffgeerling.com), but I like HPL as it's been a consistent r

I typically run a suite of workloads (see https://sbc-reviews.jeffgeerling.com https://sbc-reviews.jeffgeerling.com), but I like HPL as it's been a consistent relative performance metric for decades, especially for efficiency utilizing all available memory. There's always danger to focusing on one metric or benchmark too much, but I also enjoy comparing each system or cluster I build to the historic 'top500' list, to see what decade we're in for small clusters of mini computers.