How many cores are needed to demarcate “High Performance” these days?

I can remember when 32 cores meant you had a great deal of processing power. So what constitutes “High Performance” in your mind these days?

—

I would like to share an astute observation (first heard them from my good friends Steve Gombosi and Jeff Kuehn) – If it is to be classified as a “supercomputer” (or HPC) system, then it must be able to turn the older compute bound problem into an I/O bound problem!
So the number of cores do not matter, it is how you harness them together to achieve a quanta leap forward that makes it “HPC”

—

That’s a good one, tell Steve and Jeff I will have to “borrow” that quote. For some of what I am doing with retrograde analysis of endgames (finding very long checkmates in the game of chess and long wins in the game of checkers) I can become I/O bound very quickly with as few as 8 cores fully engaged. I am looking into making the process more parallelizeable, but I hit the Amadahl’s Law limit rather quickly because the I/O bound portion is inherently serial. Now if hard drives were only as fast as RAM, there would be no problem