^I'll have t take your word for it. I've only read maybe four or five books and a bunch of articles on SCs so I'm far from being that knowledgeable about the topic. However, I was always given to understand that there were two basic approaches currently being used to implement a supercomputer architecture, one of which is as a cluster. (The other is a grid.) And that the test for thinking of a machine as a supercomputer wasn't so much the capacity
(in flops, RAM, CPUs, etc.) as it was the capability
of the machine. Sort of like saying it's isn't so much what it is
as what it's used
for. So if a computer was capable of doing something either extremely difficult or impossible for an alternative architechture to do, then it was a de facto
supercomputer - even if the absolute 'specs' of the other machine were better.
Much like what separates 'audiophile' from consumer electronics - the audiophile products consistently produce better sound
than the consumer products do - even if the distortion and other figures are technically
better for the consumer product.
But it's been about 5 years for me...so maybe they've since changed some of that viewpoint about SCs?
@f - PS: thx for taking my comment in the way it was intended.