But isn't a "core" a "processor" and/or a "CPU"?
Evangelion said:
Whereas in the past you needed two physical CPU's to have two processing-cores, but these days you get it with just one CPU.
With single core chips, the terms "core", "CPU" and "processor" were pretty clear, and meant the same thing.
The core/CPU/processor had architectural state (program counter, integer and FP registers, status registers), L1/L2 cache, memory and bus interfaces - the whole works.
The "Pentium D" dual core chips are still two independent chips - each with state, cache and interfaces - just two chips glued to the same substrate and connected to the same socket. The FSB extends into the package, and the two chips communicate over the FSB just like a two-socket system.
Yet, some bean counter decided that this was 2 cores, 1 processor for the purposes of licensing. Suddenly, for no technical reason, at the licensing level "core" and "processor" no longer meant the same thing. (AFAIK, nobody really defined whether "CPU" meant "core" or "processor" in this new-speak, however.)
Now, we have two cores on a single chip, and depending on the design they might share L2/L3 cache and the bus interface. (And SMT (Hyper-Threading) chips also have unique state per logical processor...)
_________________________________________________
If we only use the terms "core" and "socket", there's little ambiguity - both the bean counters and the engineers agree on what those mean.
"CPU" and "processor" are no longer useful terms - their meaning depends on context and the application that you intend to run. If you look at a Windows dual-dual system, it will say that it has "two physical processors and four logical processors".
Similarly, the people who are misusing the term "SMP" to mean "multi-socket" are applying a term from a former reality to the current situation. The "P" in "SMP" can really only be interpreted to mean "core" in classical terminology.
