[Pc_Support] Are "Core Duo" and "Dual Core" synonymous terms? <EOM>

[Bryan J. Smith]

Damien McKenna wrote:
> "Dual Core" means there are two CPU cores to a physical CPU chip.
> Core Duo is a marketing term from Intel, as is Core Solo, Core 2 Duo
> and the upcoming Core 2 Quadro.
> Every Core Duo or Core 2 Duo CPU is dual core.
> Not every dual core CPU is an Intel Core Duo, e.g. AMD have had dual-
> core CPUs on market for ages,

Understand that "Core" is an Intel _brand_name_.
It represents Intel's new "Core" x86-64 design.
Again, it's a _brand_name_ just like "Pentium" or "Centrino."

There are "Core Solo" and "Core Duo" to reflect one or two cores.

> and Intel also have been selling Pentium4-based and older
> Xeon-based dual core chips that are nowhere are good as the
> newer generation.

Those are *NOT* "dual-core."  That is _dual_threading_.
Do *NOT* confuse "core" with "threading."
Because in the near-future, we will have multi-threading over multi-core.

Intel "HyperThreading" processors are a _single_core_ (and quite
_inefficient_ at that) with a scheduler hack so they look like
multiple CPUs from the OS' standpoint.  Make no mistake, regardless of
how they "fool" the OS, there i _only_ *1* set of cache, controller,
arithmetic logic and floating-point units.  That makes them 100%
_single_core_, and no other interpretation is even remotely valid.

With true "dual-core," you get 2 sets of cache, controller, arithmetic
logic and floating-point units.  That means you get _at_least_ single
core performance, and typically 60-80% more.

With "dual-threading" on a "single-core," you get 2 virtual threads
_sharing_ 1 set of cache, controller, arithmetic logic and
floating-point units.  The result is that for well-threaded
applications, up to a 20% bonus is possible.  But for poorly-threaded
applications, as poor as -10% (a performance _hit_ over 1 thread) is
possible.  This is due to the _signficiant_ overhead required for the
threading.

BTW, Intel does _not_ offer "HyperThreading" on Core processors.  Why?
 Because the Core processor is 50-100% more efficient than the
NetBurst/P4 architecture -- keeps its pipes 60-80% full, instead of
only 40-50%.  "HyperThreading" was a hack to take better advantage of
the fact that there are many stages being _unused_ in the NetBurst/P4
architecture.  That is _no_longer_ the case in Core.

The future of multi-threading is to address multi-instance
virtualization over multi-core.  Such multi-threading is not really a
desktop feature, but far more of a server one -- mainly when running
multiple OSes over multiple, virtual machines/systems/instances.