Dual Core vs 2-Way Server

Help settle a bet... assuming an application is designed to take advantage
of a two processors which is faster a two-way box or a box with a single
dual-code chip?

"swimdad16" wrote in message
.. .
Help settle a bet... assuming an application is designed to take advantage
of a two processors which is faster a two-way box or a box with a single
dual-code chip?

Since some portions of a dual core CPU are shared, they can't be
truly asynchronous. One will sometimes have to wait for the other.
I'd say the 2-way might be a bit faster in those cases.

Tom Lake

From my experience I would have to agree but it also depends on how the
mother board resolves CACHE conflicts and shares system memory between
the processors. The dual core CPUs for desktop systems seem not to have
enough CACHE and bus bandwidth but the next generation chips show lots
of promise (but I'm not counting my chickens until they are hatched).

You can design a product trying to take all the right things into
account but when placed into the real world application it doesn't
live up to expectations. Dual core CPUs are still in there infancy and
I'd wait until about 6 months after the release of the third or at
lease the second generation chips. This gives Intel and motherboard
manufactures time to resolve issues and you're not their QA
department at your expense. But the good news is Intel has been
listening, especially when criticisms affect sales. I'm patiently
waiting for the release of CONROE's reviews and recommendations for
systems.

swimdad16 wrote:
Help settle a bet... assuming an application is designed to take advantage
of a two processors which is faster a two-way box or a box with a single
dual-code chip?

Depends entirely on the efficiency of interprocessor communications
paths. In an Intel system, all communications goes over the front-side
bus (FSB), not just interprocessor communications, but i/o bus traffic,
and memory bus traffic. So an Intel system tends to get more and more
congested when more processors are put onto the FSB. So in a case like
this one would think that a dual-core system would be just the better
solution to dual-processor, because the two cores would talk to each
other over a private internal communications channel, leaving the
outside FSB channels free. Well, that wasn't really the case with
Intel's first generation of dual-core chips. The Pentium D/Pentium EE
and Xeon dual-cores were actually not talking to each other through a
private internal communications channel, they were just going out over
the FSB and back into the other core. It resulted in more traffic, not
less over the FSB. The new generation of Core Duo processors have a
proper internal communications system that doesn't go over the FSB.

In AMD's case, their dual-cores were always properly designed with
internal communications channels between the cores. Also AMD had
designed a more efficient external communications mechanism than the FSB
that Intel has. They call it Hypertransport, or Direct-Connect
Architeccture, and it was a dedicated point-to-point communications,
they separated out the i/o traffic from the interprocessor
communications and the memory traffic. So even without dual-cores, the
AMD solution was pretty efficient. Dual-cores just gives it the icing on
the cake.

Tests have shown that dual-core vs. dual-processor is almost identical
on AMD systems, not much difference; universally good. Tests have also
shown no difference on Intel's old Pentium 4-based dual-cores, due to
the fact that even the dual-cores went out over the FSB; universally
bad. However, Intel's newer Core-based dual-cores with their dedicated
internal buses between cores has shown marked improvements over their
own dual-processors.

Yousuf Khan