Dual Core Or Dual CPU - What's the real difference in performance?

[Mike Meyer]

In <200702081518.l18FIeMR002991 at cwsys.cwsent.com>, Cy Schubert <Cy.Schubert at komquats.com> typed:
> In message <17866.47828.219523.71972 at bhuda.mired.org>, Mike Meyer writes:
> > Generally, more processors means things will go faster until you run
> > out of threads. However, if there's some shared resource that is the
> > bottleneck for your load, and the resource doesn't support
> > simultaneous access by all the cores, more cores can slow things
> > down.
> > 
> > Of course, it's not really that simple. Some shared resources can be
> > managed so as to make things improve under most loads, even if they
> > don't support simultaneous access.
> 
> Generally speaking the performance increase is not linear. At some point 
> there is no benefit to adding more processors.

When some other resources becomes the bottleneck. Which resource
depends on the workload. In some cases, adding processors will slow
things down.

> To add another dimension to this discussion, hyperthreading uses spare 
> cycles in a single processor to pretend there are two processors, 
> increasing performance for some apps and reducing performance for other 
> apps.

I think hyperthreading gets a bad rap. It shares lots of resources -
like the computing units - so there are lots of workloads that cause
things to get worse when you add a processor. But the general case
should still be that it gets faster.

> Generally speaking, dual core is an inexpensive way to get SMP into the 
> hands of people who could not normally afford SMP technology as it was.

Gee, I thought it was a reaction to losing the clock rate war.

	<mike
-- 
Mike Meyer <mwm at mired.org>		http://www.mired.org/consulting.html
Independent Network/Unix/Perforce consultant, email for more information.