The problem with benchmarks like Winstone and Sysmark is that they grossly skew the importance of singular speed over cpu-availability in real usage patterns.
Few apps besides mpeg encoding personally take advantage of SMP, but there's one very, very important "app" that puts it to VERY good use -- Windows XP. Even if the individual apps themselves don't take advantage of SMP, XP can assign one thread to each CPU.
There IS one sizeable group that invariably gets overlooked and ignored by articles judging the worth of SMP, even though they're probably the gold standard of a group that can benefit from it: programmers.
Take, for instance, a developer who builds Java web applications. At any given moment in time, he's probably got all of the following running:
* Norton Antivirus
* Tomcat (JSP/Servlet web server)
* Forte, Netbeans, or Eclipse (3 popular IDEs)
* 2-10 instances of Internet Explorer or Mozilla
* a text editor for taking quick notes and use as a text scratchpad
as well as one or more of the following:
* Visio, Rational Rose, or some other UML app
* Outlook, Outlook Express, or some other mail app
* Secure CRT, or some other terminal app
* one or more command windows, or maybe even a Cygwin shell
None of those programs individually make any effort to put two CPUs to good use... but give Windows XP two CPUs to play with, and it will keep both of them spectacularly busy and give the kind of night-and-day palpable performance boost that will spoil you forever and make even a 2.4GHz single CPU system feel sluggish compared to a dualie 1800 system.
Of course, a similar argument could be made for giving such a system to a serious web developer, who's probably running:
* Norton Antivirus
* Photoshop
* Dreamweaver
* Flash
* 2-10 instances of Internet Explorer or Mozilla
* one or more components of Office XP (Word, Outlook, etc.)
In this case, it's not QUITE a night-and-day difference, but it'll still be easily perceptible as long as the system has enough ram to eliminate THAT as a choke point (say, 768 megs to 1 gig).
On the other hand, even a lowly office worker running productivity apps would probably be happier with a dualie 1400-1800 system in lieu of a 50% faster single cpu system. Winstone DOESN'T tell the whole story. Winstone measures how quickly it can perform a scripted OLE runthrough of common office apps, RUN ONE AT A TIME. How many people actually exit Word before loading IE or Outlook? I'd venture a fair guess that the following apps running simultaneously represent a FAR more realistic use case:
* Norton Antivirus
* one or more instant messaging apps
* Word, Excel, and/or Powerpoint
* Outlook or Outlook Express
* a half-dozen instances of Internet Explorer
In the office "productivity" user's case, the holdup isn't likely to be lack of raw speed... it's momentary hangs and glitches caused by one or more apps needlessly hogging the CPU or making blocking system calls -- many of which are timeout-based and will happen REGARDLESS of how fast the CPU is. Replace a single blazingly fast CPU with two CPUs that are 20% slower and together cost as much as a single faster one, and his computer experience will almost CERTAINLY improve. He'll become easier to support, too... because he'll be less likely to start filling the event queue with multiple mouse clicks whenever Windows hiccups and becomes unresponsive for a second or two -- something that happens FAR more often than anyone really cares to admit.
IMHO, a FAR better test (assuming it would actually run) to show the advantage of dual CPUs would be to run high-end Winstone and Sysmark SIMULTANEOUSLY. Assuming it worked, the dualie system would purr along. The single CPU system would utterly choke. Suddenly, the dualie's razor-thin score advantage would become INSTANTLY obvious.
Far from being a contrived, artificial benchmark, running both at once would go a LONG way towards replicating the usage patterns of REAL users who DO run multiple apps at the same time.