After so many pages of lunatic hype, except for a few cogent contributions, let's return to biologymajor's question that started this thread. Having used multiprocessor computers for roughly 10 years - Xeon's and PIII - I still kept one dual processor PIII going for backup and experimentation. I'll only address workstations in what follows. I can tell you that I never ran into a magic supercalifragilistic case where a threaded application load started at once to load up two processors. NOPE. I could only accomplish this by using C++ to schedule individual threads to specific processors. This not being the most practical thing dragging along in programming (skipping thread sync and so on), My first answer to biologymajor would be that it CAN be done, the developer willing to go through the required motions. As to the original poster's "Why?" question. In my PIII, Xeon experience, it wasn't worth the effort because in those days chip manufacturers managed to ramp up speed so quickly that reverting to a one-processor was much more attractive. Now that manufacturers are faced with the inability to increase speed, they came up with the duo chips and promise quad and even octal chips for tomorrow. However, nothing really has changed in hard or software technologies leading to "dumb supercalifragilistic" load balancing between threads of current software lacking processor/thread scheduling. Some applications may have indeed been updated to exploit this. One poster mentioned X11 in linux. I have not verified this myself, so I dunno. This problem is one of current research in software technology and afaik nobody has come up yet with a suitable paradigm. Of course, as some poster pointed out, the last chips may give you a bit more memory bandwidth which in given applications may lead to more application-processing speed but this may be very marginal in many applications as compared to late model single chips. We will all be obliged some day to buy multiprocessor chips. Let's not forget that an era has ended and that the above points remain to be addressed. nono