I'm unclear on what you mean by "server thread operations".
Generally, Intel has increased the size of the L2 and L3 caches on their CPUs as a way of increasing performance in situations where they are memory bandwidth limited. Those situations typically occur when using CPUs with lower front-side bus speeds, or in cases when using multiple CPUs (because they share the same memory bus). The larger caches are intended to help keep data flowing when memory access is slowed. It generally has nothing to do with process threads.
If I recall correctly, both the Xeon and dual core CPUs support SMT (aka, hyperthreading) which is designed to allow execution of certain non-dependent threads to occur in parallel. This can speed up processing of multithreaded applications. The dual core CPUs basically are two CPU cores attached together, sharng the same FSB (which again leads to memory access issues), but each CPU core is also capable of executing threads in parallel, which speeds up multithreaded applications even more.
So in general when choosing between two Pentium 4-based CPUs clock speed is king. The only exceptions to that would be if you are executing multiple threads simultaneously on a regular basis (heavy multitasking, some image-editing applications, or multipurpose server applications).