It depends on how many threads a "multithreaded" application spawns. If it spawns two compute-intensive threads, any chip with 3 or more cores will have similar performance. The two program threads will run at 100% on two cores and then the OS's background tasks will run on the other core or cores. Most of the programs tested in consumer-type sites are games, which tend to have one or two heavy threads and then little else to run on the CPU.
Even programs that have many compute-intensive threads may not run much faster on a quad than they do on a triple-core chip. It all depends on the memory bandwidth available and needed as well as cache usage and other system resource contention issues. A good case to demonstrate this is running SPECfp_rate on a dual Clovertown Xeon system. Performance scales well up to four threads if the OS schedules the four threads to run as two on one CPU and two on the other CPU, so the FSB bandwidth is utilized optimally. But once you cross the 4-thread barrier, the FSB starts to get hammered something fierce and you get much smaller increases in performance when scheduling more threads, even when there are idle cores available.