Again, the bottleneck is not likely the CPU in this scenario. Memory bandwidth and latency are key factors in seeing the CPU fully utilized. What's a the typical memory latency these days on x86? 12 cycles? Games must make tradeoffs in how many tasks can be run in parallel vs how efficiently they access ram pages. Cache misses kill performance on pipelined CPUs. Work is organized to minimize cache misses. This requires that some tasks are run in serial, essentially creating an upper limit on how parallel the games code can be run.