There are a couple ground-level concepts that are useful to remember when it comes to multi-core/multi-thread processors.
First, as many have covered already, not all tasks can be run in parallel. If step 8 relies on the output of step 7, which in turn relies on step 6, throwing more cores at the task won't do anything to help it run faster. You can't run step 6 on one core, step 7 on another, and step 8 on a third (well, you probably could, but you'd still be waiting for step 6 before moving on to step 7). Writing truly parallel code can be rather tricky.
Second, most high-end CPUs now support Symmetric Multi Threading, which takes advantage of CPU architecture wizardry to run two threads on one physical core (Anandtech has a great primer on how this works). Because of SMT, Windows sees a 4-core Intel i7 as actually having 8 cores. I don't have one in front of me (currently running a "lowly" i5 without SMT), but if memory serves these "extra cores" usually show up as the odd-numbered cores in Windows. Again, not all tasks are capable of taking advantage of SMT, so it is not surprising to often see all these cores not being used very heavily. I suspect this is what's happening in Gabriel Marcus' case.
