Gaming CPUs are generally at the top of a company's mid-range lineup.
Compare the Intel i7-4770k and the AMD FX-9590. Amazon actually sells the AMD chip for $2 more ($322) than the Intel chip ($320), but you'll get more performance out of the Intel chip than the AMD chip (though not a gigantic difference).
The FX-9590 has eight cores, and the i7-4770k has four. Even with half the cores, the Intel gets more performance. That's quite impressive if you ask me. The reason for that is because of single core performance.
Single core performance is more important than the number of cores (assuming you have enough cores to meet modern software demands). Each core on a chip is actually an individual CPU. In the old days, multi-CPU performance was possible only with special motherboards that had sockets for more than one processing chip, and I remember that back then AMD generally made superior chips (but that's a side point). When multi-core CPUs came out, they put multiple CPUs on the same chip. That made multi-CPU computer the norm whereas it used to be a very exotic configuration for systems where money was not a limiting factor.
But even back then, multi-CPU configurations were faster only if the chips scaled to be faster than a single chip. That required software to recognize more than one CPU and for each CPU to be sufficiently fast. You see that same software limitation today. Most programs simply can't recognize very many CPUs. Windows displays each core (or HyperThred in the case of an i7 or Xeon) as an individual CPU, just like you would see in the old days by plugging in an extra chip.
To illustrate how major the software limitation is, imaging you have one computer with a single Pentium 4, which was a single-core CPU, and you have a second computer with four separate 486 chips on the same motherboard. Theoretically, if the 486 were only 1/3 the speed of a Pentium 4, then your could scale four 486 CPUs to get a 33% speed increase (because 4 x 1/3 = 4/3). When all four of the 486 chips run at max, that may indeed be the case. But back then, almost everything was programmed to work on single-core chips because almost all computers had only one CPU. The problem was that you'd spend a lot of money on a fancy system with four CPUs, and then your software would use only one of the CPUs. Womp! When that happened (which was most of the time), then your computer would be only 1/3 the speed of a Pentium 4 system even though a Pentium 4 system would have cost much less money.
AMD today equips their chips with more CPUs that are slower than Intel's. That may actually be a very good thing when you use programs capable of recognizing many CPUs, but games can't currently do that (with a few exceptions like Crysis 3). So then you're stuck in the same situation as in the old days, where you'll have CPUs contributing nothing to a task, though they can still do other tasks (e.g. play your music in the background, record your game, whatever). I think it would be cool if we could actually have a single CPU fast enough to handle all modern tasks, but it turns out that would require too much power.
As for Mhz, that just tells you how fast the CPU is "spinning." I you have a Toyota Corolla at max speed and a Ferrari at max speed (say both are at 7,000 rpm), will both cars go the same speed? No way! Everything about the Corolla is slower even if you get the engine to run at the same rpm as the Ferrari. It's the same with CPUs (though not as cool).
While AMD doesn't put eight "Corollas" into its chips, it certainly doesn't put in eight Ferraris either. If Intel does that, then you might say that AMD probably uses something more like a BMW M5 (which is still a very fast car, just not a Ferrari).