GOOGLE and read people! It is a 3.6GHZ, 8 core processor, not 1 - 1.5GHZ core. Yes it was a stunt (probably by overclockers as much as AMD) and yes they did use liquid helium, but it still worked.
AMD state that by air/water cooling the chip should be good for 5GHz.
Multicores are one way to speed, but with added software complexity and/or plain algorithmic limitations, GHz is often king.
If you search for the video on YouTube, you can see how they did the test. It was a packaged part on an Asus motherboard -- normal chipset, DDR3, etc. In other words, a working PC. They used the BIOS overclocking menus to set the reference frequency and the CPU multiplier. The 8.429 GHz CPU clock frequency was reported by the BIOS.
The LN2 cooler was basically a 6 inch tall cylinder on top of the CPU, into which a guy slowly poured LN2.
This test was probably done on production test equipment directly on a whole (un-diced) wafer. The tester can be programmed for a 'shmoo' plot to find the best die on the best wafer. It could potentially even be a skew lot where the transistor lengths are varied for fast and slow device corners. Being AMD they own the wafer fab and control the process. This could even be from a next generation process that has not yet been qualified for production runs yet. There are a lot of knobs that can be turned. Companies that own their own wafer fabs can do this sort of thing. You won't see this on most ARM processors that are run by fabless semiconductor companies at foundries.
You have got to be kidding, if AMD even thinks of releasing FX at those speeds, they may as well just shutter the product.
I know they were having some issues (SURPRISE!) with the planned clock frequencies but that is supposed to be remedied in the silicon, and that is why we don't have parts yet.
But don't tell me that they can overclock a 1-1.5GHz cpu to 8+GHz, just not going to happen....PERIOD!