They're referring to both - or rather all three. "Stock" just means that the CPU's settings haven't been tampered with, since there's no single stock speed for the 9900K (nor many other CPUs for that matter). The cores of a stock 9900K will run at anything between 3.6 GHz and 5.0 GHz (if we forget about p-states and the like). The 3.6 GHz is the base clock speed (the TDP of 95 W is for this speed, btw) and 5.0 GHz is the maximum boost speed for a single core. The 4.7 GHz you mentioned is the maximum boost speed for all cores. So, if you have all cores maxed out, a stock 9900K will not go higher than 4.7 GHz, but if you have only one core maxed out, then that core can in theory run at 5.0 GHz. Actually reaching these speeds depends on at least CPU temperature and thus the quality of the CPU cooler, so stock speeds can be different for different users. However, tech channels usually isolate temperature as a factor (as they should), so their stock, all core boost speed should be 4.7 GHz. Single core boost speed should then be 5.0 GHz as its easier to achieve than an all core boost speed of 4.7 GHz.

Hmm. My impression is that when max performance is needed, the power saving settings shouldn't be much of a factor. They might affect the single core max boost speed, but they shouldn't affect the all core max boost speed. However, don't take my word for it. Tech channels and tech sites should always use the power savings mode that gives the maximum performance when testing CPUs, but mistakes and oversights can always happen. There are actually a few recent examples of this regarding 9900K benchmarks, namely the use of "too weak" motherboards for testing. Now, technically they were not too weak, since 9900K was listed in the supported CPU lists - so it's not really a mistake, but maybe an oversight. The problem was that the 4-phase VRMs of the motherboards couldn't handle the power draw of the 9900K at full speed, so the motherboard throttled the CPU to run at something like 4.2 GHz instead of 4.7 GHz under full load. This didn't necessarily show in short benchmark runs (because the throttling was not necessarily instantaneous), but did result in temperatures and power consumption figures that were a lot lower than what many others were seeing. You can check Hardware Unboxed's video about it here:

http://www.youtube.com/watch%3Fv=NGHiRrQ2AAo" rel="nofollow - https://www.youtube.com/watch?v=NGHiRrQ2AAo