Yes, so when the xeon has one core maxed it can turbo to 3.8, 2 cores maxed it lowers to 3.7ghz, 3-4 cores maxed (all 8 threads) it drops to 3.6ghz. The 4690k does the same thing out of the box unless it's oc'd. Mine is currently overclocked and will maintain 4.6ghz across all 4 cores maxed. How far one chip or another will oc depends on temps, the actual chips etc. Some may get 4.7 or 4.8ghz if they're lucky, or only hit 4.4-4.5ghz if they're not as great overclockers. Two for two have easily hit 4.5-4.6ghz.
Both speed and threading have an effect when it comes to dx12 by the looks of the preliminary bench's run with the win10 preview. In a sense, threading more than speed but it depends. Comparing a thread limited dual core g3258 with fast frequency to say an i3 with ht (dual core, 4 thread) the extra threading makes a huge difference. A true quad core like the i5 performs even better and an i7 with 8 thread vs the 4 of the i5 performs better yet. However the performance gain doubling the threads from i5 to i7 are much less impressive than the gains moving from a dual core 2 thread to dual core 4 thread. Speeds continue to play a role throughout.
While thread count trumps clock speed, in the end not by much. There are all varying results from the pentium to the i3 (dual core vs dual core with ht) and quad core vs quad core with ht.
ht enabled i3 with twice the threads gained 22% over the similarly clocked pentium g 3258.
overclocked pentium vs stock pentium gained 26.4% performance (roughly 1.65% per 100mhz of speed going from 3.2 to 4.8ghz)
true quad i5 vs ht enabled i3 gained 54% improvement in score. True cores definitely trump hyperthreading.
quad i5 without ht at 3.6ghz vs the ht enabled i7, hyperthreading gained 22% performance.
Overclocking the i5 to 4.6ghz using the same improvement per 100mhz as seen on the pentium (by the figures provided in the test), the oc'd i5 gains 16.5% performance. In the end, an overclocked i5 with only 4 threads closes the gap to a ht enabled chip at the same speed (such as a locked core xeon) to a difference of just 4.9%, the tiny edge given to the ht. Even in heavy threaded situations, clock speed often narrows the gap of ht to the point of being a wash. In other less threaded situations as in current games that aren't using dx12 yet, clock speed will pay out even more. Obviously the i7 4790k would have the edge having both ht and higher clock speeds. One isn't better than the other, they both matter.
http://www.pcworld.com/article/2900814/tested-directx-12s-potential-performance-leap-is-insane.html
The nice thing is what we've seen from dx12 is ht finally reaching the 20% improvement or so that it 'can' be capable of since it often doesn't shine nearly as well in most other real world scenarios. These are also just fictional speculative results based on pre-release versions of dx12 and simulated tests. The real gains in actual dx12 coded games with retail win10/dx12 may be different in actual gaming fps/performance increases.