According to various leaks ( http://wccftech.com/intel-broadwell-e-core-i7-6950x-price/ ) and ( http://wccftech.com/intel-broadwell-e-hedt-computex-2016/ ) the CPU frequencies of the upcoming Broadwell-E i7-6000 processors are as follows:
CPU Cores Frequency Turbo-boost L3 Cache
i7-6950X 10 3.0 GHz 3.5 GHz 25 MB
i7-6900K 8 3.2 GHz 3.7 GHz 20 MB
i7-6850K 6 3.6 GHz 3.8 GHz 15 MB
i7-6800K 6 3.4 GHz 3.6 GHz 15 MB
Observe how, for the top 3 processors, as the core-count increases, the CPU frequency decreases. While the L3 Cache increases – as it's shared between all cores it works out at 2.5MB/core.
Most of my computing needs are not many-core optimized, such as games – I don't do things like video-encoding, rendering, or bitcoin-mining – which means whatever is in cache would be suited to however many cores are being utilized, so the larger L3 cache would definitely help (and so be effectively larger than 2.5MB/core).
So my question is: if I got the i7-6850K, would the extra 600Mhz (300Mhz in turbo-boost) in single-threaded performance offset the smaller L3 cache compared to the i7-6950X? But contrarywise: would the 500Mhz turbo-boost in the 6950X be in-effect if I was running low-threaded programs and augment the larger L3 cache?
To be sure for certain, I'd have to wait for the Broadwell-E benchmarks to come out – though ideally I'd like to get my pre-order in first.
Best Answer
Stated turbo-boost frequency is the maximum that the cpu permitted to run, but it is not guaranteed that it will ever do.
The whole story with boost is because of thermal and power design. If all cores were running at that speed, the cpu would consume too much current and also instantly overheat. So they came up with an idea to allow couple of cores run faster, while putting the others in low power mode. This for single-threaded tasks it actually may reach its top boost freq (provided there's cooling headroom available)
Cache is useful if there are cache hits, how a program benefits from it completely depends on the program itself and it's memory access patterns, only benchmarks of particular one can help here
On the bottom line - all depends on your actual usage, both CPUs have a chance to be more suitable