http://www.anandtech.com/bench/Product/836?vs=287
Haswell is something like 10-40% faster per clock (average around 20%) so in order to match your 5.2ghz Sandy, you need a 3.7 to 4.7ghz (avg 4.3ghz) Haswell.
Most people seem to run their chips at around 4.3ghz, so your Sandy is a pretty good match for Haswell running at stock voltage. Some of the better overclocks I've seen without extreme cooling (~4.7 is unusual but not unheard of) will in some tests beat a 6.6ghz 2700K (40%).
To be fair, Haswell is 1150 and not 1155 though. You would need something like a 4.7-4.8ghz 3770K to match your chip, which is the upper range for those chips.
That hasn't been my experience at all.
I've built many many hundreds of SB/IB/HW systems, and there's nothing that indicates 10-40% IPC improvement over SB on average, when using the same RAM and drivers. That's the problem with charts like Anand's and Tom's that accumulate data over a long period of time without filling in the blanks on all the testbed data.
This is much more accurate in my experience :
http://www.pcper.com/reviews/Proces...erformance-and-Architecture/Clock-Clock-Sandy
"Finally, our x264 benchmark results show a similar 4.5% advantage for Haswell over IVB and a 10.5% edge over Sandy Bridge on the first pass. Haswell is 11% and 17% faster respectively on the second pass."
What happens is that a certain spectrum of applications (and most synthetics, but you can't run them for any useful purpose) really do take off massively on Haswell (Truecrypt for example), and that skews things a lot. HUGELY helpful if you do things on a regular basis which can take advantage, but otherwise it's a huge 'meh'.
Example, if you're a gamer, a 5Ghz SB i7 will be as fast or faster than a 4.4Ghz HW or 4.6Ghz Ivy, easily. Along with the vast majority of day to day tasks, web browsers, even h264 encoding.
A more accurate picture can be found from the AT official Haswell review :
http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/9
" The performance gains over Ivy Bridge depend on workload, but in general youre looking at
low single digits to just under 20%." <--- that 'just under 20%' doesn't happen with the majority of situations, lol.
Have you noticed that almost ALL Haswell reviews skip real gaming benchmarks almost entirely? This was Intel's direction, they wanted attention to the real gains that Haswell provides in apps that really take advantage, along with the improved iGPU and mobile capabilities.
It's super hard to even find real gaming benches with Haswell, but :
http://www.guru3d.com/articles_pages/core_i7_4770k_review,23.html
http://www.hardocp.com/article/2013/06/01/intel_haswell_i74770k_ipc_overclocking_review/5
Those are stock speed benches.
As you can clearly see, with those % in play, a SB i7 @ 5Ghz is very plainly as fast or faster than HW at even 4.4.
I think it's really a draw for overclockers/gamers.
For non-overclockers, HW is a really nice move!