Most of us with Haswell-E CPUs just forget about running 3400 MHz+ memory only the most lucky ones.
You don't need 3400 Mhz+ memory on quad memory channel X99 even 3000Mhz is almost entirely pointless! You don't even benefit to a worth while degree from memory speeds above 2666Mhz on dual memory channel Z170!
See
here and
here and
here for example....
Hell even using only 'average' speed memory on X99 in dual channel opposed to quad channel makes no appreciable differences unless your running a memory bandwidth benchmark!
See
here!
Neither Z170 or X99 are short of memory bandwidth or speed using rather 'averagely' clocked DDR4 memory...
Neither benefit to a significant degree from running memory much above 2666Mhz (unless your using the integrated iGpu on Skylake in which case your not using the CPU right anyway!)
Anyhow although Broadwell-E ups the 'stock' max memory rate up to 2400Mhz from the 2133Mhz of Haswell-E I would not expect much improvement in the max overclocked memory speed as the slower average overclocked memory speed using Haswell-E/X99 compared to say Skylake/Z170 is a function of the more complicated quad channel memory interface (on X99) over the simpler dual channel interface (on Z170). Broadwell-E is likely to allow for similar average overclocked memory speeds to Haswel-El
Haswell-E is old and the IMC is relatively weak compared to the newborn Skylake. If Broadwell-E has a much stronger IMC than Haswell-E and overclocks better on average, then why not?
Broadwell-E is a die shrunk Haswell-E chip with a few 'tweaks'. 22nm - 14nm as per the now
old school Intel Tick/Tock
Instructions per clock improvement will be minimal... within 5% max and more likely in the 2-3% range.
From what has leaked about the 10 core sku we know that Broadwell-E (
25MB cache total on the ten core 6950X) will still have 2.5mb per core cache so the same as Haswell-E.
Its also highly probable that overclocking wise that Broadwell-E will be similar to Haswell-E given what performance/frequency gains were possible from consumer 22nm to 14nm overclocked CPU's and other recent CPU releases
Broadwell-E will offer a slightly cooler (overall) and less power hungry CPU for the same core count as Haswell-E with a ten core top end SKU (6950X) instead of a octo core top end sku (5960X).
Given that the 10 core will likely cost 1200+ and that it will overclock in a similar if not worse fashion to the 5960X it will be almost entirely pointless for anyone except those who probably should be going the whole multi threaded hog and using a Xeon anyhow.
Going from 'comparable'
Haswell-E to Broadwell-E CPU's (i.e. 5820K - 6800K) is likely to rank as one of the most underwhelming 'upgrades' in Intel's history! Of course one of the benefits of the X99 platform over say the Z170 platform I that you do at least have the ability to drop it a CPU with more cores down the line where as with Z170 you can look forward to an equally underwhelming upgrade but be stuck with a 4c8t CPU going from a 6700K to the 'Kabylake' 'refresh' (7700K maybe?) as 6+ cores wont be coming to the 'consumer' lineup till Cannonlake hits (maybe in 2017?) with a whole new motherboard chipset and socket in tow and lets not forget that Intel are releasing another
new chipset to get the most out of Kabylake anyhow so Z170 boards wont be top dog when Kabylake is out anyhow
with Skylake-E in Q1 next year
Very unlikely that you'll see Skylake-E in Q1 2017 given how Intel's release dates for CPU's have repeatedly been put back for recent releases...... as noted above Broadwell-E itself was penciled in for Q1 2016 and were already well into Q2 with no release date in sight for Broadwell-E
