Reddawne
lehpron
It isn't a mystery as long as you remember that Haswell-E uses the same architecture as Haswell, therefore it will have the same IPC per core against Ivy-E. The 8-core however has a 3GHz stock speed, it will be the hardest to overclock; the sell point is having 8-cores, I think someone would be a moron to judge an inability to reach the overclocks of quad-cores.
Same goes for the 6-core?
My first sentence will apply to the 6-core models, since Ivy-E had 6-core models, i.e. the difference between the 4930K and 5930K is 3% frequency and 10-15% IPC.
Of course these estimates are based on using the same frequency of system RAM, I'm not sure how many real-world programs are affected by frequency; most of my uses from both my X58 triple-channel and HM55 dual-channel didn't perform differently from a RAM sense. The stock default of X79 w/Ivy-E was DDR3-1866 while the stock default of X99 w/Haswell-E looks to be DDR4-2133, so a 14% bump in bandwidth.
DDR spec is mainly a change in voltage and current (thus power), it doesn't affect actual operating frequency, but rather the range of potential SKUs in terms of what frequencies can be sold stably. In other words, two quad-channel kits at DDR3-2133 and DDR4-2133 will perform alike if tested independently. Unfortunately, they can't be tested independently accurately; one would have to disable cores on the Haswell-E 6-core and using only two DDR4 DIMMs to simulate against Z97/Z87 with a similar spec'd DDR3 kit and a i7-4770K and match up the frequencies. That would be the only way to see what difference DDR4 makes on a per clock, per core, per DIMM module sense. Power and heat levels will be the biggest difference.
post edited by lehpron - 2014/07/23 10:55:16