I've got a few problems with this post, entirely because it leads me to believe that you are relatively new to the hardware world.
First of:
Gamers do not need them, most professionals do not need them, and your average consumer certainly doesn't need them.
Yea, I agree. However the top end $1000 CPU's are not targeted at people who need them (because they will have likely gone with a strong Xeon setup) but rather at the people who want them as they are enthusiast parts. I certainly do not need them, but it is highly likely I will get one.
AMD's Opteron offerings in the $400-$500 range will blow away all Haswell-E desktop CPU's.
spits coffee across room
You seriously, like seriously, over estimate the performance of the opteron's. The exact one you are referencing, in terms of multicore performance barely exceeds the processing power AMD's own Fx-8350 and if people thought Bulldozers singlethreaded/IPC performance was poor, this was worse. In my mind it was an outdated chip when it launched especially for its target market.
Also from what benchmarks have already emerged (although synthetic benchs are not the b-end all of analysis) show that even the 5820K beats out the 4960x. With passmark being a pretty good round up overall performance of a CPU, you can see that the 4960x has around 150% the performance of the Opteron. So a chip that performs the better for the same amount of money is a no brainer.
Literally useless to anyone who isn't managing a datacenter. All you get are incremental clock speed increases and lower voltages. Unless you're running an APU/GPU-on-CPU-die, RAM speed (when I say speed I mean overall speed, not the clock rate) has been completely irrelevant for a long time now. There hasn't been a real world improvement from RAM generations since DDR2.
This I kind of agree with. But lower voltages are important, for a system that maybe on 24/7 there is a difference. But as with all memory revisions (of the DDR line anyway) although response of the memory hasn't really changed (taking into the account of higher transfer but lower timings) bandwidth has increased tenfold. For all performance applications, the quicker something can be moved the better, this includes gaming. As always though, you won't really notice an immediate benefit to RAM revisions at the start of its life, as its release specs are usually similar to the high end specs of the previous. Look at how much DDR3 has improved over its life, hell it was rare to see something faster than 1333MHz when it released and now its able to get past 3GHz.
Really nothing special. They're expensive boards with features that you'll never use and will most likely never get market share. (Read: Thunderbolt 2)
Like what exactly? For me I may never use the thunderbolt standard, but the additional USB3 in the spec, increased number of SATA3. (And as I have seen as major response over all here) the additional PCI-e lanes.
Again many of these things don't benefit my gaming rig, but for my productivity rig that has multiple HDD's for scratch purposes, 10G ethernet, Infiniband card, RAID cards (for the more important data) and GPU's for acceleration purposes. The more bandwidth available to these cards, the faster the manufacturers can try and make. If the potential for higher bandwidth is there, yes they can't change their existing stuff, but it will allow them to make even faster add ins.
I just feel that you haven't accounted for everyone within the computing world.