Some testing on the i7 3960X vs 2600K & 990X

[taekvideo]

Dude, you realize that even though it presents a different socket, it's still a Sandy Bridge derived CPU?
I.E. performance on limited thread counts should be equal to each other (the miniscule score differences in the first tests is not a win), it's not an automatic win.
Also some scores are influenced by the complexity of the chip, some calculational paths taken may represent a "longer distance" then the path taken in the LGA1155 variant, thus 1155 edging it out by a miniscule and not noticable margin.

Take note of 3D designers, CAD/CAM, Video editors, anything that is heavy enterprise stuff, that is coded to be very favorable to high thread counts.
In note of a 3D generation it can take 1 month of number crunching by the CPU to generate 5 minutes of film when using the i7 2600K, using an i7 3960K this time is cut down to 20 days for 5 minutes of generated film.
This is not an exaggeration actually, these are the performance gains of these things as they scale very linearly, provided it's a professional software package.

Quick Math:
30 days = 720 Hours
20 days = 480 Hours
Say that person who has to keep an eye on it costs 20 euros an hour and it also costs 1 euro an hour that the machine is doing it's number crunching, counting 8 hour workdays.

30 Day Generation, very simplified calculation
8 * 20 = 160 Euros (single day)
160 * 30 = 4800 Euros (30 days)
24 * 1 = 24 (1 euro an hour electricity burn)
24 * 30 = 720 Euro (30 day burning of electricity)
4800 + 720 = 5520 Euros cost for a 30 day generation of a project that in that world usually has a fixed price.

20 Day Generation, very simplified.
8 * 20 = 160 Euros (single day)
160 * 20 = 3200 Euros (20 days)
24 * 1 = 24 (1 euro an hour electricity burn)
24 * 20 = 480 Euro (20 day burning of electricity)
3200 + 480 = 3680 Euros cost for a 20 day generation of a project that in that world usually has a fixed price.

Net Gain in costs: 1840 Euros.

This was just gained with free spare time to engage another project, so not only do you gain cost reduction due to reduced generation time, you also gain profits from the next project you receive in the same time it took you with 30 days, so you have double earnings here.

Now please tell me, how do you not see a market for this? It's irrelevant that the i7 2600k MAY or MAY NOT be faster in limited thread count programmes.
Server CPUs aren't used for this kind of thing unless you have a ton of servers behind it, hence the server CPUs generally being far lower in clock speed and TDP, if you want to though i'll find you a full speed Xeon and show you the price tag on that and you'll notice the difference between prices.

There's an AMD 8-core server chip that only costs $250. You can use 2 or 4 of those in a single server. Would be much more efficient than using "Extreme" chips to do all that rendering.

[Evildeffy]

There's an AMD 8-core server chip that only costs $250. You can use 2 or 4 of those in a single server. Would be much more efficient than using "Extreme" chips to do all that rendering.

8 Core server chip running @ 2 - 2.2GHz (norm) vs. 6 Cores and 6 HyperThreads running @ 3,3GHz (not even counting turbo mode).

8 * 2.2 = 17,6GHz of computational power
6 * 3.3 = 19,8GHz of computational power (not even counting HyperThreads)

Which do you think is faster? Efficiency and Speed are 2 entirely different things.

[taekvideo]

8 Core server chip running @ 2 - 2.2GHz (norm) vs. 6 Cores and 6 HyperThreads running @ 3,3GHz (not even counting turbo mode).

8 * 2.2 = 17,6GHz of computational power
6 * 3.3 = 19,8GHz of computational power (not even counting HyperThreads)

Which do you think is faster? Efficiency and Speed are 2 entirely different things.

I think you missed the part where I said you could use 2 or 4 of those cpu's in each computer...
And your "ghz math" is so... wrong.

[Evildeffy]

I think you missed the part where I said you could use 2 or 4 of those cpu's in each computer...
And your "ghz math" is so... wrong.

I did indeed miss as such, whoops!
True you could stab multiple in, but you'd have to pay more on RAM as you'd require ECC for the mobos and more because every slot per CPU has to have some population of memory.
You'd require an odd case aswell due to the massive sizes of multi-CPU boards.
Not to mention architectural differences.
But you are correct, speed would alter, but still not enough i'd think, in any case Intel rigs are more expensive yes, however with the setup you're suggesting, i think you'd be spending the same.

However on my GHz math, this was assumed on an equal level speed and strength of speed (IPC) with linear scaling, it's in basis correct (very basic, but serves the point).
If we were to park in architectural differences and IPC and whatnot, that Intel machine would yet again shoot ahead.

[Cyanotical]

I think you missed the part where I said you could use 2 or 4 of those cpu's in each computer...
And your "ghz math" is so... wrong.

That is how server power is generally measured, freq * cores