Without going into much tech detail, a quick run down on the brand new SB-E platform. What happened is that the Intel engineers combined the best out of the socket 1155 and 1366. The well known ring architecture from the Sandy Bridge CPUs, of early this year, is still being used. There's absolutely no need to change a winning formula. Low power consumption, amasing performance, TurboBoost V2.0  and high overclockability for the two K skews are the main trademarks of the Socket 1155 SB lineup. Combine all that raw power with newer instruction sets, alike AVX, for ultra fast rendering and you have a very potent package in one die. The high end SB model, the 2700K, still only has got 4 cores and 8 hyperthreaded ones. Maybe one of the reasons Intel insists on calling socket 1155 a mainstream platform. But this new architecture blew many of it's more expensive high end socket 1366 CPUs away. Only the 980X and 990X had a chance to withstand this new arrival due to being hexacore based, thus allowing for faster simultaneous threading in appropriate coded applications. The engineers of Intel created the ultimate cocktail for the SB-E platform : mix most of the socket 1155 goodness together with 6 cores, plus increasing the PCI-E lane capacity of the mobo chipset for better multi GPU, all this to raise the performance bar up to another level.

As mentioned in the 3960x article, two dies are faded out, meaning 8 Cores and 16 HT ones are coming early next year. Though this 8 core CPU is meant to be for the server variant only CPU types. The well known Tick Tock schedule indicates we are still dealing with 32nm fabrication. Next Year Ivy Bridge (again socket 1155) will be produced in 22nm. Rumour there is that it will still be 4 real cores only and not 6 as speculated. But that's for next year, focus Leeghoofd, focus...

Todays tested hexacore CPU, the i7-3930K shares the same unlocked multiplier policy as the socket 1155 SB K skews. What makes it different then from the 3960X (Extreme) model, costing close to twice as much. The answer my friends is firstly the amount of shared Level 3 cache. The extreme version has got 15Mb to work with, the K one only 12Mb. Steep price to pay isn't it for just 3MB more ? Level 3 cache is important though, as we have seen in the past. Eg the difference from 3 vs 6Mb was pretty worthwhile ( eg Wolfdale CPUs)  That was double the cache amount. But going from 12 to 15Mb, doesn't sound that impressive.

Besides the smaller L3 cache, we notice one lower stock multiplier at 32 iso 33X. Nothing that can be overcome via a little overclocking :) The listed i7-3820, will have limited multiplier overclocking capabilities. Still being a quadcore and a L3 cache of 10Mb, it could be a quite interesting processor for the year 2012.

For those uncommon with Turbo Boost V2.0. I think the below picture explains more than a thousand words would ever do. It's really very fast and reacts "intelligently" to the load versus the TDP of the CPU. Some motherboard manufacturers like Asus, have locked the Turbo Boost to all cores at the max setting. Meaning eg with an Asus board you would have all cores running at 3.8Ghz iso of different multipliers for each core. This gives quite a boost at stock clock tests and it's all free... so pick your motherboard wisely if you intend to run stock clocks...

For more in depth tests like RAM dividers, Dual versus Quad Channel ram and co, I would like to refer to the previous 3960X SB-E article.