Sandy Bridge-E And X79 Are Almost Ready
There was a lot to like about Intel’s Sandy Bridge launch earlier this year. Single-threaded performance increased significantly at any given frequency. Quick Sync demonstrated commanding dominance over GPU-based transcoding from AMD and Nvidia. And, although I wasn’t over-enthused about paying extra for a K-series SKU, a mature 32 nm process easily facilitated clock rates approaching 5 GHz on air cooling.
Combined, all of those attributes took the spotlight off of Intel’s old (but still flagship) LGA 1366 interface. Even the subsequent Core i7-990X refresh, which threw six cores and a higher clock rate into the ring, wasn’t able to outperform the Core i7-2600K in enough test scenarios to warrant its $1000 price tag. The very fastest (and most expensive) Sandy Bridge-based chip could satisfy 95% of enthusiasts at less than half of the cost.
The Gulftown design’s real redeeming quality was its core count advantage, which shone most brightly in well-threaded workstation apps. But really, that was pretty much it. We even went to great lengths to show the X58’s 36 lanes of PCI Express 2.0 weren’t a real advantage over Sandy Bridge’s 16 lanes in multi-GPU configurations through an exhaustive three-part series.
At the end of the day, we had to scratch our heads and wonder how many folks would be willing to spend almost $700 more on Core i7-990X when Core i7-2600K was already so fast, and priced at $315.
But what if it was possible to cram what originally made Gulftown sexy into the Sandy Bridge mold? That’s exactly the premise behind Sandy Bridge-E, set to become the next enthusiast-oriented platform, replacing Gulftown and its LGA 1366 infrastructure.
More important than what Sandy Bridge-E is going to do on the desktop is what it’ll become in the server space. Truly, this is a design destined to drive Intel’s Xeon E5 family, comprised of 1P-, 2P-, and 4P-capable parts.
A Naming Convention, Revised
For the time being, Sandy Bridge-E is expected to reach enthusiasts in three different trims: the Core i7-3960X, the Core i7-3930K, and the Core i7-3820.
|Second-Gen Core i7 Processor Family|
|Processor||Base Clock||Max. Turbo Clock||Cores / Threads||L3 Cache||Memory||Interface||TDP|
|Core i7-3960X*Unlocked||3.3 GHz||3.9 GHz||6/12||15 MB||4-channelDDR3-1600||LGA 2011||130 W|
|Core i7-3930K*Unlocked||3.2 GHz||3.8 GHz||6/12||12 MB||4-channel DDR3-1600||LGA 2011||130 W|
|Core i7-3820*Partially Unlocked||3.6 GHz||3.9 GHz||4/8||10 MB||4-channel DDR3-1600||LGA 2011||130 W|
|Core i7-2600K*Unlocked||3.4 GHz||3.8 GHz||4/8||8 MB||2-channelDDR3-1333||LGA 1155||95 W|
|Core i7-2600||3.4 GHz||3.8 GHz||4/8||8 MB||2-channel DDR3-1333||LGA 1155||95 W|
|Core i7-2600S||2.8 GHz||3.8 GHz||4/8||8 MB||2-channel DDR3-1333||LGA 1155||95 W|
Although the model names suggest that Intel might consider this a third iteration of its Core micro-architecture, the press decks I’ve seen clearly list the three new Sandy Bridge-E parts as “second-generation Core i7s.”
By now, we’ve had increasingly-confusing names beaten over our heads by so many companies that the inelegance of “Core i7-3960X” bounces right off. Intel Core i7—OK, that part’s easy enough. The “3” is a generational reference, and the “960” is the actual model number. Incidentally, 960 doesn’t seem to compare favorably to the outgoing Gulftown-based 990. But Intel didn’t really give itself much room to maneuver there.
Even the letter suffixes are familiar by now. The “X” at the end of -3960X represents Intel’s Extreme Edition family—a designation generally reserved for one SKU at any given time at the top of the desktop stack. The “K” at the end of -3930K denotes lower-end, but still multiplier-unlocked models, also geared to enthusiasts. And the fact that the -3820 bears no modifier suggests it’ll follow in the footsteps of Core i7-2600 and i5-2500, offering limited overclockability (a handful of 100 MHz bins over and above the top Turbo Boost frequency, if history is any indication).