Intel launched its much-anticipated Nehalem-EX
processor today; the new chip could redefine performance at 4P and above. Nehalem-EX, aka Beckton, aka Xeon
6500/7500 is Nehalem at full throttle; Beckton is an eight-core/16-thread processor with up to 24MB of L3 cache or ~3MB per core compared to Nehalem's 2MB/core design. It supports up to four QPI links for glueless design implementations and features a quad-channel memory controller capable of controlling up to 16 DIMMs per socket. Intel has released a number of SKUs with different core counts, one without Hyper-Threading, varying amounts of L3 cache, and clockspeeds ranging from 1.73GHz-2.26GHz for the Hyper-Threaded CPUs and 2.66GHz one model of the new Xeon family that doesn't support it.
Beckton, like Magny-Cours
, is an evolutionary step forward; it's the CPU's interconnect architecture that makes it revolutionary. Prior to Beckton, if you wanted a four-core Xeon motherboard, you used a chipset with a configuration that looked like this:
This is Intel's highest-end Xeon 7300 chipset with four independent front-side busses connected to a memory controller with a massive 64MB snoop filter to eliminate unnecessary cache traffic. Even so, the system is enormously inefficient. With up to six cores per socket fighting for a single FSB and no way to quickly cross communicate, Intel's ability to scale its FSB designs couldn't keep up with core counts. This, incidentally, is part of why AMD
did fairly well in 4P performance comparisons even before Shanghai; the Direct Connect architecture allowed Opteron to scale more efficiently than the old Core 2 Quad-based Xeons.
Beckton changes everything.
Intel's new core now uses an AMD-style point-to-point interconnect in the 4P space and above; a move that could cut access latencies by an order of magnitude in some cases. At 16 DIMMs per socket, the CPU is built for massive big iron deployments; Intel specifically noted it'll be deploying Beckton in 256-socket configurations and above.
While Beckton should theoretically plug into any motherboard that supports Core i7 Xeons, Intel is clearly targeting enterprise servers, HPC clusters, and large-scale virtualization efforts.
Yesterday we talked about AMD's new Opteron strategy and how the company has positioned
its 8-12 core processors to challenge Intel in the 2P-8P market. Everything we said there still holds; Intel's new eight-core Xeon 7500/6500 processors offer incredible performance and even better scaling than AMD's Magny-Cours, but they're positioned for extremely high-end markets and socket configurations. The two companies will cross swords from time to time, but the price premium and raw performance of the Xeon 7500 series will create a buffer zone between the two architectures. We don't have 1K unit pricing yet, but you can bet these babies won't come cheap.
Beckton, as compared to dual-core Netburst, 65nm Core 2 Quad, and 45nm/Six-Core Dunnington.
Beckton's topology allows the CPUs to work at their full bandwidth.
The Xeon 7500 could quickly become Intel's new gravy train
; Gartner estimates that IT companies delayed upgrading one million servers due to the economic downturn. Given the tremendous performance improvements Beckton can deliver, we expect a fairly quick adoption rate in appropriate markets—if your an IT Director managing 4P systems or above, performance numbers like the ones above should make you rather happy.