To make things clear, we can mention a few examples: for the notebook CPUs, there’s a lower energy consumptions, while for the server solutions, there could be bigger cache numbers. In general, analyzing the technical characteristics from the Nehalem CPUs when compared to the Core 2 family, it’s clear how the developing team aimedf at implementing features that would bring the best performance benefits on a server level, with an approach that is somewhat similar to wht AMD has done in the past with their first K8 family CPUs, Opteron and Athlon 64.

These are the base elements on the Nehalem family CPU’s.

- Native Quad Core architecture: Intel has abandoned the Multi Chip Package approach, choosing instead what is defined as “monolithic design” for the Nehalem CPU’s. The four cores, similar to the Phenom solutions from AMD, are integrated in the same silicon block instead of pairing two dual-core dice in the same package.

- DDR3 Memory controller integrated on the processor: it’s a new feature for the Intel processors, the integrated memory controller has been on the market for a while now with the AMD CPus since the K8 family, with the first Opteron CPU presented on April 2003.

- Integrated on-die L3 Cache in all processors, up to 8MB; and also, the size of the L2 cache, specific to each core, was noticeably reduced compared to what has been done in the previous Core 2 processors. In future versions, the Nehalem processors will feature differen L3 cache versions, according to the market sector where it belongs;

- Return of the Simultaneous Multi-Threading technology, better known with the market name “Hyper-Threading”, thanks to which the Operational System can recognize the processor as featuring a double number of logical cores than what’s physically integrated. This technology was introduced by Intel with some Pentium 4 models, but it wasn’t implemented in the Core 2 Duo and Core 2 Quad solutions;

- A new set of SSE 4.2 instructions, which are extensions of SSE4 instructions introduced for the first time with the Core 2 CPUs based on Penryn cores;

- QPI (Quick Path Interconnect) technology debut: it comes to replace the front side bus on the connection between Processor, memory modules and in some CPU models, also chipset. For the first Core i7 family models, based on LGA 1366 socket, the connection between the processor and chipset will be done through a QPI link.

The first Nehalem processor versions with quad core architecture, the Core i7 family solutions, will integrate 731 million transistors, built using a 45nm fabrication process. The following evolution of Nehalem processors will feature the same fabrication process and will have a modular architecture that was implemented on the Nehalem project while in its design stages.

These processors can, in fact, be easily modified in order to implement a different number of cores, or integrate different types of features internally, in comparison to what had been presented in the first versions before the launch.

Two examples can make this flexibility more clear: the first is the Nehalem-EX CPUs, solutions that feature eight physical cores to be used specifically in server systems, which will be launched sometime this year, and that have been first announced back in the IDF Fall 2008. The second is the integration in future Nehalem versions aimed for a low-entry market, of a GPU: with this product, Intel aims at presenting its own alternative to AMD’s Fusion Family CPU’s, which feature both CPU and GPU components.

The Nehalem-EX Advantage

• Intel Nehalem Architecture built on Intel’s unique 45nm high-k metal gate technology process
• Up to 8 cores per processor
• Up to 16 threads per processor with Intel® Hyper-threading
• Scalability up to eight sockets via Quick Path Interconnects and greater with third-party node controllers
• QuickPath Architecture with four high-bandwidth links
• 24MB of shared cache
• Integrated memory controllers
• Intel Turbo Boost Technology
• Intel scalable memory buffer and scalable memory interconnects
• Up to 9x the memory bandwidth of previous generation
• Support for up to 16 memory slots per processor socket
• Advanced RAS capabilities including MCA Recovery
• 2.3 billion transistors

Nehalem is the codename that indicates the new generation of Intel processors. The first Nehalem CPUs have the name of Core i7 and are destined to medium-high end desktop systems.

With these next few articles, we will be going through the architectural characteristics on these processors, so we can better analyze the performance results for these processors when compared to the Intel Core 2 Quad solutions currently in the market.

The importance of the Nehalem CPU is very high for Intel: it’s a completely new architecture, very different from what has been implemented by the company with the Core 2 family solutions. Intel follows a known approach with the Tick / tock terms. Each year, Intel presents a new generation of processors, that can either implement a completely new architecture, or be built using a more sophisticated technology.

Taking the Nehalem CPU’s as a reference, they can be inserted into what is called a “Tock” phase, which means a completely different microarchitecture generation from the previous ones. The same happened two years ago with the debut of the Meron family CPU, which later was called Core 2 Duo and Core 2 Quad, depending on the versions; in that case, the architectural change was done to replace the Pentium D CPUs.

The “Tick” phases indicate the uses of a new fabrication process: The Nehalem CPUs are built with 45nm technology, the same used on the Penryn family (Core 2 Duo and Core 2 Quad) currently available on the market. The “tick” phases also usually indicates an architectural restyling, with the implementation of some functions that are considered “minor”; for example, with the Penryn CPUs, besides the new fabrication process, there was also the introduction of SSE4 instructions that weren’t present on the Merom family.

The evolution to a 32nm process on the Nehalem CPUs is known by the name of Westmere; the processors belonging to this family will probably not be launched before the end of 2009, and also in this case with probably small innovations and enhancements, but without any changes on the architectures.

The choice of presenting their own solutions with this scheme comes from the fact that Intel has chosen to present new architectures by using a manufacturing technology that has been already launched, to avoid risking to end up with a CPU that doesn’t reach the expectations, instead of simply combining new architectures and new fabrication processes in the same line, which would be a lot more risky.