Home / Opinions / Intel Core i7 (Nehalem) Architecture Overview



Intel Core i7 (Nehalem) Architecture Overview

Core 2 Duo sparked something of a revolution for Intel. Here at the end of 2008, the average consumer probably won't remember the bleak days of Netburst and Pentium 4 when AMD's K8 Architecture and Athlon 64 chips reigned supreme. Core 2 Duo and Quad brought brilliant performance at compelling price points, and consequently swung the balance well and truly back in Intel's favour. So, it has to be said that Nehalem, Intel's successor to Core 2, has some big shoes to fill.

Intel's last generation of CPUs, built on the Core Microarchitecture (codenamed Conroe), and subsequently enhanced using a 45nm process (codenamed Penryn) are, let's face it, fast enough for all but the most demanding of corporate consumers so there's no compelling argument for a successor to be unleashed.

Moreover, AMD's Phenom processors can't compete with Penryn chips on performance so it can't be argued that Intel is usurping the competition, either. Because it isn't. So, thus it is that I'm left with the inevitable conclusion that Intel is simply doing its best to prove Gordon Moore (who's theory states that the number of transistors in integrated circuits will double every couple of years) right, and cram yet more transistors into its CPUs. Not, I hasten to point out, that there is necessarily anything wrong with that strategy. And let's face it, we'd all rather be questioning whether we need the power of a new generation of chips, than complaining that our current hardware isn't fast enough.

It's not as if Intel isn't happy to disclose the way its development cycle is set to manifest to the public, at least in broad terms. From what we know thus far, the current Tick/Tock development cycle is set to continue until at least 2012 with a microarchitecture known only as Haswell. Each tick marks a shrinking and refinement of the manufacturing process while each tock will introduce a brand new architecture; all on a two year cycle.

Nehalem, then, is the tock to follow the tick that was Penryn and there's a lot to be excited about. Not least because Nehalem is probably the most fundamental architecture revision Intel has made in some 13 years. Anyone with a passing interest in Nehalem will know by now that this is the first mainstream Intel CPU to drop the front-side bus in favour of the Common System Interface, or Intel QuickPath Interconnect (QPI) and, in doing so, bring the memory controller on-die.

Intel's critics may want to label that as Intel copying AMD, but I'm not sure I'd agree with such a view. The company proved with Core 2 Duo it's actually quite easy to make a fast processor without an integrated memory controller as you can, to an extent, hide the performance deficiencies inherent therein by doing things like adding masses of extra on-die cache. Enough of that for now, because I'll be talking in detail about Nehalem and RAM later.

Integrating a memory controller isn't all Intel has done with Nehalem, though. The only useful development (call me controversial if you will) of Intel's Pentium 4/Netburst era, simultaneous multithreading, makes a return too. That's right, folks, HyperThreading is back with a vengeance. Thus, each Nehalem core is capable of executing two threads at any one time.

Finally Intel has also given Nehalem some pretty cool power management features. There's a big focus not only on saving power, but also on making the best use of the power being drawn by the CPU.

All in all there's a lot to talk about, so let's get started.

comments powered by Disqus