Intel, aka chipzilla, has been ruling the performance desktop processor market for what has been an age. Ever since the debacle that Pentium IV was (at least all Pentium IVs beyond Northwood), Intel has not looked back. It check mated AMD with its Core architecture. This was followed by the current Nehalem architecture delivering a technical knock out (TKO; to AMD).

Intel has seen its revenue grow from quarter to quarter despite the recent recession, while AMD has only just managed to return to black, and that too only after spinning off its foundries (now called Global foundries) and getting some much needed cash from Intel in a out of court settlement.

While Intel’s top of the line processors have no match, AMD does much better when it comes to the value sector. AMD’s recent Thuban based 6 core processors cost about a 1/3 of what Intel sells its Gulftown (Core i7-980X) processor for. They might not be able to match Intel’s prowess, but they deliver good performance for their price.

Similarly AMD’s quad core processors (Athlon II X4) can be bought for as low as US$ 100, while Intel’s cheapest true quad core processor is at least around 50 bucks more (when comparing MRSP, street prices might be different).

Intel’s latest generation of processors, the Core i series, only recently stepped into the mainstream (read budget) territory with the advent of the Core i3 series. Based on the same architecture that powers its big brothers (the Core i7 & Core i5 series), the i3 series is meant to give AMD a run for its money in the budget sector.

Today the product under my inquisitive eye is the Intel Core i3-530 processor, the youngest of the core i series.

INTEL’S REJUVINATION –THE TICK-TOCK MODEL

When AMD used to rule the performance roost with Athlon X2 series of processors, Intel was languishing with its super hot revision of the Netburst architecture (Presler, lovingly called the “pressure cooker”). It might sound strange but Intel, at that time, had a much better processor in the shape of the Pentium M, a heavily modified version of the Pentium III processor. Pentium M was meant for mobile devices (read laptops) and was very energy efficient for its time.

Fast forward to 2006 and the introduction of the tick-tock model. The same design team that was responsible for Pentium M came up with the Core architecture (more correctly the 64bit Core architecture). The desktop variants were sold under the name of Core 2 –Duos (for dual core CPUs) and –Quads (for quad core CPUs). There were oddities like the Pentium Processor for Desktop (also called Pentium Dual Core), which lead to a lot of confusion as there was already a Pentium-D (remember Presler).

With the introduction of the Core architecture a new design philosophy was adopted by Intel. This was called the Tick-Tock model.

The model basically implies either a revision of an existing architecture (the “tick”) or the introduction of a new architecture (the “tock”). Each year there has to be one tick or tock. This is nothing new; the computer graphics industry (read Nvidia) was already following this model. In fact Nvidia had a 6 month tick-tock cycle for a very long time (until the dreaded delay of the 5 series).

The following figure shows Intel’s Tick and Tock since then:

The clock started with a “Tick” as the Netburst architecture had its last showing with Presler based Pentium-D processors. The “Tock” was the storm in the shape of the Core architecture that powered the wildly successful Core2 processors. Wildly as not only that they were good (in fact very good), but also because AMD had no answer to them (and this sadly continues to be true today).

This was followed by the die shrink in the form of the E8xxx and Q8/9xxx series processors (the tick).

This process eventually led to the die shrink of the current architecture (Nehalem). The 32nm processors are Intel’s first foray into the budget segment with its new architecture. Why did Intel wait nearly two years before phasing out its Core 2 generation? Simply because they didn’t need to. AMD is still playing catch up (probably beyond Intel’s wildest expectations) and will continue to do so until its Bulldozer architecture comes to light (sometime in 2011). Intel has the luxury to slow down time as it sees fit.

As things stand today Intel’s formula seems to have worked well. Not only the tick-tock formula, but having multiple design teams working simultaneously with the “winner takes all” stake.

INTEL’S NEHALEM MICRO ARCHITECTURE

Nehalem is the name of successor of the Core micro architecture (if you are wondering how Intel applies code name to its processors –they are usually counties in USA or historical sites in Israel). Without going into too much technical detail, the Nehalem improved upon its predecessor by:

Being natively Quad core: Rather than gluing two dual core processors together, Nehalem processors were natively quad core)Integrated memory controller: Like the Athlon X2 and subsequent Phenom processors), featuring a triple channel controller (only for the Core i7-9xx series)Introduction of a Level 3 cache: A combined pool of cache that all processor cores share. There is dedicated L2 cache per core.Use of a faster bus: Called the Quick Path Interconnect (QPI) which replaced the Front side bus. (The Core i7-8xx series and lower use the DMI bus. The Core i3-5xx and Core i5-6xx processors use the QPI bus to interconnect the CPU and Graphics core & memory controller on the same chip)Return of Hyper-threading: Hyper-threading was a feature of the Pentium 4 processor, but was missing from Core2 series.Turbo boost: A processor can increase the speed of its processing cores as long as it stays within thermal parameters.

The first processors based on this architecture were the Core i7 9 series (the Bloomfield series –launched in November 2008). They featured all the improvements listed above. They co-existed for a long time with the Core2 series. The reason being that they (Core i7 processors) needed a new expensive board (QPI bus added to the cost), and DDR3 memory. It was not until the introduction of Lynnfield processors (September 2009) that Nehalem became “affordable”. Lynnfield was a much different beast as compared to Bloomfield, but was still based on the Nehalem architecture.

To reduce cost, the QPI bus was replaced by a simpler Direct Media Interface (DMI) bus. This bus did not have enough bandwidth to carry video signals (for gaming). To over come this, Intel moved all of the PCI-e lines on the processor die. This necessitated a new socket design. Bloomfields used an LGA 1366 socket; Lynnfields use an LGA 1156 socket. The thermal profile of the processors was also reduced from about 130 Watts to 95 Watts. There were some other changes as well like improved virtualization and turbo boost. The memory controller was cut down from a triple channel to a dual channel variant.

As the PCI-e lines had moved to the processor, there was no need for a “North bridge” chip. (The memory controller in already on the processor for all Nehalem based processors). This also helped to further cut the motherboard manufacturing cost. LGA 1156 boards only feature what was once called the Southbridge, which Intel now calls the Platform Controller Hub or PCH. The PCH has enough circuitry to drive USB and SATA ports.

The first processors that are based on Lynnfield revision are the Core i7 8 series and the Core i5 7 series processors. The latter does not feature hyper-threading, but otherwise carries all the features of its elder brethren.

THE NEHALEM TICK –WESTMERE & CLARKDALE

After what seems like an eternity Nehalem finally comes to the mainstream segment. The 32nm die shrink of the Nehalem is dubbed “Westmere”. The dual core variants of this architecture are known as “Clarkdale”. The 32nm Clarkdale processors are the latest edition to the Intel family of processors. Where Westmere (itself) takes the performance crown with 6 cores and 24 MB of L3 cache (and a 1000 dollar price tag), the Clarkdale series are for the rest of us. It has taken Intel more than a year to bring Nehalem to everyone. Intel could have done it sooner, but poor economy and more importantly no competition probably made them reconsider their plans.

The Clarkdale processors are branded as Core i3-5xx and Core i5-6xx. There is a Pentium branded processor in the mix as well.

What separates Clarkdale from Lynnfield? Firstly these are 2 cores/ 4 threads processors, rather than having 4 cores/ 8 threads (except the Core i5 750).

Secondly they feature an on-chip video processor. Notice I wrote “on-chip” not on die. That will happen in early 2011 with Sandy Bridge (the next “Tock”). Though the processor die is manufactured on a 32nm process, the graphics core is still fabricated on a 45nm process. The core is called “Intel HD Graphics” and is a souped up version of the Intel G45. Though better than its predecessor it is still no match for discrete graphics. For all serious gaming a discrete graphics adapter is a must. The graphics core also contains the memory controller (which has been moved off the main CPU die) as well as the PCI-e lanes.

Apart from the Core i3 series, all 32nm processors also feature better encryption thanks to the addition of new instructions. Though they won’t make a world of difference to most of us, they can speed up Windows 7 bitlocker performance.

The Core i3 series are also missing the turbo-boost feature found on the Core i5-6xx series processors.

The 6xx series is too expensive to be attractive. It is the 5xx series (starting at US$ 117) that will give AMD a run for its money. AMD has no answers (yet) for Intel’s top dogs. But it has been doing pretty decently in the budget segment, offering 4 core processors for a measly US$ 100. AMD also offers “Black Edition” (BE) processors at about the same price as Multiplier locked processors (though without a Heat-sink fan One is not need –No one who buys a BE will run it at stock!). With its first near 100 dollar processor, Intel finally has a weapon to counter AMD.

The following table lists all the current Clarkdale processor and specifications.

All processor have the same amount of L2 cache i.e. 256KB per core.

Intel HD Graphics characteristics are detailed below

INSIDE THE CLARKDALE PROCESSOR

The Clarkdale processor is a two die solution. The smaller of the two is the CPU fabricated on a 32nm process. The larger of the two is the HD Graphics die which is fabricated on a 45nm process.

The 32nm CPU features 383 million transistors in an area of 81 mm2. The larger 45nm HD graphics/ Memory Controller has an area of 114 mm2 with only 177 million transistors. The disparity is because of the difference in fabrication process. The smaller 32nm process results in a final product that packs more transistors in a smaller die.

As you can see the Memory controller and the PCIe lanes have been kicked off the main CPU die and onto the HD graphics die. The two are connected by the ultra fast QPI bus. Moving the memory controller off the CPU die will do no favors for RAM latency. The memory controller is still the same as for the Lynnfield series of processors (Core i7-8xx and Core i5-7xx processors) –dual channel.