Intel’s Mobile Nehalem on Calpella – What You’re Missing

Written by: J. F. Amprimoz • Edited by: Lamar Stonecypher
Published Feb 22, 2009 • Related Guides: Core i7 | Intel | CPU

Mobile Nehalem and the accompanying Calpella platform have been delayed again, but what are laptop users missing out on during this delay? We look at Calpella and the CPUs it will run: Clarksfield and Arrandale.

Laptop users are none too happy about Intel holding off on Calpella, and delaying Centrino 2 price cuts, until laptop vendors can get more of the old laptops sold. But what benefits are laptop users not getting until, at the soonest, September?

First Mobile Nehalem: The Quad-Core Claksfield

Clarksfield will be Nehalem’s first foray into the notebook market. It will likely share a lot of nice features from the desktop Core i7 line. This includes Hyper Threading, which makes each core able to process two threads of instructions, bringing a huge increase in performance for multithreaded apps.

Some version of what Intel calls Turbo Boost Technology will be included. On the desktop Core i7 this shuts off cores that aren’t in use. Using the heat and voltage headroom left by shutting off those cores, TBT lightly overclocks those still in use.

Hopefully, the laptop version will be configurable based on the laptop’s power state. Specifically, it would be great if it could use the overclocking side when plugged in or above a certain battery level, under which it would just do the shutting off unused cores part.

Clarksfield and the Calpella Platform

Intel 5 Series Platform for Clarksfield with DMI

click to enlarge

Changes to the CPU that affect the platform as a whole are numerous. Most noticeably, the PCI-e graphics and memory controllers move from the northbridge to the CPU. Clarksfield, and versions of the Calpella platform it will use, do not have integrated graphics. With the controllers on the CPU and no on board graphics, the Calpella doesn’t need a northbridge.

On the Ibex Peak-M chipsets used by Calpella (called simply Ibex Peak in desktop form, -M is for Mobile), the Clarksfield connects directly to the southbridge by Intel’s Direct Media Interface, as we see on the right.

DMI is currently used to connect the north and southbridges of Intel’s Series 4 chipset, where the graphics and memory controllers were on the northbridge, and connected to the CPU by Front Side Bus. See diagram below.

Intel 4 Series Platform with DMI and FSB

click to enlarge

Next Year’s 32nm Mobile Westmere Nehalem: The Dual-Core Arrandale

The next step for the Intel Nehalem’s march to mobility is Arrandale. Along with its desktop counterpart, Clarkdale (note there is no s, as opposed to the Clarksfield above), they will be the world’s first 32nm processors. It is similar to the Clarksfield, sharing Nehalem features like TBT and Hyper Threading, and Ibex Peak features like not having a northbridge.

click to enlarge

Other than the 32nm, Westmere, process (Clarksfield is still on 45nm, or Penryn) and being two instead of four core, the biggest difference between Clarksfield and Arrandale is that the later has integrated graphics. Yup, the graphics are integrated right onto the processor (well, right beside it, see picture at left).

That has many benefits in terms of size, less bottlenecking, power consumption, and even cost reduction. Unfortunately, the first CPUs with integrated graphics will use a somewhat improved version of Intel’s GMA 4500 currently found on Series 4 chipsets. The choice to modify the uncompetitive GMA 4500 for integrated CPU graphics is discussed here.

The resulting difference to the Calpella platform is that Arrandale version will have something Intel calls Flexible Display Interface to move display output to the southbridge (picture below). It still has DMI for other CPU-southbridge communication.

Intel 5 Series Platform for Arrandale with DMI and FDI

click to enlarge

Calpella as a Whole

Most of Calpella’s more noteworthy changes from Centrino 2 are related to taking advantages of the CPU, and we discussed those above. Calpella has a few other neat tricks, though, mostly on the power saving front. After all, the processor itself should offer a big boost on its own, if Core i7 performance is anything to go by. Similar to TBT (mentioned above) which can shut off unused cores, there should also be functionality to shut down chipset components that aren’t in use. Intel is also planning an improved WiFi system that should add 30 minutes to battery life.

When Calpella does arrive, the Ibex Peak-M chipsets will have similar names to the desktop Ibex Peak lineup of P, Q, H, Series 5 chipsets. This continues Intel's chipset naming system that began with Series 3 with one small change. Now that the graphics are on the CPU instead of the chipset, H replaces G as the chipsets that support CPUs with integrated graphics (the Arrandale described above) for home users. Q is the business offering, and P doesn’t have the FDI to support CPU-integrated graphics, as its meant to be used with discrete graphics.

As with the platform codename itself, add an M for Mobile. So P55 and Q57 on the desktop would become PM55 and QM57 on the notebook, respectively.

No Calpella Until You Finish Your Centrino 2

Nehalem’s performance isn’t needed for day to day tasks, and the numerous cores and Hyper Threading don’t really help with gaming. For those who have demanding professional apps to run, though, they are a real boon. Getting these processors mobile would actually allow for laptops that have some claim to being ‘portable workstations.’

And everyone can use less power consumption and more battery life. But, Intel won’t be rolling out Calpella until laptop vendors can clear existing Centrino 2 inventory.