The interconnect
As with the first nForce, the two chips in the set will talk to each other over a HyperTransport link at up to 800MB/s. Because of the common interconnect type, NVIDIA's north and south bridge chips can be mixed and matched at will, even across generations. Chips from the first nForce will stick around as "value" parts, and NVIDIA expects mobo makers to pair up, say, an original nForce IGP with an nForce2 south bridge in order to build a cost-effective solution with support for all the latest I/O standards.

Incidentally, NVIDIA's use of HyperTransport probably means its south bridge chips will require little or no modification in order to work with AMD's upcoming Hammer processors.


The south bridge
NVIDIA calls its south bridge chips "MCP," for Master Control Program, the evil computer program from the movie "Tron."

However, that's just around NVIDIA HQ. To the rest of the world, MCP stands for Media Control Processor, which sounds much less threatening. The MCP handles various and sundry input/output functions, including the PCI bus. Because the MCP hangs off a high-speed HyperTransport link, I/O functions incorporated into the MCP (instead of on an add-in card) have gobs more bandwidth available to them than PCI devices. Also, HyperTransport can act as a traffic cop, allocating reserved bandwidth for critical or sensitive applications.

Like the north bridge, nForce2's MCP chips will come in two flavors. The big daddy MCP is called the MCP-T, and it includes the following improvements over the nForce MCP:

  • USB 2.0 support — The MCP-T will enable up to six USB 2.0 ports. USB 2.0 claims to be as fast as 480Mbps, which is one heckuva lot of bandwidth for peripheral devices.

  • IEEE 1394a — Yep, the nForce2 has Firewire support.

  • ATA/133 support — The latest (and last?) rev of the ATA spec before Serial ATA begins its inevitable takeover. But we know how "inevitable" things sometimes go.

  • Dual Ethernet controllers — To find a home in corporate PCs, mobos often need the 3Com name on their Ethernet implementations. NVIDIA maintains its Ethernet controller is faster and consumes fewer CPU cycles, but to keep OEMs happy, nForce2 will have a 3Com NIC integrated as well.

    Yep, I said "as well." NVIDIA just stuck both network controllers on the chip. With the MCP-T, some nForce2 boards may have two Ethernet ports—with no need for extra chips.


  • Dolby Digital 5.1 audio — OK, so this isn't an improvement over the original nForce, but I'm going to list it here anyway, because it's worth mentioning. The MCP-T will include the APU (Audio Processing Unit) from the first nForce, which supports real-time Dolby Digital AC-3 audio for true 5.1 surround sound. Unfortunately, many of the original nForce boards didn't include a digital S/PDIF connector, so the chipset's AC-3 capabilities weren't accessible. NVIDIA is looking to rectify that situation this time around with its SoundStorm branding initiative. Boards and systems with all the right connectors will get a SoundStorm tag on them to vouch for their completeness.
The other flavor of the nForce2 MCP is called simply "MCP." The MCP is a feature- and price-reduced version of the MCP-T. It will have only a single NVIDIA Ethernet controller, no Firewire, and AC97 audio instead of NVIDIA's APU. Beyond that, it's just like MCP-T, including support for USB 2.0, and ATA/133. Like the IGP/SPP, the two MCP versions will be pin-compatible with one another, so motherboard makers can switch between the two as needed.