Introduction

While NVIDIA officially “introduced” the nForce 590 SLI and 570 SLI chipsets for the Intel platform when Intel’s Core 2 CPUs launched last July, as any enthusiast will tell you, finding motherboards based on these chipsets has been a tough endeavor.

In August ASUS released their nForce 570 SLI-based P5NSLI motherboard (which we reviewed in our ASUS Core 2-ready motherboard roundup), and Dell uses NVIDIA’s nForce 590 SLI chipset in their XPS 700 systems, but other than that, nForce 500 motherboards for Core 2 have been hard to come by. It has been rumored that motherboard manufacturers have been passing on nForce 570/590 because it doesn’t scale well to higher FSB speeds. Based on end user overclocking reports with NVIDIA’s nForce4 SLI X16 chipset, which the nForce 590 SLI is largely based on, this theory may certainly be true, as these chipsets barely overclocked beyond FSB speeds of 266MHz. In comparison, Intel’s own P965 and 975X motherboards have proven to hit speeds of 400MHz or more; that’s 1.6GHz effective FSB!

With nForce 570 SLI and 590 SLI getting off to such a lackluster start, NVIDIA quickly went back to the drawing board and has now cooked up another new chipset for Core 2: the nForce 600 family.

While the nForce 600 moniker suggests this is an entirely new chipset platform, actually it’s not. NVIDIA uses the same basic South Bridge chips for nForce 600 that were used for nForce 500 previously. That means all of the South Bridge features such as audio, networking, storage, and USB carry over unchanged, however NVIDIA has employed an entirely new North Bridge chip for nForce 600.

nForce 680i SLI

Fundamentally the new North Bridge in nForce 680i looks a lot like the North Bridge NVIDIA used previously in nForce 590, it supports many of the same key features found in nForce 590 such as 1066MHz FSB operation, but nForce 680i adds support for DDR2-800 and NVIDIA’s 1200MHz “SLI-ready” memory (nForce 590 was officially limited to just DDR2-667 memory). NVIDIA has also completely redesigned the nForce 680i’s memory controller. Latency has been reduced in the prefetch unit of the new memory controller and it has been optimized to run better with Core 2’s new unified memory architecture.

Like the nForce 590 SLI, the nForce 680i SLI splits the PCI Express graphics duties between the North and South Bridge of the chipset. You’ve got one 16-lane connection on the North Bridge, and a second 16-lane graphics connection hanging off the South Bridge. Linking the two chips together is an 8GB/sec HyperTransport interface.

NVIDIA’s 680i chipset also provides a third PCI Express graphics slot, but it only supports 8 lanes. The premise behind this third graphics slot is for GPU-based physics processing. Ideally you’d be able to mix and match GPUs, so you could run a pair of GeForce 7900s or GeForce 8800s in SLI mode for graphics processing, and say a GeForce 7600 running in the third x8 slot for physics, but NVIDIA hasn’t announced anything yet so this is all speculative.

Built for 1333MHz(+) FSB

Of course, the feature enthusiasts want to see most nowadays is scalability. Simply hitting 300MHz isn’t good enough for most overclockers. Here NVIDIA’s nForce 680i chipset really delivers the goods. The North Bridge of the chipset has been designed from the start to hit 1333MHz FSB, even though 1333MHz desktop CPUs aren’t even available yet. In fact, NVIDIA’s documentation lists full support for 1333MHz FSB, and if you talk with an NVIDIA engineer about the topic, they’ll mention that speeds of 1400MHz or more aren’t out of the realm of possibility if the motherboard is built for the task. This is due in part to the chipset’s new 90-nm manufacturing process.