Asus denies existence of Nforce FX dual chipset

Yesterday, [H]ardOCP ran a news brief and a diagram of the chipset, which can be seen here.

The fact that Asus denies the board's existence doesn't actually mean it doesn't exist -- companies often decline to comment on products they then launch days later.

The problem with discovering the board's existence is that it simultaneously fits into the categories of "wishful thinking" and "damn good idea." A properly designed dual AMD board could catapult Nvidia's reputation as a serious provider of business/workstation level products in addition to chipsets aimed at the home user in an arena its competitors are unlikely to match.

Via's name tends to bring to mind high performance but rather fussy equipment and SiS has been a budget chipset provider for most of its life.

AMD would probably be the best candidate for producing an updated dual AthlonMP motherboard -- just as businesses like to mate an Intel board with an Intel chip, they'd do the same on the other side of the equation -- but NVIDIA is probably number two on the list.

The most exciting thing about a potential GeForce FX chipset is the potential for NVIDIA to use its dual channel DDR technology and provide each CPU with a single channel of memory independent from the other. Historically, Symmetric Multi Processing (SMP) systems have been limited by the need for the two or more processors present to share a single front-side bus and memory controller. Adding four CPUs might quadruple your raw number crunching ability, but only so long as the bus didn't saturate before hand.

AMD's 760MP(X) chipset gave the AthlonMP two independent FSBs by connecting a Hyper Transport* link to each, but the two CPUs still share memory bandwidth. An Nvidia-implemented dual DDR channel could lift this restriction and potentially give a significant boost to SMP performance in bandwidth-intensive applications.

That factor alone would make the board attractive to dual enthusiasts but such a design would bring a host of new goodies to dual AthlonMP, including 8X AGP, SATA support, DDR400 support, and FireWire. We've written in the past on the need for AMD to update its 760MPX chipset, but if the company lacks the resources to do so, what better solution than to team with Nvidia to provide a third party solution? Even though Nvidia is relatively new to the Socket A market, the company's close partnership with AMD the last two years has doubtlessly given them insight on how to effectively build a Socket A chipset—the success of Nforce and Nforce2 proves it.

The largest reason AMD might shy away from such a chipset is if it felt it would interfere with Athlon 64, which will ship in a dual form as well, though not immediately at launch. Since we wouldn't see boards based on a GF FX chipset until early fall at best, this new Socket A chipset could be establishing itself in the dual market just as AMD is trying to push Athlon 64 into that range.

While there is the chance that AMD could end up competing with itself by pushing dual 32 and 64 bit solutions simultaneously, there's also a chance for tremendous gain. A dual Hammer system should still be significantly faster clock-for-clock than a hypothetical Athlon MP system, but being able to offer a strong 32-bit solution could give AMD additional leverage to position dual Athlon 64 at a higher price point and with a greater profit margin.

Perhaps this is one case where fantasy should make the jump to reality. The hypothetical dual GeForce FX chipset isn't a high-flying idea making a weak swipe at usefulness in the here-and-now—this is a product that would deliver solid, real performance increases, yet doesn't look so advanced as to be impossible for modern design to create. We welcome any clarifying information Nvidia, Asus, or AMD can provide. Here's hoping we don't have to wait until Computex in June to find out. µ

* CORRECTION Course it used the old Alpha front side bus, not Hypertransport, which was just a gleam in someone's eye way back then.