Today NVIDIA announces nForce2, the follow-up to their first foray into chipset market with nForce. As you would expect nForce2 expands on the lessons learned from nForce and takes integration to a new level, whilst also removing some key areas of integration in some configurations.

Here we will have a look at what comprises the nForce2 chipset and what configurations will be available to motherboard manufacturers and systems integrators.

Video or No Video?

One of the main marketing features of NVIDIA’s original nForce chipset was the fact that it contained a GeForce2 MX graphics core integrated into the chipset -- this was the most powerful integrated graphics chipset available at the time. However, this was as a point of discussion amongst many of the retail PC up-graders out there as most of them would not be using the integrated graphics, but the more powerful discrete add-in graphics boards available. The chipset obviously had plenty of attractive features however there was the question of need for paying for the cost of the integrated graphics when many users will not be making use of it.

The original nForce made significant inroads into the available AMD OEM market, partly due to the inclusion of this integrated 3D graphics and the fact that the strong brand ‘GeForce2’ was associated with it. However, NVIDIA have been analysing the AMD market and realised that it is heavily channel based, which tends to utilise add-in graphics boards, and they had not made much of an impact here as they had with the OEM market. With nForce2, however, NVIDIA intends to have a product for everyone, and to this extent they will be introducing essentially two ‘Northbridge’ parts: the nForce2 SPP for discrete graphics implementations and nForce2 IGP with integrated graphics.

The nForce2 SPP’s basic premise is to achieve the highest possible system performance possible with an add-in graphics board, so this will not feature any kind of integrated graphics at all. The IGP, however, carries the torch from nForce with its integrated GeForce2 MX and furthers that by utilising in integrated GeForce4 MX core. Using the GeForce4 MX core does mean that ostensibly this is DX7 compliant graphics capable, but it does feature all the elements that the discrete GeForce4 MX does, which includes Accuview AntiAliasing and the VPE with its comprehensive support for NVIDIA’s nView display and multi-monitor technologies. Because of this the nForce2 based products that utilise the IGP will have capabilities for natively supporting two VGA’s/CRT’s and will even have integrated TV-out functionality.

As graphical power will play such a key role in any of nForce2 configurations both variants will be of AGP8X specification. Motherboards that use the IGP can still be populated with an AGP8X connector so that users will still be able to upgrade from their integrated graphics should they wish.

DualDDR

The original nForce chipset featured NVIDIA’s ‘TwinBank’ memory which utilised two 64bit busses so that the integrated graphics performance did not impact on the overall system performance. With all configurations of nForce2 NVIDIA will have a newer memory interface, dubbed DualDDR, with the onus not just on better integrated graphics performance, but also overall system performance because they are targeting the add-in market as well.

Likewise with TwinBank, DualDDR also features two independent 64bit memory controllers which, when operating together, will give a total of a 128bit memory bus. With these busses operating simultaneously, or slightly out of phase, this also has the effect of reducing the overall latency of the system, as, for example, one controller could be reading or writing to main memory whilst the second controller can be preparing for the next access. NVIDIA have also worked on their algorithms to further extend the possible performance of the DualDDR bus.

When utilising the twin 64bit busses in conjunction with DDR-400 RAM this can deliver up to 6.4GB of data bandwidth – with AGP8x in operation, which potentially utilises 2GB of bandwidth, and the Athlon FSB also using more then 2GB’s bandwidth a standard 64bit bus using DDR-400 RAM would not meet the maximum operating requirement of these two busses only delivering a total potential bandwidth of 3.2GB’s.

The DualDDR memory controller features 3 memory address busses, each capable of addressing 1GB of RAM, meaning that if all 3 banks are filled on a motherboard a total of 3GB of system RAM can be used. NVIDIA have also taken their asynchronous clocking memory bus successes from their graphics cards and utilised them effectively in nForce2; with the Althon FSB being 266MHz to support RAM other than DDR-266 an effective asynchronous bus must be used to support DDR-300 and DDR-400 memory types, which the nForce2 does.

nForce2 should also appeal the overclocking crowd. Because of the use of asynchronous memory busses NVIDIA will be providing the ability to tweak the speeds of the various busses independently from one another.