NVIDIA's recent announcement
of extending all its multi-GPU technologies to Intel chipsets-based socket LGA-1156 motherboards, even as the company reportedly plans
its own chipset, comes in at no better time than this, when rival AMD has a decent lineup of GPUs, processors, and desktop platform technologies, all of which well-oiled. To beat AMD in the game, and propagate its own GPU and multi-GPU technologies, some sort of loose alignment with Intel is inevitable, especially considering ATI CrossFireX has been freely available to motherboard makers for several product generations now.
In a recent presentation circulated to sections of the media, NVIDIA put forward a sort of quasi-platform to rival AMD Dragon, although it isn't named or defined, NVIDIA refers to it as "Power of 3". Part of its key components include Intel socket LGA-1156 processor (from the Core i3/i5/i7 series) running on a motherboard with Intel P55 chipset, Windows 7, and two or more NVIDIA GeForce GPUs. To deal with two or more GPUs, NVIDIA defines its existing "NVIDIA SLI Ready" marker and the seemingly new "NVIDIA PhysX Ready" marker. The difference between the two is that the latter lets you install a second (or third) graphics card that is dedicated to PhysX.
While doing so, it also grades P55 chipset motherboards into three tiers: There's a "mainstream" motherboard that has at least one PCI-Express 2.0 x16 slot for graphics, and a second PCI-E x16 slot that is electrically PCI-E x4, that should provide sufficient bandwidth for a PhysX-dedicated accelerator. Then there's a "performance" tier in which, the two (or more) PCI-E x16 slots rearrange as two PCI-E x8 slots when both are populated. This allows two GeForce accelerators to work in a 2-way SLI setup. In some motherboards, a third slot is electrically x4, this can handle the PhysX-dedicated card. Finally, there's the "extreme segment", in which motherboards usually make use of PCI-Express multiplex chips such as the NVIDIA BR-03, to ensure higher operating bandwidth for the graphics cards. This platform allows 3-way SLI, and in some boards with a fourth electrical PCI-E x4 slot to handle a PhysX-dedicated card too. Whew.
NVIDIA further made the licensing part a little more affordable for motherboard vendors. For the X58 platform, motherboard vendors reportedly have to pay a royalty of US $5 per SLI-supportive motherboard they sell. For the new platform, this royalty has been reduced to $3 per board, plus an upfront license fee of $30,000 a motherboard manufacturer has to pay once. A licensee then gets a BIOS micro-code entry that lets NVIDIA GeForce drivers recognise the motherboard as a qualified platform.
With quad-core Core i5 processors starting at US $196, and motherboards starting at well within the $150 mark, NVIDIA claims that it will provide consumers a better performing, and higher value for money platform (compared to AMD Dragon). As for AMD, it is on the brink of unveiling a new generation of GPUs, and will launch newer chipsets as the year progresses. The battle promises to be bitter for the long-standing market rivals, and hopefully sweet for you and me.
Sources: Hardware Canucks, Expreview