Whoosh, now certified for Intel, natively.
Nvidia will not limit its SLI multi-GPU technology to next generation X58 motherboards that contain the NF200 chip. Previously Nvidia has limited its SLI technology to its own chipsets but without a QPI license on the table which is needed for the next generation Nehalem (Core i7 and others) platform, Nvidia had to limit it to future motherboards containing its PCI-Express splitter chip.
We've heard unconfirmed reports that Nvidia charges $30 for the privilege, or $20 if you only make X58 motherboards with SLI on them but as far as we know no one in Taiwan has taken up this offer.
, uh, reports, that just after Nvision finished, Nvidia gathered together press to tell them about this very recent decision. This is fantastic for us consumers - you can now buy any standard X58 board with the SLI stamp on it and have ONE platform for both AMD and Nvidia multi-GPU. Finally.
But then again, why now pay more for an X58 motherboard with NF200 bolted on, when non-NF200s will do? Will it infuriate those already invested time and money into designing boards with NF200 chips on them?
Nvidia is claimed to have stated that it realised the decision to limit SLI to niche very high-end motherboards was a business mistake and that it had effectively removed itself from the larger mainstream and most of the enthusiast market. It was keen to note though, that motherboards must adhere to certain conditions in order to be granted with SLI certification; these include submission to Nvidia's Santa Clara labs where the it will be tested for functionality, slot placement, product compatibility etc.
Oh and the testing won't be free - so naturally the costs will be uniformly passed onto the consumer, so (cynically speaking) even if you don't use SLI you'll be paying for it in one way or another. This is without doubt a clever decision by Nvidia to put the focus on its SLI brand rather than an optional, physical chipset.
Nvidia will then provide an encryption key (called a "cookie") that'll go in the motherboards BIOS which will be sniffed out by the Forceware drivers and will unlock functionality. No doubt the key will also be an identifier to a particular motherboard, so any mobo maker caught skipping testing of future products might find its SLI license taken away in future Forceware updates.
Our immediate question is that of economics: how much does the certification process cost and will it force the overall X58 price up enough to compensate for the lack of NF200 chipset, since those already with the NF200 won't need to be certified? In some respects, we expect no motherboard manufacturer will want to shy away from not having SLI support since it has massive brand leverage and is great marketing material. However, given the fact that the X58 development cycle is already stretching the motherboard manufacturers to the limit (as does every major chipset launch), we also half-expect SLI certification to come post launch with future BIOS updates. In addition, what stops an enterprising individual making beta BIOS', hacking a cookie from a certified BIOS and dropping it into non-certified ones though?
In the past we've seen hacked Forceware drivers allow SLI on Intel chipsets since it was just a matter of PCI-Express lanes and software, however Nvidia claims its SLI now specifically requires the two major functions of its NF200 chipset: PW Short and Broadcast writes. PCI-Express 2.0 has peer-to-peer writes as a standard feature so there is no need for PW Short, and the Broadcast writes could now be done in software. If that's the case, what about SLI for the AMD 790FX then?
The problem for Nvidia is that by opening up SLI uniformly like ATI has done, it loses on Nvidia chipset sales and a big chunk of its business and essential part to its platform-ideology. As AMD/ATI only now makes chipsets for AMD CPUs (I still cry for the RD600), getting ATI cards on a competitor platform in multi-GPU is almost a bonus. Nvidia makes chipsets for both CPU types though and wants to keep leveraging its position as dominant GPU manufacturer to take advantage of this with extra features like ESA, EPP(2) etc where it can directly control them.
So is this good or bad news for those wanting an X58+i7? Let us know your thoughts, in the forums