Review Index:

NVIDIA's SLI Gets Busier - Notebook and Quad SLI Versions

Manufacturer: NVIDIA

Quad SLI Rendering Modes

The rendering modes available on the quad SLI technology are mostly very familiar ones and work just the way you think they would.

First, the quad SLI version of AFR simply uses each GPU core to render a frame, alternating over fourth frame instead of every other frame. 

Quad SLI is also compatible with a split frame rendering mode that also does what you would expect; instead of breaking the area of the screen into two pieces it divides it into four.  Since SFR uses slightly more CPU overhead when it has to divide up the processing, it would only make sense that quad SLI SFR mode is slightly less efficient than the dual SLI SFR mode, though the additional GPU processing power should more than make up for it. 

A new rendering mode being introduced with quad SLI is called AFR of SFR and is also pretty intuitive.  Basically, the logic goes that you take two GPUs and run a standard split-frame rendering technique off of them and you do this for both pairs of GPUs.  Then you use an alternate frame rendering method to alternate between the SFR frames being generated. 

Quad SLI is also introducing a new SLI AA option along with it dubbed 32X SLIAA.  The process by which NVIDIA's drivers generate a 32x anti-aliasing blend is pretty complicated, but a quick diagram might help.

Basically, each GPU is rendering a 4x MSAA blend and then it is combined between all four of the GPUs to get a final output that is technically equivalent to a 32x sample.  Each GPU on the same card shares the data to generate a single 8xAA sample and then each 8xAA sample is shared across the individual cards and blended into 16x samples and then finally combined to generate a 32xAA sample that is sent to the screen.

In all honesty, I don't expect this to be a very fast performing operation or one that will generate results that will be noticeably different than just standard 8xAA or 16x SLIAA.  As soon as we get our hands on one to test you'll know for sure.

Obviously when you are combing four of these super powerful GPUs into a single system, power constraints are a big concern.  NVIDIA didn't have final number, but they did mention that a power supply would need to be around 800-1000 watts in order to support quad SLI efficiently

In this sample system running at IDF 2006, we saw that there were four complete PCIe power connectors internally, one to each GPU, while in a previous design we saw external power supply connections on the cards.  I am unsure as of now which design NVIDIA has finalized but system builders are most likely to require the internal connections only in order to keep their high end system appeal in place.  Additional flimsy, external cables aren't going to do that. 

All the quad SLI products that are being announced today are going to be coming solely from system builders like Alienware, Voodoo PC and ABS.  NVIDIA did say they planned to productize the quad SLI technology for e-tail and the end user, but it wasn't going to happen right away and they wouldn't give a timeline on when it might occur.  Are you prepared to hop on and drop $1800 for a graphics system?  That is the same debate that NVIDIA is having internally. 

I also asked NVIDIA about how quad SLI was affecting the overhead the Detonator driver created and how much overall efficiency was affected by a move to four GPUs in a single system.  Logically, dividing work up between four cards instead of two introduces additional work to both the driver and the GPUs but NVIDIA says that because the dual SLI connections were built into each G71 core the additional overhead is very little. 

The law of diminishing returns is hard at work here as while you can always throw power at a problem (in this case 3D rendering) there is a sensitive line that is crossed where users don't see enough return on their additional investment to make it worth it.  Luckily for NVIDIA, this line has been steadily moving towards the more expensive cards with the 6800 Ultra, 7800 GTX and now 7900 GTX bound to be great successes. 

That doesn't mean there won't be some issues with CPU scaling of course.  If your system running on a 3500+ was hitting a CPU bottleneck with your 7800 GT, then moving to a quad SLI 7900 GTX system would not be smart without moving to at least an FX-60 processor.  Intel's Conroe parts look very promising as well and might be around right when quad SLI is coming out for purchase by the enthusiast at e-tail.  A match made in heaven?