NVIDIA unveils Hybrid SLI, 780a chipset at CES 2008

NVIDIA used its CES platform to offer some information on the company's upcoming products, but the announcements made fell short of generating any sort of "wow" factor. That's not necessarily a bad thing—small, incremental, but ultimately unexciting product improvements are just as important as next-generation bolts of lightning—but anyone hoping for news of an ATI killer or dominant new chipset ended up disappointed.

Related Stories

Hybrid SLI
NVIDIA spent a good deal of time talking up what the company is callingHybrid SLI. The term refers to a set of two technologies, GeForce Boostand Hybrid Power, both of which must be present in order to earn theHybrid SLI moniker. GeForce Boost is a performance-enhancing optionthat melds an integrated GPU with a discrete GPU. Once combined in thismanner, the two GPUs render different frames, with the results beingpassed to the integrated GPU for display.
The real-world performance gain from using GeForce Boost is unknown.NVIDIA claims a 40 percent improvement in 3DMark 2006 when GeForceBoost is enabled, but gives no information on which integrated anddiscrete solutions were combined to achieve that number. It's hard totell how the system's increased 3DMark score correlates to other games,but NVIDIA was kind enough to include the additional bar graphquantitatively demonstrating the performance benefits of GeForce Boost(and no, we didn't crop out any percentages).

Hybrid Power, meanwhile, is aimed at users interested in minimizingpower draw without sacrificing video performance. A HybridPower-compatible motherboard and video card can shut down a discretevideo card when it's not needed, but power it up again and use it intandem with the integrated solution when a game is launched.
The upcoming NVIDIA 780a SLI, 750a SLI, and uATX GeForce 8200 chipsetswill all be compatible with Hybrid SLI. All future NVIDIA video cardswill support Hybrid SLI as well, but two current models—the 8400GS and8500GT—will support GeForce Boost when paired with an appropriatemotherboard. As for availability, the new boards and chipsets will beavailable in the first quarter of 2008 for AMD CPUs, with Intelavailability to follow in the second quarter.
Upcoming chipsets and HD playback support
NVIDIA's upcoming 780a SLI chipset is closely related to the Intel 780ichipset that the company launched last December. There are somesignificant differences between the two, however, as shown by theirrespective chipset diagrams below. The Intel 780i is first, followed bythe AMD 780a.

The 780a appears to be superior to the 780i in several respects.Both chipsets use the nForce 200 crossbar—it's still unclear why NVIDIAhasn't integrated this functionality into the northbridge the way ATIand Intel have—but the distribution of PCIe connections and the totalbandwidth allotment is significantly different. The 780i SLI offers twoPCIe 2.0 slots off the nForce 200, but the third slot is doublyhandicapped by its PCIe 1.0 limitation and its attachment to the 780isouthbridge. It's far from an ideal arrangement, and it leaves thechipset's "Tri-SLI" support looking more like a marketing bullet pointthan an actual feature.
The 780a, on the other hand, attaches all three SLI slots directly tothe nForce 200 chip and splits the bandwidth between them in anx16/x8/x8 configuration. NVIDIA also claims that the nForce 200 chip iscapable of performing a point-to-point write from one GPU to anotherwithout touching the CPU. When the CPU does need to feed data to theGPU(s), the NF200 processor can take a single input from the CPU andwrite it to all connected GPU's simultaneously. NVIDIA has yet toclarify if the nForce 200 crossbar on the 780i is also capable of thesetransactions.
Finally, the 780a includes an integrated GPU. That's not a typo—NVIDIA is serious enough about the Hybrid SLI concept that it intends to offer the solution on both its 780a and 750a chipsets.
GeForce 8200
The exact status of the GeForce 8200 is a bit puzzling. Ostensibly,it's NVIDIA's next-generation IGP with full DX10 support and full videodecode hardware for the H.264, VC-1, and MPEG-2 code. The actualGeForce 8200 term, however, is referred to differently in differentplaces. In some places, NVIDIA refers to it as a "motherboard GPU" andthe lowest price GeForce DX10. As you can see from the image below,however, the GeForce 8200 is also a chipset despite its lack of annForce designation.

NVIDIA has yet to state whether or not the GeForce 8200 IGP is the same IGP solution it intends to include in its 780a and 750a chipsets, or what relation it will have to the IGPs built into those motherboards.
Conclusion
NVIDIA's CES presentation was rather disappointing. The company'sHybrid SLI announcement isn't a surprise, given ATI's Hyper Crossfirepreview last month. The two implementations may not compete in exactlythe same area—NVIDIA's Hyper SLI appears to have a broader focus thanATI's Hyper Crossfire—but they'll probably end up looking an awful lotalike.
Similarly, the 780a chipset is a definite improvement over the 780i,but NVIDIA failed to mention whether its new chipsets would offer lowerpower consumption than the nForce 5xx and 6xx series. It would've alsobeen nice to see a south bridge attached to the 780a diagram, giventhat the 780i's south bridge is nearly two years old. It's also not allthat encouraging to see NVIDIA's chipset business so closely hitched toAMD, given that company's precarious financial condition.
As for video cards, NVIDIA had little to say at its presentation, but we may find something out later in the week. Stay tuned.
To be fair, however, NVIDIA's presentation meshes with the rest of whatArs has seen at CES thus far. Updates, incremental progress, and smallsteps appear to be the order of the day. Plenty of progress, in otherwords, but not much sparkle.