Sli != Sli
Wrong wrong wrong.
Voodoo SLI = Scan Line Interleave - this allowed the Voodoo to render up to 1024x768 on a hardware platform that could only do 800x600, and gained a bit of performance to boot. Yes, with 3Dfx, timing was painfully critical. Voodoo's SLI ran each card at 1024x384 and performed hardware scan-line interleaving.
nVidia SLI = Scalable Link Interface - this offers two rendering and one anti-aliasing method for splitting the work between the video cards; Split Frame Rendering, Alternate Frame Rendering and SLI Antialiasing. Timing is not critical, but if the GPU's aren't matched well, performance will worsen.
Hybrid SLI is designed specifically to work with unmatched GPU's. It is broken into two different modes: Hybrid Power - this will switch between integrated and discrete GPUs on the fly as demand changes, and GeForce Boost - this will use BOTH GPUs to improve the 3D performance of games at the sacrifice of battery life.
Why does nVidia's site say only the 9100, 9300 and 9500 support GeForce Boost? All GeForce 8-series GPU's and higher technically support all available forms of SLI, since they're all software schemes. Therefore, pairing a 9800 with a 9100 would not improve frame rate, so nVidia limits this mode to chips they know would improve frame rates.
However, the 9400M is brand new, and I suspect that nVidia has yet to update all their webpages. It is not a slightly faster bump of the 9100M, and should be able to run well along-side the 9600M. On both the pages specifically on the 9400M and 9600M, both clearly indicate that they are Hybrid SLI enabled (and in detail, both say they are Hybrid Power and GeForce Boost enabled).
So, this is only up to Apple to include support for this at the kernel level.
Great post amongst the madness. Implementing Hybrid SLI at this point with Snow Leopard around the corner would not have been wise. It would be a huge undertaking. Consider the syncronization nightmare that is involved between a discrete GPU with its own memory and that of an onboard GPU, which DEPENDS on main memory. If main memory bottlenecks during intensive processing and slows down the 9400, the frames processed by the 9600 would have to wait or the 9600 would have to go back and fill in what the 9400 couldn't process in time. SLI with 2 identical cards with their own memory allows for much easier syncing...None of these issues have to thought of. I think people are really underestimating the work that is needed to do Hybrid SLI. Frankly it's a feat the technology exists period. If you really think about, the overhead involved in syncing one card with another that's 2.5x faster may not give you a boost worth the development at the 9600m level. Even IF the boost could warrant the development (which costs $$$), could you blame Apple for not putting the time into an OS that is going to be replaced next summer? Can you imagine the possible bugs that could occur that they would have to field and fix? I wish Hybrid worked too, but cut Apple a little slack here.
