NVIDIA SLI performance scaling - GeForce 8800 GT 512MB
If 2007 was the year of DirectX 10, then 2008 is looking like it may well be the year of multi-GPU rendering. Sure, using more than one graphics board to accelerate 3D rendering is nothing new by a long stretch, and we've seen a handful of attempts to bring such technology together on a single board from both ATI and NVIDIA in recent years, but this time around there's a feeling amongst many that we'll be seeing a concerted, full-on attempt to make multi-GPU rendering a big part of the future of discrete graphics.
With GPUs fast approaching one billion transistors, and manufacturing processes quickly being pushed to their limits, this makes a certain amount of sense - why create one massive, expensive, hot GPU with low yields when you can tie together two (or more) smaller, more efficient and cheaper GPUs to do the same job? Intel and AMD jumped on this bandwagon with CPUs, adding cores over clock speed, so it was only a matter of time before the graphics industry went the same way.
Of course, GPU manufacturers face some rather difficult challenges ahead if they really want to bring us into a full on multi-GPU future, but that's a discussion for another day. All we do know is that with ATI's Radeon HD 3870 X2 (sporting two RV670 cores) released this week, NVIDIA's dual G92 GeForce 9800 GX2 on the way, and rumours that ATI's next-generation R700 part will in fact be two RV770 cores on a single board, multi-GPU rendering in some shape or form is here to stay.
Thus (and with our Radeon HD 3870 X2 review coming very soon), what better time is there to take stock of the current state of multi-GPU rendering? To see just how well two graphics boards scale in current game titles on the NVIDIA side of the fence, we've taken two hugely popular GeForce 8800 GT 512MB parts, an nForce 780i SLI motherboard, and put this configuration through its paces to see what it can offer. Interested? Of course you are. So let's get testing!