Shutting the card (or cards) down is handled by the SMBus channel (which is a part of the PCI-Express specification) and the discrete graphics card has to be HybridPower-enabled in order to take advantage of this functionality. The reason for this is that it requires additional PLLs and power circuitry to allow the GPU to be turned off.
Since you’re turning off the discrete graphics card (or cards) with the provision of them eventually being switched back on when it’s needed the contents of its frame buffer must be copied across the PCI-Express bus into the IGP’s frame buffer (i.e. local system memory).
However, even when you’re running in performance mode—with the discrete GPU enabled, but with the display still plugged into the motherboard—the contents of the frame buffer need to be transferred to the front buffer, which in this case is located in main system memory. Nvidia said that PCI-Express 2.0 provides more than enough bandwidth for both of these processes and it also says that latency isn’t an issue either.
Despite Nvidia’s reassurance, I’m still a little concerned about the performance implications here because you’ve got a situation where a graphics card with massive
amounts of local memory bandwidth is having to send its data to another, much slower, memory partition before it’s displayed on screen. In the worst case, the result could be some rather annoying input lag and that would certainly put a lot of gamers off using this technology.
It’s worth remembering that being able to turn off your power hungry graphics cards doesn’t just reduce your power bill, it should also quieten your system down quite a bit too because the fans on the cards that have been put into a sleep mode won’t require any active cooling. Instead, your system will be as quiet as a typical mainstream PC when it’s running HybridPower mode.
Click to enlarge
Nvidia talked a lot about the benefits of HybridPower and they’re at their most profound at the high-end – after all, they’re the GPUs that use the most power and it’s fair to say that Nvidia’s power saving technologies in its desktop GPUs aren’t quite as aggressive as those recently introduced into AMD’s competing products. The high-end is where the focus is going to be for now on HybridPower, but that doesn’t mean Nvidia won’t bring the technology to more affordable graphics cards in the future as well – yes, the power savings won’t be as profound, but they are just as important in my opinion.
We’ve spoken to several reliable sources about HybridPower in order to try and gather some information on when we can expect to see HybridPower enabled products – at least, with a more specific timeframe than the vague Q1 release schedule Nvidia has indicated officially. None of them seemed to know exactly when we can expect to see product – I’m guessing it will be Nvidia’s next high-end product
and I’m not even sure that it’ll be supported right out of the gate, because there’s a lot of work to be done on drivers – and from what I have been told, the driver development is incredibly complex.
Because of this, Hybrid SLI is still quite a way from its official launch and the live demo system we’ve seen was described as ‘alpha’. And while I’m talking about the live demo system used to show the power saving benefits of HybridPower technology in passing, it’s worth mentioning that the card installed was recognised as a GeForce 8800 GT by both the operating system and the RTHDRIBL demo that was used to create 3D load. Nvidia described this as an unreleased graphics card launching in the future, so we’ll assume that whatever the card was is based on the G92 graphics chip and was masked as a GeForce 8800 GT by the driver.
Now, when I asked Nvidia about support on the GeForce 8800 GT and GeForce 8800 GTS 512—existing products based on the G92 graphics chip—the company said that current cards don’t have the required PLLs to enable HybridPower. That doesn’t mean that they couldn’t theoretically support the technology in the future after some slight design modifications though.