Source: Tom's Hardware – Keywords: nvidia, hybrid, sli
Categories: Graphics
Power, Performance, and Efficiency
The big thing that separates the nForce 780a SLI from lower-cost GeForce 8000-series chipsets is its nForce 200 PCI-Express 2.0 bridge, which adds support for two high-end cards at full bandwidth to what might otherwise have been a mundane product. This feature lends itself to the ultimate in gaming system efficiency by shutting off the unneeded cards when the user exits a game, but our motherboard died before we could test this feature on a 9800GX2.
Everyone knows the 9800GX2 is a "power hog" that really deserves such energy-saving technology, but any discussion of efficiency must also consider the amount of work being done by that energy.
We might not have any HybridPower data, but we can at least consider the performance gained by the lesser cards in GeForce Boost mode before making any judgments about the 9800GX2’s basic efficiency.

The low-cost 8400GS and 8500GT got their greatest GeForce Boost in Supreme Commander, but both cards lost some performance in Prey with this technology enabled.
No performance gains are free, and the onboard graphics processor certainly requires some power to function. We tested each configuration in both 2D and 3D mode, using the HD Rendering 1 benchmark (Canyon Frow_even) from 3D Mark 2006 to find the maximum requirement. The following chart shows "global" power consumption for the entire system.

The onboard graphics controller added only a few watts to each low-end configuration, while the 9800GX2’s extreme power consumption leaves us devastated over the loss of our test motherboard prior to HybridPower evaluation.
It’s still important to remember that efficiency is a ratio of work output to energy input, and the 9800GX2 is an impressively powerful card. At the other end of the market, the 780a’s onboard graphics processor’s scant performance is compensated by its miniscule power consumption. Our motherboard failure may have prevented us from answering any questions today about HybridPower on the 9800GX2, but is the other side of Hybrid SLI - known as GeForce Boost - also a boon to efficiency?

It may seem hard to believe, but the 9800GX2 is actually the most efficient card in our evaluation, thanks to its enormous performance advantage over low-cost cards. GeForce Boost also does well, providing a performance boost that surpasses its power consumption.
Conclusion
Today we considered the GeForce Boost capabilities of NVidia’s newest performance chipset, but it’s really a shame our test board didn’t live long enough to provide any answers about the efficiency of HybridPower. Next week we’ll see how the 780a SLI stacks up against AMD’s 790FX in general performance.
- Previous page Benchmark Results, Continued
- How To Overclock Your Graphics Card
- PCI Express 2.0 Graphics Cards: How Much Extra Performance Do They...
- Nvidia GeForce 9800 GTX Review
- Nvidia GeForce 9800 GX2 Review
- The Best Gaming Graphics cards for Your Money: March 2008
- Nvidia's GeForce 9600 GT Reviewed
- Fresh from Canada - ATI's Radeon HD 3450 and HD 3650
- The Best Gaming Graphics Cards for Your Money: February 2008
- ATI R680: the Rage Fury MAXX 2?
- Crossfire meets PCI Express 2.0 – More Lanes, More Frames?
Can someone answer me the question. Why is onboard graphics so much better at preserving energy per performance amount than a normal graphics card.
You can't play most games at 100x100 on an onboard chip so there is nothing to use any energy.
You thrown in a 3d Card which has 1000000x more transistors and you have to power them up to get the game moving. I hope the next war for gfx card speed will be fps per watt. 1watt card running at 60fsp will do me.
My PC broke last year and there was a dip in my bill for the month that I didn't play games. Thats a worry for me. Soon it wont be the cost of the game; but how much I have to play to power my rig per level.
will you be recieving a new motherboard to run the tests again with the 9800GX2??
That last graph (performance per watt) is completely flawed because it doesn't take into account the energy used by the PC. Unless overall system performance per watt is what you wanted, in which case it gets a lot more complicated...
btw looking at the impact on power consumption of limiting performance via v-sync (or another mechanism) would be interesting for a future article on power usage.