Performance Leap: NVIDIA GeForce 6800 Ultra

Power Consumption, Continued

This measurement is taken when Windows is sitting idle after it has finished booting and loading the drivers. Again the Radeon 9800XT is slightly more power hungry than the GeForce 6800 Ultra. The GeForce FX 5950 Ultra is much less demanding than either of these cards, though.

These values represent the maximum power draw that we measured in the different Game Tests in 3DMark 2003 at different resolutions and quality settings (FSAA, anisotropic filtering). The requirements seem to vary from game to game, but at the same time, ATi and NVIDIA cards also seem to respond to the games differently. In FarCry, for example, the cards need much less juice than in 3DMark 2003. And while the power requirements of the NVIDIA cards spike in Game Test 4, the Radeon 9800XT appetite for electricity is most pronounced in Game Test 1.

Basically, ATi’s Radeon 9800XT and NVIDIA’s GeForce FX 5950 have very similar power requirements, although the ATi card proves to be just a touch more economical. The GeForce 6800 Ultra, on the other hand, obviously enjoys taking a good swig from the power socket, but less than one might originally expect upon seeing the dual aux power connectors. In the end, the new card draws about 24 Watts more than its predecessor. Adjusted for the 69% efficiency of our power supply unit, that would make it 17 Watts more than the GeForce FX 5950, even though the new card uses more power-efficient GDDR3 memory.

We can also extrapolate the power requirements of the remaining cards from these numbers. Assuming that NVIDIA’s quoted maximum power draw of 110 Watts for the 6800 Ultra is correct. Let’s also assume that we reached that worst-case scenario during our tests. That would mean that the Radeon 9800XT has a maximum power requirement of about 91,5 Watts, while the FX 5950 needs 93,5 Watts.

Note : The numbers quoted here can not be directly compared with the numbers from our previous article . In this test, we used a different motherboard, which alone would change the results. Additionally, we had to adjust the power supply of our review system, as it was running 5% below spec, since the control panel featuring the PSU voltage regulation wasn’t attached (Antec specific issue).