Unfortunately, we could not carry out our power measurements tests for technical reasons, but we did measure the temperature of our multi-GPU configurations. To put them under load we launched the test from Aliens vs. Predator (2010) and ran it in five cycles at 2560x1600 with the highest graphics quality settings and 16x anisotropic filtering but without FSAA. It is at such settings that the GPUs reached their highest temperatures.
This test was carried out with the graphics cards installed into a closed system case. Here are photographs of them, followed by their results:
Quite expectedly, each configuration is very hot. The GPUs of the Radeon HD 6970s were as hot as 94 and 90°C, their fans rotating at 3750 and 2660 RPM (the numbers refer to the top and bottom cards, respectively). The GPUs of the GeForce GTX 570s were 90 and 87°C hot, their fans working at 3270 and 2790 RPM. As the numbers suggest, both configurations are very hot and very noisy. Temperature sensors of different cards and GPUs are calibrated in different ways, but we can say that our GeForce GTX 570 tandem seemed to be less noisy than the Radeon HD 6970 pair. Anyway, we guess the best option for such configurations is a liquid cooling system with full-cover water blocks.
But do the cards differ much in terms of their GPU temperature when they are working alone rather than in a multi-GPU tandem?
The single Radeon HD 6970 was as hot as 90°C but its fan was working at 2470 RPM only. The GeForce GTX 570 had a GPU temperature of 87°C and a fan speed of 2610 RPM. So, the temperatures are the same as when the cards work in multi-GPU tandems but the noise level is lower.