Skip to main content

Curbing Your GPU's Power Use: Is It Worthwhile?

Benchmark Results: Cinebench R11 And PowerDirector 9

Desktop Usage: Cinebench R11

We've included average power consumption numbers, in addition to peak power consumption. There are some interesting things to note here. For one, we can see that the default values (clock speeds and voltage) for UVD mode are already optimal. We were only able to get away with a core voltage of 0.98 V for default UVD clocks, and the savings we see in return are very small. We also notice that lowering clocks without adjusting voltage does not translate to lower peak power consumption. Just look at the underclocked settings with the base voltage (1.175 V).

Running the card at DXVA clocks is good enough to offer some savings (between 3 to 9%) compared to the default stock clock. If you do have a graphics card with an adjustable voltage, there's a fairly good chance you can access additional savings (around 13 to 19%). Dropping memory clocks allows us to use a lower voltage and come up with even lower power consumption. At 500/1030 MHz and an operating voltage of 0.85 V, the Radeon HD 6970 we’re using is able to match the undervolted settings for the Radeon HD 5870.

H.264 GPU-Accelerated Encoding And Filters: PowerDirector 8

Keep in mind that the Radeon HD 5670 and 5770 are running at their UVD clocks. Both also feature hardware-accelerated decoding enabled. In contrast, hardware-accelerated decoding is only enabled when the Radeon HD 6970 is running at its default clocks (500/1375 MHz @ 1.0 V).

If you want an example of a best-case scenario, look at the H.264 and Bloom filter test results. The 6970 can finish the task in almost half of the time it takes AMD’s Radeon HD 5770 and 5670 to complete it.

The Radeon HD 6970 easily uses less power for the entire test, achieving nearly the same power levels as a Radeon HD 5770 and finishing in almost half of the time.