Measuring Power Consumption: Let's Recap
What do these numbers tell us? First, the obvious: if you’re running an general-purpose application capable of utilizing your GPU, you aren’t going to load it down the same way you would if it were a game. This is why power measurement results in these scenarios don't yield the same elevated numbers seen in a more taxing benchmark like FurMark. As expected, they fall somewhere between idle and full load values we generally see in graphics cards reviews. But by how much? That naturally depends on the degree to which they utilize the GPU.
Here's a handy table with all the values from the various tests.
|Header Cell - Column 0||785||Radeon 2900 XT||Radeon HD 5670||Radeon HD 5770||Radeon HD 5870 1 GB||Radeon HD 5870 2 GB|
|Crysis @ 1920x1080||120||307||167||224||279||329|
|Adobe Photoshop CS4||120||193||134||154||183||205|
|Cinebench R11 OpenGL Test||132||232||175||202||220||253|
*Peak Power Consumption in Watts
Of course, these applications don't just differ in GPU utilization, but also other resources as well (CPU, RAM, and hard disk), so those factors have to be taken into account, too. However, they affirm the claim that GPUs don't consume peak power numbers every time they're used.
The results above are quite interesting. Cinebench R11 actually pushes the processor more than Crysis. We can see this from the higher base system power consumption with Cinebench (121 vs. 132 W). Factor in that higher base system power consumption and you'll see that the actual power consumed by the graphics card is still lower in Cinebench than Crysis.
If we take out the base system's power consumption, the table above ends up more like the one below. Granted, there's some overlap in these numbers, since we do not know exactly how much power the integrated graphics consumes.
|Header Cell - Column 0||Radeon 2900 XT||Radeon HD 5670||Radeon HD 5770||Radeon HD 5870 1 GB||Radeon HD 5870 2 GB|
|Crysis @ 1920x1080||187||47||104||159||229|
|Adobe Photoshop CS4||73||14||34||63||85|
|Cinebench R11 OpenGL Test||100||43||70||88||121|
*Peak Power Consumption attributed to each graphics cards in Watts
These numbers are important because they suggest the power draw we can realistically expect out of these cards when used in non-gaming/graphics applications. We now know just how much power is actually consumed, rather than just knowing it falls somewhere between idle and full load. We can also see the difference compared to the base system's power consumption. Should a graphics card with better power management comes along, we can point out the differences, and by just how much it improves.
Second, we've learned how managing power for different kinds of workloads really helps keep total power consumption under control. This can be seen from the PowerDVD results with the Radeon HD 5670. During H.264 playback, the Radeon HD 5670 hardly consumes more power than the integrated Radeon HD 3300, thanks to both processors’ UVD blocks. That's very efficient power management, which is even remarkable if we compare both chips side-by-side.
Another example can be seen in the Radeon HD 5870 1 GB. Running at UVD clocks (instead of full–speed) in PowerDirector allows the board to consume only 15 to 21 W more than the Radeon HD 5770. If it were to run at full speed, power consumption would be much higher.
Third, we were able to measure total power consumption. This kind of measurement can be done with similar applications, where the GPU is used to run a task completed in a certain amount of time. It’s useful in determining which graphics card offers the best performance and power consumption in that specific application. In this case, we were able to use PowerDirector. After testing, only 2 to 3 Wh separates the four Radeon HD 5000–series boards. That's quite an engineering feat.
The important element here is balance--a balance between performance and power consumption. The additional performance from the faster cards helps them finish the task and return to idle mode quickly. Overall consumption is still low enough so that we don't end up using significantly more power. It's doubtful we could possibly be aware of this just by looking at idle and full load power consumption numbers. In fact, this is the same type of comparison we like to make when looking at processors, but we’ve never really applied the same principles to graphics.
In more traditional graphics workloads, that same performance headroom benefits gamers who want higher frame rates or the ability to use slightly higher resolutions. Undoubtedly, you will be able to play Crysis at higher frame rates with the Radeon HD 5770 compared to the HD 5670. The same also applies to AMD’s Radeon HD 5870, which allows you to enjoy more demanding settings than the HD 5770.
Oh, wait, this just in:
My next PC will be used mostly for movie DVDs and Diablo 3. Apparently if I get a 5870 1GB I get the best of both worlds - speed in Diablo and low power consumption when playing movies.
How about nVidia cards, would I get the same behavior with a GTX 480 for example?
Next questions: First, where does the HD5750 fall in this? Second, if you do the same kinds of manual tweaking for power saving that you did in your Cool-n-Quiet analysis, how will that change the results? And finally, if you run a F@H client, what does that do to "idle" scores, when the GPU is actually quite busy processing a work unit?
I'd love to see nvidia cards and beefier CPUs used as well. Normal non green hdds too. Just how big of a difference in speed/power do they make?
Thank you for sharing.
Thanks for reading the article.
Have no 5750 sample yet, but they should relatively be close to 5770. For this article, we simply chose the best bin for each series (Redwood, Juniper and Cypress).
The second question, what will happen when you tweak the chip? Glad you ask!! I can't say much yet, but you'll be surprised what the 5870 1 GB can do.
As for NVIDIA cards, I'm hoping to have the chance to test GF100 and derivatives very soon.