Defining Energy Efficiency

Over the last year, we've been inundated with tech products branded as ECO-this and Green-that. While we've always been interested in energy consumption as an indicator of aspects of technical design and performance, we're far from convinced that you can actually save significant cash by buying a-n-other company's ‘power saving’ product.

In this article, we're going to be busting some myths about computers and power consumption - both our own assumptions and the superfluous claims made by manufacturers.

We're going to measure the real world differences between actual performance PCs. By measuring the power at the wall socket we have a direct readout of the immediate benefit provided by changing one component to a lower-power alternative, but it's important to combine this with the time it takes to complete a computational task - a CPU that's 1 per cent more power efficient when video encoding but that takes 50 per cent longer than its competitor isn't going to be beneficial. From these two variables (power use and time taken), we can work out the product's efficiency as a factor of power use over time. For example, if a particular product uses 50 per cent more power, but it's also twice as fast (100 per cent) at completing the task at hand, then this means its 33 per cent more efficient.

Letting a represent speed and b power
a/b = E
2A/1.5B = 1.33 E

CPUs - AMD

Test Setup

• Gigabyte GA-790FX-UD5 (790FX/SB750, F7 BIOS)
  Common Components
  
• 4GB OCZ DDR3 PC3-15000 / 1,866MHz / Platinum Edition @ SPD: 1,333MHz, CL7
  
• 650W Seasonic X-Series PSU
  
• 1TB Seagate Barracuda 7200.12 HDD
  
• ATI Radeon HD 5770 1GB
  
• ATI Catalyst 10.1 WHQL
  
• Windows 7 Home Premium x64

We tested the CPUs in the standard systems above with our complete Custom PC benchmark to get an average, overall result, and we compared this to their full load power use in order to get an efficiency factor. The Custom PC benchmark comprises three tests - editing a series of photographs in GIMP, encoding an MPEG-2 video using H.264, and a multi-tasking test where the H.264 video is played while 7-Zip is used to compress a large number of files in the background. The tasks are always the same amount of overall work, but obviously different CPUs take a different amount of time to get through them. Achieving a higher score (completing the set tasks in faster times) at a lower power use means the result is better for any particular CPU. We rounded up a batch of processors, some within the same TDP rating, to measure the perceived effect of this TDP factor and whether buying lower power CPUs simply means a slow processor, or actually one that is more efficient.



We would have to say we're quite surprised at some of the results. Despite the middle-of-the-rad 2.5GHz clock speed of the 905e, its efficiency factor is actually one of the highest at 8.06. The performance of the 905e, the latest quad-core Athlon 635 and our favourite Phenom II X3 720 BE all generate around the same result, however the 905e requires several watts less than the triple-core and they both use 20-25W less than the Athlon II X4! To be honest, this just goes to show, again, how much of a false economy the Athlon II X4 parts are: the 635 is the same price as the Phenom II X3 720 Black Edition, yet, it uses more power despite the fact the Phenom has 6MB of L3 cache it cannot power down. The Phenom obviously has the added advantage it can be potentially 4th core unlocked and overclocked via multiplier only too, increasing its performance considerably.

It's also worth noting that despite the fact AMD rates the Phenom II X3 720 BE as a '95W' TDP part and the Phenom II X4 905e as a '65W' TDP part, there's only a 5W difference under load and nothing between them at idle.

Despite the fact AMD's kingpin Phenom II X4 965 BE is clearly the fastest by a large margin, the fact it also uses 40-50-60W more than the other triple and quad core products drops its efficiency factor right down. At the other end of the scale, despite the fact that AMD's Athlon II X2 240e is over 40 per cent slower than its premium Phenom II, it requires just half the power, rendering its efficiency factor much higher than everyone else. Well done AMD for putting this together into a £60 part.