Power Consumption

While we've already covered the card's incredibly robust power circuit, we haven't really talked about one of AMD's major design goals with Cypress: to reduce the idle power consumption to a minimum.

The RV670 made some huge strides forward in idle power consumption with the introduction of ATI PowerPlay technology on the desktop for the first time. We saw idle power consumption for the GPU drop to around 25W, which is about as little power as any relatively speedy GPU sips when it's not doing anything.

Unfortunately, the technology ended up getting lost in the post somewhere between RV670 and RV770, because the latter's idle power consumption was pretty horrific compared to what Nvidia had achieved with its GT200 GPU. Nvidia had rolled out its own power saving technology which had various modes, including one for Blu-ray playback, to ensure power consumption was kept under control when the user didn't require all of the horsepower available.

As a result, the Radeon HD 4870 was a card that never really cooled down, because it never truly powered itself down properly. Idle power was close to 100W on RV770 and AMD claims that Cypress's idle power is less than a third of that at just 27W.


Yup, idle power consumption has improved pretty significantly - a 150W idle power draw (for the whole PC, remember) is nothing to be sniffed at on a system with as much power as the one we're using for testing these cards. For reference purposes, our system was the same as what we use for all of our graphics card testing and is as follows:

• Intel Core i7 965 processor (3.2GHz: 133MHz x 24)
  
• Asus P6T V2 motherboard (Intel X58 Express)
  
• 6GB Corsair TR3X6G1333C9 DDR3 memory (3x 2GB, 1,333MHz, CL9)
  
• Corsair X128 SSD
  
• Corsair HX1000W PSU
  
• Windows Vista Home Premium x86-64 (with Service Pack 2)
  
• Antec Twelve Hundred Chassis
  
• HD 5870: Catalyst 8.66 RC6 Windows Vista/7 (September 11th Build)
  
• All other ATI: Catalyst 9.9 WHQL
  
• Nvidia: ForceWare 190.38 WHQL

Testing the power consumption with the Canyon Flight test of 3DMark06 was an interesting conundrum, because it didn't really tell the whole story. We saw higher power consumption from the Radeon HD 5870 in other tests we ran, but this only shows that power consumption varies depending on what you’re doing – possibly other tests loaded the cores more, or thrashed the memory.


Looking at the power meter during some other tests, the Radeon HD 5870 managed a peak power consumption of 306W in 3DMark Vantage's Perlin Noise test, for example. However, running a game-like scene, the HD 5870 consumed as much power as the HD 4890, which is pretty good considering that the HD 5870 is twice the GPU of the HD 4890.