Articles: Graphics

Bookmark and Share

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 ]

Power Consumption, Temperature, Noise, Overclocking

It is the first Radeon HD 6950 product with 1 GB of video memory in our lab, so we decided to check out its power consumption as well. We used the following testbed: 

  • Intel Core 2 Quad Q6600 CPU (3GHz, 1333 MHz FSB x 9, LGA775)
  • DFI LANParty UT ICFX3200-T2R/G mainboard (ATI CrossFire Xpress 3200 chipset)
  • PC2-1066 SDRAM (2x2 GB, 1066MHz)
  • Enermax Liberty ELT620AWT PSU (620W)
  • Microsoft Windows 7 Ultimate 7 64-bit
  • CyberLink PowerDVD 9 Ultra/"Serenity" BD (1080p VC-1, 20 Mbit)
  • Crysis Warhead
  • OCCT Perestroika 3.1.0

The new testbed for measuring electric characteristics of graphics cards uses a card designed by one of our engineers, Oleg Artamonov, and described in his article called PC Power Consumption: How Many Watts Do We Need?. As usual, we used the following benchmarks to load the graphics accelerators:

  • CyberLink PowerDVD 9: FullScreen, hardware acceleration enabled
  • Crysis Warhead: 1600x1200, FSAA 4x, DirectX 10/Enthusiast, "frost" map
  • OCCT Perestroika GPU: 1600x1200, FullScreen, Shader Complexity 8

Except for the maximum load simulation with OCCT, we measured power consumption in each mode for 60 seconds. We limited the run time of OCCT: GPU to 10 seconds to avoid overloading the graphics card's power circuitry. Here are the obtained results:

The power consumption measurements do not tell us anything new. The PowerColor HD 6950 1GB needs somewhat more power than the reference Radeon HD 6950 2GB in 2D mode and when decoding HD video. When it comes to 3D applications, its power draw is somewhat lower. 161 watts is comparable to the power consumption of the GeForce GTX 560 Ti, so the PowerColor card is quite competitive in its class in this respect.

Take note that in the video playback mode the numbers reflect the peak power consumption as is clear from the diagram above whereas the average is about 40 watts. The second power connector (it is closer to the PCB edge and is labeled “12V 6-pin” in the table) has a higher load than the first one, but even OCCT cannot load those connectors much higher than their recommended limit of 75 watts.

Now let’s check out the performance of the card’s cooler. We've got the following results at an ambient temperature of 25°C:

The results are far from extraordinary and do not actually differ much from those of the PowerColor HD6870 PCS+ but the HD6950 1GB has higher heat dissipation. Besides, the cooling system of the PowerColor HD6950 1GB looks very good in comparison with the reference card's cooler.

The PowerColor HD6870 PCS+ would produce a lot of noise in our tests. Is the new card from PowerColor any better in this respect?

Oddly enough, this depends on the position of the BIOS switch. When you use the read-only factory BIOS, the noise level is within reasonable limits. At a background noise of 38 dBA, the PowerColor HD6950 1GB is as good as the reference Radeon HD 6950 2GB in terms of noisiness in 3D applications. But when the switch is set into the other position, the fans accelerate and get much louder. This looks like some protection against overheat during overclocking experiments. The high level of noise can hardly keep off true overclockers, so we don't mind the manufacturer's cautious approach. Of course, we ran the card in the b1 mode (i.e. with the factory BIOS) in our tests, including the temperature tests.

The PowerColor HD6950 1GB didn’t do particularly well in our overclocking tests. We managed to make it stable at the following clock rates:

The GPU frequency of 860 MHz is hardly a big success but we achieved a rather high memory frequency of 5400 MHz. We are going to benchmark our PowerColor HD6950 1GB both at its default clock rates and at the frequencies we achieved in our overclocking attempt.

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 ]


Comments currently: 9
Discussion started: 04/01/11 02:23:20 PM
Latest comment: 04/09/11 05:30:33 AM

View comments

Add your Comment