Power Consumption, Temperature, Noise and Overclockability
To check out the electrical parameters of our GeForce GTX 480 and GeForce GTX 470 we are going to use our standard 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?. This device facilitates and automates the measurement process. 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 limit the run time of OCCT: GPU to 10 seconds to avoid overloading the graphics card's power circuitry. Here is what we managed to obtain using this testing methodology:
A chip with over 3 billion transistors, even though manufactured on 40nm technology, cannot be expected to be economical. However, the developers of the GF100 did their best to keep its power consumption within reasonable limits and succeeded. The peak power draw of the senior model of the new GeForce series is much lower in 3D applications compared to that of the GeForce GTX 295 (but somewhat higher compared to the Radeon HD 5970).
When displaying the Windows Desktop, the flagship products of both companies compete as equals. It’s unclear which is better at playing HD video, too. Take a look at the graphs: the GeForce GTX 480 seems hungry at first, but 20-25 seconds into the test file the driver switches the card into power-saving mode, reducing the power consumption on each line, save for the loaded +3.3V one. As a result, the average power draw of the GeForce GTX 480 is that mode is 50-60 watts. Quite acceptable for its class. The Radeon HD 5970 behaves in a similar manner, dropping its power draw from 100 to 50-55 watts about 10 seconds into the clip. So, the products from AMD and Nvidia are equal in this test.
As for the distribution of load, the 8-pin PCIe 2.0 connector is expectedly the most burdened one. Its peak load is nearly as high as 13 amperes. At 12 volts, this means about 150 watts. The peak current in the line going through the 6-pin PCIe 1.0 connector is not higher than 6.2 amperes, so the load is not higher than 75 watts. These numbers fit within the permissible limits specified for these types of PCIe power connectors. The power section of the PCIe x16 slot is not higher than 41 watts under any load.
The junior GF100-based model does not look economical in comparison with the Radeon HD 5870. It has a peak power draw of nearly 200 watts as compared to the AMD solution’s 170 watts. Alas, this is the other side of the tremendous complexity of the new GPU developed by Nvidia. The Radeon HD 5870 has only half the power consumption of the GeForce GTX 470 in idle mode but the latter looks better when playing Blu-ray thanks to the same method as the GeForce GTX 480 uses. The card consumes a lot of power at first, but as soon as the driver realizes that it is dealing with HD video rather than a 3D game, the average power draw lowers to 25-40 watts, which is somewhat better than the result of the Radeon HD 5870.
The distribution of load among the power lines of the GeForce GTX 470 is not as interesting as with the GeForce GTX 480 as its connectors are both of the 6-pin variety and are loaded equally in every mode. The maximum current is 8.3 amperes, which is about 100 watts at 12 V, but we know that the 75W limit specified for the PCIe 1.0 connector is rather a recommendation. We did not notice the connectors get too hot or burn out during our tests.
Summing everything up, the new solutions from Nvidia are good enough in terms of power consumption, even though they are somewhat inferior to the Radeon HD 5800 series in this respect. Given the complexity of the GF100 chip, this had to be expected. After all, we are talking about premium-class products capable of competing with the RV870 Cypress based solutions.
Of course, the previous generation of graphics cards from Nvidia, based on the 55nm G200b or on the older 65nm G200, cannot match the newcomers, just as expected. So, if your gaming platform is equipped with a GeForce GTX 285, GeForce GTX 295 or Radeon HD 4890, its power supply will suffice for a GeForce GTX 480 or GTX 470. And if you are building your system from scratch and want to use one of these cards, you should choose your power supply carefully. This advice is true for every top-end graphics card.
As for the temperature factor, the GPU of the senior model of the new GeForce series can be as hot as 90°C and more but the Radeon HD 5970 is just slightly cooler. The cooler of the Nvidia card works more effectively in idle mode although we must keep it mind that the Radeon HD 5970 has two rather than one GPU. The new card is really hot at work. After running some 3D applications, it may scorch your fingers if you touch the metallic plate above the heatsink. You should wait a while for the card to cool down before uninstalling it from your computer.
The GeForce GTX 470 is not exceptional in terms of temperature, either. It is about as hot as the GeForce GTX 285, but that’s an achievement considering its performance and in comparison with the Radeon HD 5870 and 5850. The plastic casing of the cooler on the GTX 470 won’t allow you to scorch your fingers, whatever the card’s temperature may be.
At an ambient noise of 37 dBA, the GeForce GTX 400 series cards are both quiet in 2D mode, but not exactly silent. When under load, the senior model quickly accelerates its fan to 70% and you can only then stifle its noise by stepping up the volume in your speakers or headphones. When the load is removed, the graphics card quickly slows down to 44% fan speed and its noise becomes normal.
The GeForce GTX 470 dissipates less heat and is less noisy. It is very quiet in 2D mode and rather quiet in 3D mode. The fan did not speed up more than 60% during our tests. These results are just what you could expect from products of that class. Top-end gaming cards just cannot be silent unless you use a liquid cooling system. By the way, some makers of such coolers, e.g. Koolance and Danger Den, have already announced water blocks designed for the GeForce GTX 480/470. EVGA and Inno3D offer such cards with the water-blocks already installed.
We had not expected much from the Gigabyte GeForce GTX 480 in terms of overclockability but the card managed to surprise us.
We could increase the main GPU frequency by as much as 13.5%, from 701 to 795 MHz, and the shader domain frequency grew up proportionally from 1401 to 1590 MHz. The memory chips were stable at 975 (3900) MHz, i.e. 5.5% above the default frequency. This must be due to the 384-bit memory bus. We guess this overclocking result is quite substantial, so we will benchmark our Gigabyte GTX 480 at these increased frequencies as well.
Overclocking the Gigabyte GeForce GTX 470 was even more rewarding.
We achieved a GPU frequency growth of 19.3%, from 608/1215 MHz to 725/1450 MHz but the memory chips could only be overclocked by 3.1% or to 863 (3452) MHz. Anyway, this is a substantial frequency growth and we will benchmark the card at the overclocked frequencies, too. We will see shortly what performance benefits this overclocking provides.