Articles: Graphics

Bookmark and Share

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

Power Consumption Measurements: Testbed and Methods

In our reviews we use a digital multimeter Velleman DVM850BL (0.5% accuracy) to measure the current on different power lines. The testbed is configured like follows:

  • Intel Pentium 4 560 CPU (3.60GHz, 1MB L2)
  • Intel Desktop Board D925XCV (modified)
  • PC-4300 DDR2 SDRAM (2x512MB)
  • Samsung SpinPoint SP1213C HDD (Serial ATA-150, 8MB buffer)
  • Microsoft Windows XP Pro SP2, DirectX 9.0c.
  • ATI Catalyst 6.6, Nvidia ForceWare 84.21
  • If not mentioned specifically, we used reference graphics cards from ATI Technologies and Nvidia. For your convenience we also provide the links to corresponding detailed reviews of the mentioned products.

We use the same methodology since 2004 (for details see our article called Power Consumption of Contemporary Graphics Accelerators. Part II: NVIDIA vs. ATI) and think it has proved its worth: we measure the current in shunts that we connect into the +3.3V and +12V power lines of the mainboard’s graphics slot. Similar shunts are also connected into the alternate power connector (the additional power connector on the graphics card’s PCB).

According to the PCI Express x16 1.0a specification, up to 72 watts of power can be transferred through the +12V lines of the mainboard as well as through the additional power connector, i.e. 144 watts in total. Additionally, some power can be consumed from the mainboard’s +3.3V line (this line is used to simplify power conversion for some chips on the graphics card). It means that a modern graphics card cannot have a power draw of over 150W, at least mainboard and PSU manufacturers make their products basing on that supposition.

Some of our readers inquired after the distribution of power consumption among the different lines and we do offer this information for some of the tested devices. These numbers aren’t, however, important for certain products, so you should excuse us for not giving you this information for every tested graphics card.

We created the Peak 3D load by launching the first SM3.0 graphics test from 3DMark06 and running it in a loop at 1600x1200 with enabled 4x full-screen antialiasing and 16x anisotropic filtering. The Peak 2D load was created by means of the 2D Transparent Windows test from Futuremark PCMark05. In the Idle mode the graphics card was displaying the Windows XP Desktop.

Noise Level: Testbed and Methods

The level of noise produced by graphics cards is measured in our labs on the computers that we use as the testbeds for benchmarking the same cards.

  • AMD Athlon 64 FX-60 CPU with Zalman CNPS9500 cooler working at its minimum rotation speed.
  • Asus A8N32-SLI Deluxe mainboard (nForce4 SLI X16) for Nvidia GeForce graphics cards
  • Asus A8R32-MVP Deluxe mainboard (ATI CrossFire Xpress 3200) for ATI Radeon graphics cards
  • OCZ PC-3200 Platinum EL DDR SDRAM (2x1GB, CL2-3-2-5)
  • Maxtor MaXLine III 7B250S0 HDD (Serial ATA-150, 16MB buffer)
  • Enermax Liberty 620W PSU (ELT620AWT, 620W)
  • Microsoft Windows XP Pro SP2, DirectX 9.0c
  • ATI Catalyst 6.6, Nvidia ForceWare 84.21
  • If not mentioned specifically, we used reference graphics cards from ATI Technologies and Nvidia.

We employ a digital sound-level meter Velleman DVM1326. This instrument has a resolution of 0.1dB and allows measuring noise level in a range up to 130dB with A or C weighting.

We minimize the influence of external factors by performing the measurements at night and with closed windows; the background noise level is about 36dBA then. The sound-level reading is about 40dBA at a distance of 1 meter from the test platform when a graphics card with passive cooling is installed in it. These are the two reference numbers we base our judgments upon. The noise is measured when the system case is open.

Just like when measuring the maximum currents, we check out three operating modes – Idle, Peak 2D load, and Peak 3D load – which are described in the previous section.

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


Comments currently: 87
Discussion started: 07/31/06 03:26:14 AM
Latest comment: 01/07/17 08:38:45 AM

View comments

Add your Comment