I tested the cables from OCZ Technologies on a testbed with the following configuration:
- Golden Power GP-460 PSU;
- Intel Pentium 4 3200MHz CPU (Prescott core);
- ASUS P4C800 Deluxe mainboard;
- 2x512MB TwinMOS PC3200 DDR SDRAM, CL2.5;
- NVIDIA GeForce FX 5900 Ultra reference graphics card.
I attached one OCZ cable to the graphics card’s additional power connector and took the oscillograms with the help of an M621 digital oscilloscope from ETC. I performed the tests in two modes: “Idle” (the Windows Desktop on the screen, 1280x1024@75Hz display mode) and “Burn” (a botmatch running in Unreal Tournament 2004, in 1024x768 resolution, with forced 4x full-screen antialiasing and 8x anisotropic filtering, and at the maximum graphics quality settings).
So, let’s first see how well the cables suppress the noise the graphics card receives from the +5/+12v lines through the additional power connector. In the oscillogram below, the signal in the +12v line on the cable’s input is shown in green and on the output (i.e. on the graphics card’s power connector) in yellow. The unit of the X axis is 1 microsecond, and the unit of the Y axis equals 50 millivolts. This and other oscillograms were taken in the alternating current mode.
As shown, there are current surges in the input signal (green) that repeat each 3 microseconds, i.e. at a frequency of about 300 kHz. This interference is evidently produced by some system component like a CPU voltage regulator. By the way, this interference of the same value is observed both in the +12v line and in the +5v line.
But what about the OCZ cable? The oscillogram shows that the cable does reduce the interference considerably – its swing was about 75mv on the input, but only 20mv on the output. That’s an excellent result. The next oscillogram is taken under the same conditions, but on a different scale: one unit of the X axis now corresponds to 200 microseconds.
It is clear that the interference level is much lower on the output. The same goes for the +5v line, too. The interference is successfully damped by the cable’s LC filter, and there’s no sense in taking more oscillograms.