Stress testing system components

Stress testing involves running applications which stress components of a system (either individually or collectively). Its most common uses are for diagnosing potentially faulty components or testing the stability of overclocked hardware.

See the Memtest86 guide.

There are a number of applications which can be used to test these components. Generally the same principles are followed:

• determine an appropriate temperature monitoring application (often one comes with the motherboard on the driver CD). Alternatively,
  • Speedfan supports most motherboards;
  • as does Everest (discontinued free edition available here); and
  • the CoreTemp Utility is reportedly the most accurate.
  • PCWizard 2006 is a good all-round stress-tester and system monitor.
• run a stress testing application, ensuring that temperatures are well controlled (as a guide, <50 degrees on an AMD Athlon 64 or <65 degrees on a Core 2 Duo or Pentium 4 or Pentium D)

If the stress testing application returns an error, or the computer freezes or restarts, then the system is not stable.

Prime95 is a favourite for overclockers. If a system is able to run the blend test for 6 or more hours, it is generally considered to be completely stable.

The blend test stresses both ram and CPU. The small FFT and large, in place FFT stress tests only stress the CPU.

Stress Prime 2004 has a handy front end for Prime95, which is not especially user-friendly. For dual core processors, the Orthos edition referred to on that page is required.

There are a few methods you can use to stress test a graphics card:

• Run one of Futuremark’s 3DMark benchmark utilities, which can be downloaded here.
• Spend a reasonable amount of time (30 minutes or more) playing a graphically intensive 3D game.
• Run rthdribl, the Real-Time High Dynamic Range Image-Based Lighting demo. Unlike 3DMark, this demo just does the same thing over and over, and will max out your GPU’s load temperature the fastest. It will reveal any instability with artifacting, but it’s a good idea to also run 3DMark when doing general stability testing. You’ll need a card that supports Pixel and Vertex Shader versions 2.0 and later.

An unstable graphics card can cause freezes, restarts, a crash to desktop, or graphical anomalies (such as artifacts).

In all cases, temperatures should be monitored. Most newer graphics cards have the ability to display temperatures, and both ATI and nVidia have utilities to do this which form part of their standard drivers (and in particular on the most up-to-date drivers).

Rivatuner supports GeForce and Radeon 9 series cards, and has a number of useful functions (including overclocking and temperature logging).

Suitable maximum temperatures vary from graphics card to graphics card.