Stress Testing a Video Card

Video cards, video drivers, and computationally intensive simulations and games cause a significant percentage of crashes for home and business users. Many of these crashes are caused by bugs in the applications, drivers, game engines (such as the Unreal engine), and 3D rendering software (such as OpenGL or DirectX), but some are also due to a slow failure of the system's video card. It is possible to stress test a video card by running applications that require significant video memory and processing. Crashes can range anywhere from a garbled screen and hung system to a blue screen of death and total system shutdown. In addition, many vendors (such as AMD/ATI and NVidia) have monitoring/tuning tools for the graphics cards that allow other key parameters (such as temperature and GPU performance) to be monitored.

One of the easiest ways to stress test a graphics card is to use the Really Slick Screensavers. These are a set of screensavers that I discovered in high school during my early days of running RedHat/Fedora Linux. Most of the screensavers can be tuned to fully utilize the hardware (particularly the lattice screensaver) and may help to implicate the graphics card as a failing component. The last time I visited the site, there was both an installer and a zip file containing the screensavers. Once the screen savers are installed, they can be configured and testing can begin.



Looking at the settings for the lattice screensaver,



The computational complexity and the stress applied to the hardware can be increased by increasing parameters such as the latitudinal and longitudinal divisions of the tori as well as the depth and density. Not all shutdowns indicate hardware failure, they may also indicate other issues such as a failing/underpowered power supply and overheating.

See also,
Windows Crash Dump Analysis 
Identifying Cooling Issues
Troubleshooting Memory Errors
Stress Testing a CPU to Detect Hardware Failure
How to Detect a Failing Hard Drive