We’ve been building a few servers recently, and as part of our testing we performed a memory test on each of the machines. In doing so we discovered that one of the RAM modules was defective. We confirmed that the memory, not the cpu nor motherboard, was at fault by then testing the RAM in another machine.
System memory can be defective without being overly apparent, allowing you to install your operating system and use it normally for a period of time. After a while programs may begin to crash unexpectedly, and other system errors may occur.
By running Memtest86 before shipping the servers to the datacenter we have saved ourselves countless hours of remotely debugging the issue and coordinating a replacement of the hardware.
Memtest86 is free and can be downloaded as a bootable iso and in many other formats here. It is also bundled on Ubunutu installation disks, just boot from the cd/dvd and run the “memory test” option.
Below is a screenshot detailing over 400,000 memory errors. Working memory should not return any errors at all, one single error is cause for concern.
Check out the Memtest86 website for more details on how to determine which memory module is at fault, and whether your bios or voltage settings may be to blame.
