Computer Crash? Google Says Blame the Memory
Posted 10/08/09 at 01:15:41 PM by Paul Lilly
Has your PC been on th fritz lately? If so, there's a good chance it's the system memory causing all those headaches, according to Google's research. Google, which has several thousand computers in its data centers, collected real-world data on its systems and wrote a research paper (PDF) titled "DRAM Errors in the Wild: A Large-Scale Field Study."
"We found the incidence of memory errors and the range of error rates across different DIMMs (dual pin-line memory modules) to be much higher than previously reported," according to the paper written by Bianca Schroeder, a professor at the University of Toronto, and Google's Eduardo Pinheiro and Wolf-Dietrich Weber. "Memory errors are not rare events."
The results might surprise you. Google's research reveals that correctable memory errors occur in one of every three of the company's servers each year, and one in a hundred suffer an uncorrectable error, which usually leads to a crash.
It's important to note that Google's servers use ECC (error correction code) memory, yet each module, on average, suffered nearly 4,000 correctable errors per year. So what's the big deal if they're correctable? A correctable error on a Google machine is likely an uncorrectable error on your PC, says Peter Glaskowsky, an analyst at the Envisioneering Group.
However, Glaskowsky also points out that most consumer PCs aren't manipulating tons of data in memory.

Image Credit: Bianca Schroeder
ECC
Submitted by jwalch.hawk on Thu, 10/08/2009 - 2:51pm
The quote from that Peter guy is bad. ECC is not the only way memory errors are corrected. Desktop RAM will use parity to *detect* (not correct) errors in memory. This raises an exception, and (in a very general sense) the operation just gets repeated until the memory gets it right. The fact that desktop RAM does not come with built-in error-correction doesn't mean that it doesn't detect errors, nor does it mean there's no way to correct them. It just means that they have to be taken care of by something else.
Justing changing out the OEM
Submitted by dag1992 on Thu, 10/08/2009 - 1:36pm
Justing changing out the OEM Ram in my laptop not only made it zippier but a lot more stable. When it really comes down to it, all components can make your pc crash is some form or fashion.
It's "Dual Inline Memory
Submitted by colinjm0517 on Thu, 10/08/2009 - 11:27am
It's "Dual Inline Memory Module" not "Dual Pin-Line Memory Module"
Windows 7 is the King of all OSes
That's what I was thinking.
Submitted by COMMANDER_COOK on Thu, 10/08/2009 - 6:12pm
That's what I was thinking.
Heh, Right On the money Google. Add Video drivers too.
Submitted by JohnP on Thu, 10/08/2009 - 11:18am
I built a new computer around the i920 chip and had nothing but troubles from day one. The Op Sys kept crashing or getting wiped out after a couple of weeks, taking the hard drive with it. Had to reformat or fix 3 different hard drives and reload the op sys 5 or 6 times. Someone told me to look at the memory. I moved the chips around and reseated them and POOF! all my woes went away. In my case (pun), Google got it right. WAY to embarrassed to admit that I never bothered to run a memory test on the new machine.
As for broken memory chips, I have had a few and they are really tough to figure out what is wrong. Now, any time I have an unexplained crash or trouble with the Op Sys, I run a memory test and an fdisk before doing anything else.
Remember all the troubles with Vista when it first came out? 75% of the issues folks had were with video drivers causing errors in Vista. I still see them crop up occasionally after an update. Be careful out there!
Feature
Review
Feature
Feature
Feature






