http://blogs.zdnet.com/storage/?p=638
DRAM 1000 times more errors than we thought?
-
The report seems pretty solid in terms of providing data to back up the claims, but the real question is whether these errors cause any tangible negative costs: downtime, data loss, system reliability, etc. The article says nothing about the practical ramifications of the findings.
-
It could be accurate, but I think the writers missed an important part.
When you RMA memory, usually the manufacturer asks to have you RMA the memory in a kit.
So If I bought a 8 x 2Gb memory kit, and only one card was defective, Micron asks to send the entire set back.
So done a few hundred times, and you get quite different numbers.
Just my $0.02 on that
K-TRON -
The study was done at google. I dont think they buy memory in anything as small as groups of 4, or even 40.
The study is available on line and is linked in the article. -
A server at google that is being used 24/7/365 that is taxing its memory is going to have scads more errors than anything a consumer would use.
-
Why would it matter anyways, with ECC memory?
DRAM Error rates
Discussion in 'Hardware Components and Aftermarket Upgrades' started by prime, Nov 3, 2009.