[conspire] question about EIDE ATA hard drive to buy
rick at linuxmafia.com
Wed Dec 20 21:20:52 PST 2006
Quoting David E. Fox (dfox at m206-157.dsl.tsoft.com):
> All HD's have had issues, but I remember people having them over
> Western Digital, Maxtor and even IBM (the so-called *DeathStar*. PS I
> have one of those, in continuous use since 2000.
Earlier today, I recapped some the rhetoric on the Net about the
Deathstar scandal, which rhetoric peaked 2001-2004, and lead to a
class-action lawsuit, now settled, and (some would argue) helped
motivate IBM's sale of that product line to Hitachi.
Just for context, these are all 30 and 40 GB drives (the 60GXP and 75GXP
product lines), using then-new Giant MagnetoResistive encoding
(http://www.research.ibm.com/research/gmr.html) to achieve higher
densities. Most people claim the problem was at a Thailand
manufacturing house -- though a small minority claim Hungary.
There was some chance that the passage of five years might settle the
question of what happened. Well, as it turns out, sort of:
http://techreport.com/onearticle.x/6292 mentions one flamboyant bit of
fallout from the class-action suit: A _Maximum PC_ Feb. 2004 two-page
article: http://www.sheller.com/PDF/2004.01.09_Maximum_P1.pdf It
includes a number of eye-opening IBM internal memos and e-mails.
o There was a deliberate public-denial compaign, at the time
that insiders knew better. They lied fluently and often.
o The failure rate was known within the firm to be 10x higher
than was being publicly admitted.
o After an OEM returned 5000 units for having a 2.5% factory
failure rate (10x higher than normal), an IBM executive wrote
of intention to turn right around and resell them to other
customers "without rework".
o IBM documents admitted that the firm itself was unable to
distinguish in its records "Rev 2" units with improvements
intended to fix the problems from "Rev 1" drives.
None of the available documention furnishes a technical smoking gun,
but I can hazard some educated guesses, based on some telling facts:
o Certain customers had very high failure rates during the warranty
period; others had entirely normal rates, ones within expectation.
o The lion's share of the complaints concerned the 7,200 RPM units
rather than the 5,400 RPM ones.
o Those drives were known for running relatively hot.
o 7,200 RPM drives were a pretty new category, at the time.
Here's the thing: HEAT BUILDUP KILLS DRIVES -- and many customers,
especially in the desktop PC world, are utterly clueless about that.
For a given generation of hard drive, heat output is roughly
proportional to spindle speed: A 15k RPM drive is hotter than a 10k RPM
drive, is hotter than a 7200 RPM drive, is hotter than a 5400 RPM drive.
And yet, people just clamp these things into their dinky little
underventilated PCs, close the cases, and never bother to check or
I'll bet *I* could have run farms of those 60GXP and 75GXP drives, and
so could any sysadmin, without failure rates above industry average --
because we actually _check_ how hot equipment is running, and
fix hot spots before something gets fried.
We'll probably never know for sure, but I'd lay long odds on that hunch
Cheers, "< > Kernel support for JAVA binaries (obsolete) (NEW)"
Rick Moen -- Linux kernel v2.2.19 configuration
rick at linuxmafia.com
More information about the conspire