Critical Design Flaw Found in WD Caviar Green HDDs

Semple · 25 Apr 2011 at 11:40

Think i'll give mine a try when i get back.

kmufc77 · 25 Apr 2011 at 12:28

Does this look ok guys?

wonko · 25 Apr 2011 at 12:45

Hi there, LCC increasing at about 1 per hour seems fine. It would take a few decades to get to 300K at that rate (and that's not likely to kill it).

DragonQ · 25 Apr 2011 at 20:37

wonko said:
What version of dos and wdidle3 are you using? It sounds like the tool isn't running properly for some reason.

Dunno, just downloaded the ISO from the page linked to in this thread and booted from it...

AtaRo · 26 Apr 2011 at 00:36

Make your own bootable usb drive and try to use this ver. WDidle + Utility .Worked with my WD 2TB WD20EARS drive.

DragonQ · 27 Apr 2011 at 00:14

I think I found out why "wdidle3 /r" crashes for me...my older 1.5 TB WD Green appears to be dying, as my 2 TB one did a month ago. >_>

Armadillo · 28 Apr 2011 at 03:55

Adjusted mine. My 2TB green was cycling itself about 8 times per hour and that's with light usage as it's just a storage drive and I haven't really got anything on it yet, so wasn't accessing it much.

Tool set it fine, although it did freeze on the /r command, only when trying straight after /s300. Set it, reboot and it read it fine and said it was set to 300.

FreeStream · 28 Apr 2011 at 13:09

Well, that's annoying.

One of my WD20EARS died a while back, didn't save the SMART log unfortunately.

I've got three drives in external enclosures so it would be a PITA to run this software, I'll run it when I add them to my planned UnRAID server.

Given they are going to be spending most of their lives powered off anyway I don't think it matters!

DragonQ · 28 Apr 2011 at 13:38

It looks like the "wdidle3 /s300" change did take effect - my WD20EARS's load/unload count now extrapolates to ~12000 a year, which is much better than before.

My WD15EADS is now completely screwed, filing an RMA now.

EDIT: Or maybe not...WD's website produces an error when entering serial numbers. Typical. :/

exia · 28 Apr 2011 at 14:24

Just installed my WD20EARS and the LCC seemed to increase at an alarming rate,

anyhow ran the tool and set widdle3 /d (disable), all seems fine so far.

Sem · 28 Apr 2011 at 23:29

the WD2001FASS has the same problem

my head loads/unloads reached 32k after 6 months use before i disabled it

Bigpops · 30 Apr 2011 at 01:05

exia said:
Just installed my WD20EARS and the LCC seemed to increase at an alarming rate,

anyhow ran the tool and set widdle3 /d (disable), all seems fine so far.

I've had the opposite happen.

My load count has almost doubled in 150 hours, from 5000 to over 9000 since I ran wdidle3 /d.

So it took 6 weeks (from new) to get to 5000, then 150 hours to get from 5000 to 9000.

I tried it with the timer set to 300 seconds too for a couple of days and it didn't seem to make much difference, but didn't climb at the alarming rate it just has.

I guess I'll be turning the timer back on.

JimAroo · 30 Apr 2011 at 03:58

Hmmm, I've had a 1TB WD Green (WD10EACS) for over a year. It's got 5467 write cycles / 15990 Power on Hours count.

Now I just checked my new 2TB WD Green (WD20EARS), only had it for like 1/2 month. It's had 11,762 write cycles / 1045 Power on Hours count.

I'm guessing that's extremely bad?

Muel · 30 Apr 2011 at 09:35

I always laugh at stuff like this. "Critical"? It's minor at a push. Shock horror! Hard drives can die quicker if you have increased usage on them!

It even says in the article:

some software and operating systems are incompatible with the Intellipark feature causing endless head parking movement as the HDD continuously goes in/out of idle mode

That says to me that it could be the software/OS at fault just as much as the drives.

HangTime · 30 Apr 2011 at 10:07

chippy08 said:
Apprently when it reaches 300,000 cycles it will fail funny there no mention of this problem before you buy these drives.

Yeah in true MI fashion the drive self-destructs as soon as the cycle count hits an arbitrary round decimalised number of cycles. Bigbops has somehow circumvented this feature and got his to over 380k cycles - better get someone on the case now.

Quoted lifecycle values for hardware are usually based on statistical probability and just because something is rated to a value of 300k it doesn't mean to say that a good proportion of drives won't still function after 400k, or 500k, or 1 million or whatever.

wonko · 30 Apr 2011 at 10:23

JimAroo said:
Now I just checked my new 2TB WD Green (WD20EARS), only had it for like 1/2 month. It's had 11,762 write cycles / 1045 Power on Hours count.

I'm guessing that's extremely bad?

"Extremely bad" is really overstating it TBH. 1045 hours is 1.5 months running 24x7, so at that rate I make it over 3 years before reaching 300K LCC, which itself isn't likely to be a problem.

For your usage, the heads are parking on average every 5 minutes. The issue is where the disk is being systematically accessed at an interval a bit greater than the default idle3 timer, e.g. 15 seconds or so. I'd argue that that isn't intended usage for drives like these, and rather than go messing with the disk it would be better to look at the OS and applications and why it is polling the disk so frequently.

Bigpops said:
I've had the opposite happen.

My load count has almost doubled in 150 hours, from 5000 to over 9000 since I ran wdidle3 /d.

I can't see how increasing / disabling the timer could cause that - are you sure the way the drive is being used/accessed hasn't changed? I have read on some forums that the /d switch doesn't properly disable the timer on some drives or where there is more than one drive attached when running wdidle3 - did you try /S0 or /S300 or similar?

JimAroo · 30 Apr 2011 at 13:40

wonko said:
it would be better to look at the OS and applications and why it is polling the disk so frequently.

That's just it, I'm using it as a storage drive, all I store is a few apps, game patches, ISO files. I'm hardly using it.. in fact I'm only using 165GB of the 2TB available. My other storage the 1GB version, I have been streaming movies/music/saving my downloads to it for years. Yet the HD that I hardly use compared to the other has had more than double the number of write cycles.

Bigpops · 30 Apr 2011 at 14:17

I ran wdidle3 /s300 (2x wdc drives, 2x samsung and a Seagate connected) and monitored that for a couple of days. It didn't seem to make a difference.
So, I ran wdidle3 /d on them (both drives connected still) 4 or 5 days ago, and my 2TB hard drive has gone nuts by the looks of the SMART data. Used wdidle3 /r to read it back and it said all was good. If I don't have any success, I'll have to dig my server out of the corner and disconnect all the sata drives and do them one at a time.

My WDC drives are just storage drives. No OS files are stored on them. I've just set them to spin down after 20 minutes with hdparm (I thought I'd done this already, but maybe not), so hopefully this will help things a little. The drives are only accessed while I'm watching or listening to something, and when a download completes and the resulting files are copied across (by scripts) to where they belong. 80% of the time they're idle, as I only use my HTPC for a couple of hours a day usually, even on weekends.

I've just tried a linux alternative to wdidle3 called idle3-tools (http://sourceforge.net/projects/idle3-tools/) to set the timer back to 300 seconds. You can specify the device with that (/dev/sdb and sdc in my case), so hopefully that will make a difference.

I knew it was a bad sign when it took me 9 hours to successfully format the 2TB due to the 4k sectors and it lying to the OS about it. Should have bought another Samsung.

wonko · 30 Apr 2011 at 15:02

Bigpops said:
I ran wdidle3 /s300 (2x wdc drives, 2x samsung and a Seagate connected) and monitored that for a couple of days. It didn't seem to make a difference.
So, I ran wdidle3 /d on them (both drives connected still) 4 or 5 days ago, and my 2TB hard drive has gone nuts by the looks of the SMART data. Used wdidle3 /r to read it back and it said all was good. If I don't have any success, I'll have to dig my server out of the corner and disconnect all the sata drives and do them one at a time.

My WDC drives are just storage drives. No OS files are stored on them. I've just set them to spin down after 20 minutes with hdparm (I thought I'd done this already, but maybe not), so hopefully this will help things a little. The drives are only accessed while I'm watching or listening to something, and when a download completes and the resulting files are copied across (by scripts) to where they belong. 80% of the time they're idle, as I only use my HTPC for a couple of hours a day usually, even on weekends.

I've just tried a linux alternative to wdidle3 called idle3-tools (http://sourceforge.net/projects/idle3-tools/) to set the timer back to 300 seconds. You can specify the device with that (/dev/sdb and sdc in my case), so hopefully that will make a difference.

I knew it was a bad sign when it took me 9 hours to successfully format the 2TB due to the 4k sectors and it lying to the OS about it. Should have bought another Samsung.

Mmm, wish I had an obvious explanation... don't doubt what you've seen but am still puzzled - a rise of 4000 in 150 hours is 27/hour or one every 135 seconds on average, and I can't see how changing the idle3 timer could cause the rate to increase to this if it started out at the default 8 secondds. Unless wdidle3 has other effects :confused:

?

TBH, the intellipark isn't really that 'intelli' - it doesn't seem too far fetched to have a idle3 timer that is dynamic based upon the usage of the drive to avoid this issue.

I'd be interested to hear if you have any success with the linux alternative to wdidle3.

Bigpops · 30 Apr 2011 at 18:41

I'll be keeping an eye in it for sure.

The drive could be lying about what it's doing just to worry me for no reason at all. I think I read that something like that happened a couple of years ago with these drives. My 1TB drive (now with 390K cycles) has never missed a beat so far.

EDIT:

I just divided the amount of load cycles on my 1TB drive - 386685, by the amount of hours it's been powered - 15196 , and guess what I get? 26.1, so you weren't far off with 27 an hour. Now, the drives should only be accessed for a few hours a day max during the week, as I'm at work in the day time, and no one else has access. I really don't get it.