Settling the HD debate WAS: Re: [mythtv-users] A warning about Samsung HDDs

[Steve Adeff]

On Sunday 30 October 2005 22:26, Brady wrote:
> Let me chime in on this..
>
> I worked for a large fortune 500 company a few years back. We did
> product qualification on virtually every computer component known to
> man. Mainboards, Hard Drives, PDAs, etc. The Hard Drive testing was
> slick. We would hammer a hard drive until it died in order to make
> sure that our products wouldn't be the failure points. During this
> time I saw every major brand fail in ways that you can only dream of.
> Spindle locks, head contact, 5V feedback that caused small fires, etc.
>
> Anyways, on with my point. Without disclosing anything that is
> protected by NDA I must say that I found out three major things during
> this time. Do not buy Deskstar series hard drives. Western Digital,
> while only average on there drive quality, ranks pretty high for there
> support department, and Seagate drives are unpredictable. Not in the
> way you would think though. I could tell you exact ways to fail any
> drive except a Seagate. Run the voltage a little high and then run the
> clock a little slow and some drives would fail in 30 hours of constant
> testing. Seagates on the other hand defied most failure reproduction.
> While they didn't usually last longer than other drives they typically
> failed in boring ways. Things like cache failure or just plain old
> power down failure. Most times we could recover the drives.
>
> Why do I even mention this you ask? Well, I felt that Seagate drives
> where just well built enough to withstand the tests we put them
> through and tended to only fail due to normal "wear and tear." There
> was no easy way to kill a Seagate. They seemed a little more bullet
> proof than other drives.
>
> Anyways.. Its late (I think, I just moved across a time zone boundary
> and then day light savings happened.. I have no clue what time it is
> anymore) and I am starting my new job tomorrow so I should stop
> ranting.
>
> On 10/30/05, Todd <nospam at tippyturtle.com> wrote:
> > I am sure this is going to sound like I am some type of employee...but oh
> > well, I am not...
> >
> > I have always used Seagate drives and had extremely good luck.  I started
> > using them because they were the first IDE drives to run 7200rpm at
> > 2-3bels...dead silent in comparison to the other drives 5 years ago.  I
> > believe they last longer than most because:
> > 1.  I still haven't had a Seagate die on me (I am sure there is a bunch
> > luck in there)...and I have, over the years had serious issues with WD,
> > IBM. Maxtor, Fujitsu and others.
> > 2.  Seagate has consistently had LONGER warrantee periods.  I think they
> > others are now matching Seagate, but even when it was in vogue to leave
> > your customers high and dry with 1 year _from_time_of_build_ warrantees,
> > Seagate had 3 year from purchase.  I assume they are trusting their
> > drives to live that long because the design them to.
> >
> > ...ok, now start the long threads of "I was screwed by Seagate."  :-)
> >
> > Todd
> >
> > ----- Original Message -----
> > From: "Dewey Smolka" <dsmolka at gmail.com>
> > To: <mythtv-users at mythtv.org>
> > Sent: Sunday, October 30, 2005 4:22 PM
> > Subject: [mythtv-users] A warning about Samsung HDDs
> >
> > > Hey all.
> > >
> > > I just wanted to post a very unpleasant experience as a warning to
> > > anyone looking for storage.
> > >
> > > I just had a Samsung HDD (my MBE's boot disk/recordings volume) come
> > > to a literal screeching halt less than 30 days after I purchased it.
> > > The part # is SP2514N (250GB, 7200rpm, 8M, PATA), and I bought it at
> > > Microcenter -- it's an OEM drive, so it came without any packaging.
> > > And I do mean it came to a screeching halt -- while I was watching the
> > > football game, the unit started making some really unnerving sounds
> > > and froze the system. When I tried to reboot, it wouldn't even get to
> > > GRUB.
> > >
> > > I still have no idea what happened, but can rule out a couple things.
> > > It wasn't a heat issue as I had been regularly monitoring drive temps
> > > since setting up my new rig. It generally ran at between 35 and 38 C,
> > > and maxed out about 42 C.
> > >
> > > The PSU is also new, and at 450 W shouldn't have any problem driving
> > > my P4 2.5 with 3 HDDs.
> > >
> > > I had thought Samsung drives were fairly solid, but this is the third
> > > HDD of theirs that I've had fail on me in the last four months or so.
> > > The oldest (an 80 GB ATA drive) was less than two years, the next
> > > oldest (60 GB notebook drive) was barely a year although I blame that
> > > one on overheating because the notebook's vid card was right next to
> > > the HDD.
> > >
> > > There was no indication from SMART that anything was amiss.
> > >
> > > The bright side is that I didn't lose much, and nothing that's not
> > > easily replacable or that I don't mind losing. I am also fortunate
> > > that I saved the 20 GB HDD from my old MBE, and was able to swap it
> > > into the new machine to access my music and video storage, which are
> > > on Maxstor drives.
> > >
> > > I know I can get the Samsung replaced, but it will still take me the
> > > better part of a day to reinstall everything and redo the menus etc
> > > how I had them before. This is not what I wanted to do this week.
> > >
> > > If you're looking for an HDD, I'd have to recommend you avoid Samsung.


I'm gunna stop everyone short here. 

First of all when you tested the Deskstars they were under control of the IBM 
harddrive division which for a a good stretch of time made excellent, if not 
the best quality harddrives then had a couple years of problems with their 
deskstars. Subsequently the user backlash forced them to sell to Hitachi.

Second. I've had seagate drives die on me. Moral of the story, every drive has 
the chance of dieing. I've never found a company to be found to make all 
inferior drives. Drive series have had issues (see deskstar and one of the 
Maxtor lines a while back that everyone seems to use to rip on Maxtor), but 
not drives by entire companies.

Third. Seagate had a period where they lowered their warranty to 1yr when 
everyone else was. Shortly after Hitachi raised their warranty to 3 years, so 
WD and Maxtor did as well. Seagate, who had been having some trouble with a 
couple of their models revamped and changed to a 5yr warranty, after which 
Hitachi changed as well. WD and Maxtor stayed at 3 for their consumer drives 
and iirc now have 5yr on their scsi and server series.

Fourth. Drives have varying electronics. So testing a drive one year could 
mean nothing in regards to that drives model the following year if they 
changed the electronics without changing the model name. I've got two WD 
drives of the same model number but if you look at their case and circuit 
board they look different because I bought them ~1yr apart. I've also seen 
the same with Maxtor drives.

Fifth. Two things will affect the life and performance of a drive without a 
manufacturing flaw (ie those old deskstars were gunna die on you, which is 
why IBM had such a consumer backlash about them). Power and heat. Keep your 
drives cool. They can get pretty warm and still be in their rated range, but 
the closer you can keep them to ~65-70degF (18-24C)the better. 

Power is also essential. Drives require two voltages, +5 and +12. The +5 is 
for their electronics and the +12 is for their spindle motor. This is where a 
good, proper operating, power supply comes in. A few folks here have already 
mentioned (along with my personal experience) that they had a PSU kill a 
harddrive. 

A fluctuating 5V on the low side will cause the electronics to turn on and off 
rapidly, damaging the circuit chips. A fluctuating 5V on the high side can 
damage components due to electrical "burn". Drives don't have voltage 
regulators in their electronics since they rely on the PSU to not fluctuate. 
A fluctuating low 12V will cause the spindle to start and stop randomly and 
can cause a head crash. Hard drive heads use the Bernouli (sp?) effect to 
keep the head off the platter. As the platter spins the wind it creates 
forces the head off the platter. Modern drives are near impossible to get a 
head to smash into a platter under normal operating conditions. If you look 
on the case you will see a G-Force rating, which is the max ammount of force 
the drive's head can withstand, and from what I remember this rating is based 
on impulse (which is always higher than constant). A fluctuating high 12V, as 
can be imagined, can burn spindle motor winding.

So make sure you buy a quality PSU and monitor the voltage levels. Every 
modern motherboard will give you a pretty good indication of the PSU voltage 
levels. These monitors aren't great though. Some boards the error is known 
(do a google search) some are not. But the closer to 5V and 12V the more 
accurate, but a change of .2V can make the error quite large on some boards.

If you think anything might be amiss, check the voltages with a multimeter 
while the computer is running and make sure the fluctuation you see is within 
the specifications of your PSU. If any of them fluctuate beyond that, then 
make sure your not outputting more than the rated current per V level. If you 
need more power, buy a new PSU with a higher Watt output, if your within the 
Watt rating and within the current per V level then check if your still under 
warranty or buy a new PSU.
As to how to check the PSU with a multimeter, find an open drive power 
connector (or use a splitter to make one) and while the computer is running 
(preferably under full load with a good amount of drive access) check the 5V 
and 12V voltage. Since its voltage it will be the same value at every 5V and 
12V connector connected to the PSU (except for dual PSU setups obviously...).

PSU's, while remaining under their Current per V level *should* output a 
voltage within their rated margin of error for the extent of their warranty 
period, usually longer. Cheaper PSU's will not live as long as a quality PSU 
though. I find the good PSU's will live well beyond their warranty period. 
Once you ask the PSU to supply more current than it is rated for you put it 
into an area where you are causing it to create more heat than it is designed 
to produce and will quickly cause the PSU to no longer be provide even its 
rated values.

Sixth. SMART. Older versions of SMART were not bulletproof, and I can't vouch 
for the newer versions, but I've found the newer SMART chips to not only 
provide more monitoring parameters but to also have a more stringent 
parameter range that help give a quicker indication of a problem. 

I have my smartctl logging every SMART value change. Even if a value hasn't 
passed its threshold, the speed at which it approaches can be an indicator of 
a problem. The parameter *delta* can be as much an indicator as a paramter 
failure. I currently have a drive that has a rapidly increasing parameter 
count for bad CRC's. So I know something is up. I need to call and see if 
this makes it eligable for warranty, but at least I know there is an issue to 
be aware of! 

To summarize
1. No two harddrives are made the same, even if they are from the same 
company. Don't let a harddrive failure taint your view of that company as in 
the real world it means nothing.
2. Warranty period is only an indicator of the companies willingness to 
replace their drives. Is 5yrs better than 3yr? only if you plan on operating 
that drive into its 4th year. It will have no bearing on the drive operating 
into its 2nd year.
3. Heat. Keep the drive cool.
4. Power. Monitor your power and make sure you have a properly sized PSU. 
Spend good money on your PSU as well, a cheap PSU will come back to haunt 
you. 
5. SMART. Its not just waiting for it to give you an error, its monitoring the 
change in parameters. A failed SMART parameter means the problem has come to 
an Emergency status.
6. No two harddrives are made the same, even from the same company!

Steve