[conspire] Reliable 1.5TB SATA drives?

Rick Moen rick at linuxmafia.com
Mon Jul 26 15:49:39 PDT 2010

Quoting Mark S Bilk (mark at cosmicpenguin.com):

> First of all, please delete my post to the list that didn't 
> go through because it had an attachment.  I resent it 
> without the attachment (the jpg showing the System Monitor).

Um, whatever this was about, it seems to have worked out -- but, for the
record, there was nothing from you in the mailing list's admin queue.

FYI, also, my mailing lists don't hold messages merely on account of
having attachments.  I know that some do, and apparently the logic is 
varied, sometimes on account of a desire to protect people against
MS-Windows malware.  Whatever.  We don't do that.  (Mailing list admins
who _do_ attempt such filtering should be advised that they'll have
problems with, among other things, MIME-encoded PGP/gpg signatures,
which present as attachments.)

If a posting of yours ever _is_ held in the admin queue, you'l receive
an e-mail informing you of that fact, and approximately why it happened.  
Included in that e-mail will be a URL you can visit to cancel the held
post, if you wish.  So, no need to ask for help; it'd be within your

> I bought the drives last week from Frys in Sunnyvale.  The 
> salesman paid no attention to what I was saying, and then 
> dropped one of the boxes containing a Seagate FreeAgent 
> Desk 1.5TB USB drive about 10 inches onto the table.  When 
> I said, "What are you doing?  That's a delicate machine!",
> he replied, "Don't worry, the box is tightly sealed."

Heh.  Par for the course.
Contrary to stereotype, I have gotten some good technical help from
Fry's employees.  One just shouldn't count on it.

> Apparently Frys not only shrink-wraps the returned drives, but 
> even puts on those heavy reinforced plastic straps.  But they do 
> put orange labels on the boxes.

And this is claimed to be standard practice, and I have no reason to
think it isn't.  I just note that it's a widespread issue.

> As far as I know, other stores won't take items back for a full
> refund within 15 or 30 days without a hassle.

True.  I hope you see that there's at least one potential downside.  If
retailer Foo is extremely popular with people who appreciate the right
to return with no questions asked, then a large amount of merchandise
will probably be returned with no questions asked.  So, if nothing else,
you'd better hope that orange-label policy is consistently followed.

> The suggestion from tech support was that the FreeAgent Desk 
> external USB drive was doing frequent recalibrations (lots of 
> clicking) because it was sensitive to noise pickup in the USB 
> cable, and that I should try a different cable.  I did, and 
> the problem went away.  So the drive itself was OK.  I don't 
> think smart would have been relevant to this.

No, you most definitely always need to use quality cabling.  If you
don't do that, then drive quality is kind of beside the point.

On the other thread about transfer rates, I hope you do know that the
rating figures are theoretical best-case figures at best.  Often, they
are just ceilings on what the data bus in question _could_ achieve
before saturation if only you _could_ crank data into them that fast.

For hard drives, the theoretical best case would be sequentially ordered
data (on physically consecutive clusters) within a physical cylinder,
under circumstances where absolutely nothing else is going on -- such
that one is bottlenecked by raw read/write speed and not by seek times
or rotational latency.  Real data under real conditions on real
operating systems tends to be pretty much nothing like that.  

So, what I'm saying is, you really need to read those figures in
context, and not just assume you can approach them in arbitrary
real-world scenarios.

I also note that you didn't really detail your physical configuration,
e.g., are these internal SATA drives.  

I also note that you're using an Nvidia HBA on Linux.  Well, you know
what I'd be saying about that, right?  Non-cooperative with the Linux
community, with usual consequences thereof, including effects on
available drivers.

Finally, I second Luke's comment about addressing modes for the driver.
Don't assume DMA unless you've verified.  You say you think DMA on 
account of metrics.  Hmm, well, I'd want to verify it more directly.
You say nothing in dmesg?  That's bizarre.

It could be a limitation of your driver.  Did you say what driver you
are using?  I cannot remember your having done so.  For libata drivers, 
see what Jeff Garzik says here:

More information about the conspire mailing list