dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
3
share rss forum feed


Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12
reply to koitsu

Re: [hard drive] Concerned about HDD health

said by koitsu:

Sorry I couldn't be much more help. Welcome to one of the many reasons why I hate RAID (hardware, software, or BIOS-level -- doesn't matter which): lack of visibility into problems like this.

I'm certainly beginning to feel the same way. I originally went for it for the speed and size, but I've never had fantastic results with speed either (avg read is 120 MBps). I have heard of some people getting faster reads off single Caviar Blacks!
--
If we lose this freedom of ours, history will record with the greatest astonishment, those who had the most to lose, did the least to prevent its happening.


koitsu
Premium,MVM
join:2002-07-16
Mountain View, CA
kudos:23

Agreed. With RAID 0 and those disks, you should be getting between 240-320MBytes/second on reads and writes. That almost implies that you may have 1 drive in the pair which is... well... being a jerk somehow.

I should note I do have (well, did until they were recently sold) some WD Caviar Blacks that do not push 170MBytes/sec any longer. They did when I bought them, but they don't now. Their power-on hour counts were in the mid-to-high 20000 range, maybe 30000. I forget.

My guess is that performance of MHDDs gradually lowers over time due to magnetic substrates diminishing in quality as time goes on (I imagine this is caused more by writes than reads). Drives then (internally) have to do more ECC (when reading) and make heavier use of the ECC block per sector than they did when they were new.

Again: this is just a theory/guess, not based on actual scientific fact. The physical science involving hard disks is something I don't quite understand (as I said in another thread, I'm just a simple caveman... ).

I cannot wait until SSDs reach 1TB capacities combined with affordable prices. I would love to replace the MHDDs in my FreeBSD box (two 1TB WD10EFRX drives in a ZFS mirror, and one 2TB WD2003FYYS drive used for nightly backups of everything on the mirror, OS disk, and my Windows workstation) with SSDs for a lot of reasons -- power and noise would be the main two!
--
Making life hard for others since 1977.
I speak for myself and not my employer/affiliates of my employer.



Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12

Click for full size
Here's the before/after of the spare Samsung HD155UI 1.5 TB I am going to put the RAID on before taking it apart (I incorrectly thought it was a WD Green).

What the hell is going on with B5?


koitsu
Premium,MVM
join:2002-07-16
Mountain View, CA
kudos:23

1 edit

Sadly I'm not very familiar with Samsung drives, so I'm not quite sure what some of these attributes are. Nothing stops a vendor from using a "commonly defined" attribute for something else (e.g. Western Digital could use 0xB5 as 'Snakes On Platters' while Samsung could use 0xB5 as 'Banana Counts').

It would really help if I could see smartmontools output instead of HD Tune Pro, since smartmontools has an internal drive database of many drives, adjusting the name and decoding method of the attribute based on model/firmware/etc.. HD Tune Pro doesn't have this (well okay, it has a very tiny one for some models of SSD, but it's no where near as accurate/correct as smartmontools).

For example, temperature on this model of drive is vendor-encoded, and attribute 0xB5 could also be vendor-encoded. It's hard for me to say at this point.

Ignoring attribute 0xB5 for a moment, and assuming 0xC8 is correctly labelled (again, smartmontools... ), then the attribute indicates at one point during the zeroing the drive did experience a very, very low number of anomalies during the zeroing. Since a drive remaps an LBA to a spare sector only on a write, there may have been a few LBAs written to which caused the drive to have to issue re-write attempts before they passed, thus increasing the rate slightly. Obviously there are no remapped sectors (successful or failed for that matter), but something did happen during the writes.

I would still use the drive regardless of this attribute having incremented, but you may want to run a full read scan of the drive (reading every LBA) to make sure nothing happens there. You can do this in HD Tune Pro by using the Error tab, but make sure to uncheck the "Quick scan" option. Also be aware that HD Tune Pro and some other utilities have a known bug/issue where the error scan may suddenly start returning errors for every LBA past a certain point; this is a software bug. (It can also happen when using the Erase feature)

Finally, be aware that this model of Samsung drive is one of the drives which is known to have a catastrophic firmware bug. You can't look at the firmware number to determine if your drive has the fix or not (Samsung chose to not increment the firmware number). The only place I was able to find this fixed firmware was here.
--
Making life hard for others since 1977.
I speak for myself and not my employer/affiliates of my employer.



Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12

Is that exe not for Windows 7 64-bit? It's saying it's not for my version of Windows...

I also found it here, but some problem, it won't run.

»knowledge.seagate.com/articles/e···223571en
--
If we lose this freedom of ours, history will record with the greatest astonishment, those who had the most to lose, did the least to prevent its happening.



Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12
reply to koitsu

Click for full size
Click for full size
Click for full size
Here's the first drive. Not sure what's going on with the read speeds starting at 100 MBps and ending at 50.


Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12

Click for full size
Click for full size
Click for full size
Second drive. Finally something concrete. Do you know what C5 means for Caviar Blacks?


Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12

To make things even more confusing, the Western Digital Data LifeGuard Diagnostics tool passes that drive. It shows C5 as Current Pending Sector Count as well (200/200 as Value/Worst, but 0 for Threshold and 0 for "Warranty" column), but says both drives check out OK.

I am going to go ahead and zero both drives and see what the SMART data looks like on the other end.
--
If we lose this freedom of ours, history will record with the greatest astonishment, those who had the most to lose, did the least to prevent its happening.



Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12
reply to Krisnatharok

Click for full size
Click for full size
Click for full size
Finished zeroing the first drive. Here's the details.

Granted the juicier details will come from the health data after the second drive.

Still disappointed with the speed test.
--
If we lose this freedom of ours, history will record with the greatest astonishment, those who had the most to lose, did the least to prevent its happening.


Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12

Click for full size
Click for full size
Click for full size
And now the second drive. Offline uncorrectable increase by 1, but the warning disappeared.

Think it's still good?


Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12

Click for full size
SAMSUNG Spinpoint F4EG HD155UI 1.5TB 5400 RPM
Click for full size
OCZ Vertex 2 120 GB SSD
Read tests for the Samsung Eco, which trounces the Caviar Blacks, and the Vertex 2, which really jumps around. Is it because it is the boot drive, or is something else at play with my SATA controller?
--
If we lose this freedom of ours, history will record with the greatest astonishment, those who had the most to lose, did the least to prevent its happening.


Krisnatharok
Caveat Emptor
Premium
join:2009-02-11
Earth Orbit
kudos:12


New Raid 0 Speed Test, 2x 640 GB WD Caviar Blacks
Better read results for the fresh RAID, but still low. Hey, at least it's more consistent across the entire test.

I wonder if having data on the other two drives affected the read test.
--
If we lose this freedom of ours, history will record with the greatest astonishment, those who had the most to lose, did the least to prevent its happening.