Question Pattern of Slow Sectors While Performing Read Test

[MichaelRoscopic]

I was running a read test on a 6TB Toshiba HDWE160 using Hard Disk Sentinel and started to see an obvious pattern of (very) slow sectors part way through the scan. None of the sectors came back as "bad", but there is obviously something fishy going on here with the somewhat organized pattern of slow sectors.

Here is an image for reference:
(the darker the green, the slower the read for that particular block)

Any ideas for what could be causing this pattern of slow sectors?

 

[mdd1963]

Perhaps the sectors located nearer the inside of the platters, as heads must travel the farthest to access them? Certainly not all sectors will have identical access times for this reason alone...

 

[MichaelRoscopic]

I suppose that could be a contributing factor.

I also came across a similar pattern posted by someone who was running tests on an SSD. I'm not sure exactly how this would differ from HDD results, but the explanation given for the pattern of changes in speed was related to the flushing of the cache.

I can try running the test again with a random or reverse order (rather than sequential) and see if the pattern persists. That way I can determine if the pattern is related to the sectors themselves, or to the nature of how the drive reads/buffers data.

 

[fzabkar]

It looks like one head is weak.

AFAICT, the drive has 5 heads. The pattern appears to suggest that the user area is traversed in serpentine fashion, with each head reading approximately two blocks before switching to the next head.

You can confirm this with a short stroked read benchmark in HD Tune.

How to determine number of heads using HD Tune:
http://www.hddoracle.com/viewtopic.php?f=59&t=650

If you would like to read more about serpentine traversal ...

HDD from inside: Tracks and Zones. How hard it can be?
http://hddscan.com/doc/HDD_Tracks_and_Zones.html

Edit: Google suggests that the drive has 5 platters and 10 heads. So it would appear that two consecutive heads are slow?!

 

[MichaelRoscopic]

If two consecutive heads are weak, would it seem unusual that the pattern only really shows up after about 30% of the test has been completed?

Looking at the shades of green, it also seems to get progressively slower as the test continues. The greens where the pattern first shows up are lighter when compared to the greens below.

I'll give the HD Tune benchmark a try and see what the results look like. Right now I'm just waiting for a "reinitialize disk surface" to complete and see if that has any effect.

Edit: I think I'll also try running one of these tests until the pattern is clearly visible. I will then cancel the test and immediately start it over again to see if the pattern still takes 30% to show up, or if it is immediately visible in the results.

Another Edit: I suppose it might make more sense to do some tests starting at around block 3000 instead of starting at 0 to see what happens.

 

[fzabkar]

The disc is divided into zones. Each zone has its own calibration parameters, aka "adaptives".

 

[MichaelRoscopic]

Looks like I'm going to be returning this HDD, since I just got it about a week ago.

I also tried looking up the specs and apparently it has 6 platters and 12 heads.

After running another read test for a bit longer, you can see that the pattern of slow blocks repeats faster and faster as it gets closer to the end:

It starts out with a single slow block repeating every 11 blocks. At about 83% into the scan, it is three slow blocks repeating every 8 blocks.

Is there a reason that the drive doesn't detect this as being a problem? I would assume that a "performance" drive with a whole bunch of sectors which can only be read at about 1MB/s (or less) would be an obvious sign that something is very wrong.

 

[fzabkar]

Your test results illustrate why I don't trust the HDD manufacturers' tools.