[SOLVED] 2 identical HDDs - which one would you return?

[Wroblewski]

I have two identical new TOSHIBA MG09ACA18TE, "First" and "Second". I want to keep only one. The SMART values of both disks look OK and about the same, and while the HDDScan graphs look quite similar, I observe something I don't understand in the graphs (please see the imgur link provided):

• First has this steep downslope at the innermost (?) sectors
• Second has this sudden "up-jump" in speed somewhere betwenn LBA 5bn and 7.5bn

Are these of any significance? If so, what's your advice on them? Should I prefer First above Second or vice versa?
Thanks.

HDDScan graphs here

 

[Ralston18]

A coin flip could be used to decide...

Very likely that the software and hardware involved provides LBA values with some + or - error range.

So the variances may actually be moot or meaningless. I.e. Margin of error...

If there no performance problems or issues - I would not worry about the indicated SMART values.

Keep both drives. Always good to have a spare - maybe just for a backup clone.

Just my thoughts on the matter. There may be other ideas and suggestions.

 

[fzabkar]

I'd probably keep the second one. However, since you are running these tests under Windows, which is a multi-tasking OS, it could be that any "blips" in the curve may be due to other tasks competing for the attention of the CPU.

Try running a read benchmark with HD Tune. Select a reduced LBA range for your tests (I don't know whether HDDScan can do this). You will then see the individual performance of each head.

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

 

[USAFRet]

Flip a coin.
Rely on actual backups, in case it dies.

I just had the 16TB version of that die on me. 7 months old.
Took Toshiba 3 months to figure out it was faulty and send me the replacement.

 

[Wroblewski]

Thanks for every reply, I really appreciate! And I've learned something, especially about testing heads (yes, HDDScan can be limitied to a certain LBA range).

Please allow one question @fzabkar : You'd keep the second one. Because of that steep downturn in speed towards the end ? What could be the reason for this? A head? TIA

 

[fzabkar]

I don't know what that downturn is. I just don't like the fact that the curve isn't smooth. However, if that was my only drive, I would keep it. If you run the test again, do you still see that dip?

By the way, if this drive runs out of warranty and one of the heads goes bad, it is possible to switch off that head and rebuild the drive with a reduced capacity. I am not aware of any tools that can do this yet, though.

Enterprise Capacity HDD MG09 Series Product Manual:
https://toshiba.semicon-storage.com...rise/P_L_MG09ACA16TAY_210202072836_1_0_0E.pdf

These new ATA commands return the status of each "physical element" (read head), and you can then remove one or more heads and truncate the capacity accordingly (pages 11, 12, 15).

SATA/ATA commands:

12h GET PHYSICAL ELEMENT STATUS
7Ch REMOVE ELEMENT AND TRUNCATE

SAS/SCSI commands:

9Eh/18h REMOVE ELEMENT AND TRUNCATE (16)
9Eh/17h GET PHYSICAL ELEMENT STATUS

 

[Wroblewski]

Thanks again! Will run the test once more and report.

 

[mdd1963]

Are these not 1 TB WD drives, worth about $20-$25 each these days? Keep both, run them in a mirror, and, also have a detached periodic full backup.

 

[USAFRet]

Are these not 1 TB WD drives, worth about $20-$25 each these days? Keep both, run them in a mirror, and, also have a detached periodic full backup.

18TB
https://www.newegg.com/toshiba-mg09aca18te-18tb/p/1B0-0011-00322

 

[Wroblewski]

Also with HD Tune the curves don't look really smooth:
First (left column) and Second (right column), run 3 times with 3/10/20GB in short stroke mode.
Plus, HD Tune displays a pattern that isn't very clear, in the sense that there are no (well, I don't see) clear thresholds. However, if the repeating glitch represents one head (?), I guess both HDDs seem to have a weak/slow head (sorry, my comprehension of these things is quite limited).
While HDDScan can be limited to a LBA range, I didn't succeed to finish a partial test successfully yet, will retry.

 

[fzabkar]

The performance graphs wouldn't bother me, especially if HDDScan reports no slow sectors. As for differing performance between heads, that's normal. It's a consequence of manufacturing tolerances. Some heads have a better frequency response and their platters will be recorded with a higher bits per inch spec. This enables you to count the heads in a HD Tune short stroke test.

The following article explains it all.

HDD from inside: Tracks and Zones:
https://hddscan.com/doc/HDD_Tracks_and_Zones.html

 

[Wroblewski]

Thank you!