20MB/s difference between identical drives

I have eight WD Red drives - four 3TB and four 2TB.  The four 3TB devices I bought about a week ago.   My machine has four different SATA controllers in it.  Two on board and two external.  All of the drives listed below except for /dev/sdj test around 140-148MB/s using hdparm -t /dev/ 

/dev/sdj consistently gets 124MB/s regardless of what controller it is on.  This is concerning since this is a disk array and this 20MB/s drop effectively slows down all read access.  Has anyone had this happen?  If so can I get the drive replaced? (NOT with a refurb.)

ata-WDC_WD20EFRX-68AX9N0_WD-WCC1T0427128 -> …/…/sdl
ata-WDC_WD20EFRX-68AX9N0_WD-WCC300511344 -> …/…/sdg
ata-WDC_WD20EFRX-68AX9N0_WD-WMC300410333 -> …/…/sdd
ata-WDC_WD20EFRX-68EUZN0_WD-WCC4M0069116 -> …/…/sde
ata-WDC_WD30EFRX-68EUZN0_WD-WMC4N0337722 -> …/…/sdk
ata-WDC_WD30EFRX-68EUZN0_WD-WMC4N0506148 -> …/…/sdi
ata-WDC_WD30EFRX-68EUZN0_WD-WMC4N0507375 -> …/…/sdj
ata-WDC_WD30EFRX-68EUZN0_WD-WMC4N0521785 -> …/…/sdb

It could be that the slower drive has an extra platter, ie lower data density.

I would weigh each drive, including the 2TB models. I expect that the difference between the 2TB drives and the fastest of the 3TB drives should amount to one platter and two heads.

Another test would involve a shortstroked read benchmark using HD Tune. I don’t know of any equivalent method in Linux.

How to determine number of heads using HD Tune:
http://malthus.zapto.org/viewtopic.php?t=650&p=1878#p1878

So the 3TB drive in question actually failed. It’s still under replacement with the store so they swapped it out. The replacement isn’t running at 125MB/s, nor is it running at 140MB/s but instead about 132MB/s. I’ll have to see how to check the number of heads…