http://support.microsoft.com/kb/982018
"Issue 2
Before Advanced Format disks were available, NTFS assumed that the
logical sector size that was exposed by the disk interface was equal
to the physical sector size of the disk. Although NTFS was originally
designed to theoretically support larger sectors sizes, NTFS assumed
that these two sector sizes were equal. When NTFS performs buffered
writes at the end of a file in Windows 7 and Windows Server 2008 R2,
NTFS pads the write operation to the end of the sector size which is
exposed by the disk interface (the logical sector size). This behavior
guarantees sector alignment. However, because the size of a logical
sector of a 512e drive is 512 bytes, the buffered write does not align
to the physical sector size of the drive’s hard disk. This causes the
Advanced Format disk to internally update the 512-byte logical sector
within its 4KB physical sector, which can result in performance and
reliability issues.
This hotfix introduces behavior to NTFS which makes sure that buffered
writes at the end of the file are padded to the reported physical
sector size of the disk.
This issue can manifest itself with everyday applications on an
Advanced Format disk. However, this issue occurs most frequently
when you perform many small writes, such as using the ImageX
command-line tool in Windows 7 or in Windows Server 2008 R2 to
apply a Microsoft Windows Image (.wim file) to an Advanced Format
disk.
Without the hotfix, ImageX takes significantly longer to complete the
operation than if you were to apply the same operation to a
traditional hard disk.
"
You could have a disk with 4KB internal sectors, with 512e emulation.
And then it's a matter of what the OS is able to determine, about
the true dimensions of the drive, such as physical or logical
sector size. Windows 7 likely could have handled a declared 4KB
sector just fine, but the 512e emulation provided for operation
with older OSes, may "slightly shoot in the foot" lots of things.
If you're bored, I'm sure there is plenty of material to read.
I've noticed some weirdness about my 500GB drive, that makes
me wonder how it really works inside. And if you can't get
straight answers from reporting utilities (or from the product
datasheet), it's pretty hard to conclude anything.
One thing I noticed, is optimal commands in "dd" on older drives,
are no longer optimal on my new 500GB single platter drives.
The new drive does better with "counter-intuitive" smaller block
sizes (which means, finally, the cache memory chip is actually
being used). It can make a factor of three difference in some
of the things I attempt with it. And it means fiddling around
until I find the most efficient command to use (re-start the
command, if the measured performance is wrong). I use the
Performance plugin, amongst other things, for monitoring.
Your "Transfer Rate Minimum" is pretty low. That could be
some bad blocks. Try the HDTune bad block scan, and let
it run from end to end on the disk. You're likely to see
solid green blocks, so it's not like I expect to see any
red blocks in the test. Then, rerun the transfer rate benchmark
right after that. See if the curve looks at all different.
HDTune should be less affected by alignment issues (but you
never know absolutely for sure with these things). My 500GB
drive "perked up", after being scanned once - technical
explanation unknown, since resolving pending bad blocks
is done on writes. Sequential reads really shouldn't do anything,
except to "sweep the dirt off the platter"
Paul
