WD Ultrastar DC HC520 / HGST He12 50% slower in the center

HGST Internal SSD Internal SSD
1

The WD Ultrastar DC HC520 / HGST He12 should have a transfer rate of 243MiB/sec.
This is only true for the first two TBs, but than it degrades to 50% in the center of the drive.

Used the following script to run 5 cycles. For each cycle measured the transfer rate at each start of a TB.

#!/bin/bash
for run in 1 2 3 4 5
do
	for i in 11 0 1 2 3 4 5 6 7 8 9 10 11
	do
		echo "RUN=$run SKIP=$i"
		b=`expr 1024 \* 10`k
		s=`expr $i \* 100`KB
		c=`expr 1024 \* 1`
		dd if=/dev/sdg bs=$b iflag=direct skip=$s count=$c status=none | pv -b -t -r >/dev/null
		/usr/sbin/smartctl -a /dev/sdg | grep ^194     # show temperature
		echo
	done
done

b=10240k, as this is an optimal buffer size for these drives
s=skip, will skip number * b size
c=count, will read count * b size
iflag=direct, no OS buffer is used, raw/direct reading from the drive

Each cycle showed the same transfer rate (there was no variation) at a specific TB start, independent of temperature. The drive was not mounted.

SKIP=0 10.0GiB 0:00:42 [ 242MiB/s]
SKIP=1 10.0GiB 0:00:42 [ 242MiB/s]
SKIP=2 10.0GiB 0:00:43 [ 235MiB/s]
SKIP=3 10.0GiB 0:00:45 [ 226MiB/s]
SKIP=4 10.0GiB 0:00:46 [ 220MiB/s]
SKIP=5 10.0GiB 0:00:48 [ 210MiB/s]
SKIP=6 10.0GiB 0:00:50 [ 202MiB/s]
SKIP=7 10.0GiB 0:00:53 [ 192MiB/s]
SKIP=8 10.0GiB 0:00:57 [ 178MiB/s]
SKIP=9 10.0GiB 0:01:01 [ 166MiB/s]
SKIP=10 10.0GiB 0:01:08 [ 149MiB/s]
SKIP=11 10.0GiB 0:01:18 [ 129MiB/s]

RUN=1 33Celsius
RUN=2 39Celsius
RUN=3 42Celsius
RUN=4 43Celsius
RUN=5 44Celsius

Did somebody else notice this 50% drop of transfer rate for these drives?
Did you receive a sustainable transfer rate through the drive?
Hope to have some users running my script and showing their results for similar drives.

I noticed the 5 seconds clucking sound in my drive when active/idle as described in Recurring "clucking sound" every 5 seconds (Western Digital Gold)
Although during the transfer, the rate which is shown every seconds does only vary some 10MiB/s, which is in my opinion caused by reading/flushing the buffer.
Could this still have something to do with it?

Drive information:
Device Model: HGST HUH721212ALE600
Firmware Version: LEGNT3D0
SATA Version is: SATA 3.2, 6.0 Gb/s (current: 6.0 Gb/s)

Didn’t find any firmware updates. Is there an updated firmware available?

2

The optimal buffer size for a hard drive is the size of its write cache. So here you should set:
bs=256M
A drastic reduction in write/read speed when the capacity is >75% is normal. This is because the inner tracks carry much less data per rotation. Reading speed should always be better than writing.
I would be happy to receive an adapted script for writing with new measurement data.

3

I use Windows and I find the HD520 works fine even when filled near capacity. When full I tend to get more disks and maybe a NAS box or 3.