Difference between revisions of "Advanced Format"

From ArchWiki
Jump to navigation Jump to search
m (→‎Disable via hdparm: Removed extra row between headings)
(→‎Current HDD Models that Employ a 4k Sectors: - ST1000DL002 has 4096 physical sector size)
Line 38: Line 38:
| Samsung        || HD204UI    || 2.0 TB || 512  || 512
| Samsung        || HD204UI    || 2.0 TB || 512  || 512
| Seagate        || ST1000DL002 || 1.0 TB || 512  || 512
| Seagate        || ST1000DL002 || 1.0 TB || 512  || 4096
| Seagate        || ST2000DL003 || 2.0 TB || 512  || 512
| Seagate        || ST2000DL003 || 2.0 TB || 512  || 512

Revision as of 11:47, 18 April 2012

This template has only maintenance purposes. For linking to local translations please use interlanguage links, see Help:i18n#Interlanguage links.

Local languages: Català – Dansk – English – Español – Esperanto – Hrvatski – Indonesia – Italiano – Lietuviškai – Magyar – Nederlands – Norsk Bokmål – Polski – Português – Slovenský – Česky – Ελληνικά – Български – Русский – Српски – Українська – עברית – العربية – ไทย – 日本語 – 正體中文 – 简体中文 – 한국어

External languages (all articles in these languages should be moved to the external wiki): Deutsch – Français – Română – Suomi – Svenska – Tiếng Việt – Türkçe – فارسی


The 'advanced format' feature reduces overhead by using 4 kilobyte sectors instead of the traditional 512 byte sectors. The old format gave a format efficiency of 87%. Advanced Format results in a format efficiency of 96% which increases space by up to 11%. The 4k sector is slated to become the next standard for HDDs by 2014.

More Detailed Explanation

The main idea behind using 4096-byte sectors is to increase the bit density on each track by reducing the number of gaps which hold Sync/DAM and ECC (Error Correction Code) information between data sectors. For eight 512-byte sectors, the track also holds eight sector gaps.

By having one single sector of size 4096-byte (8 x 512-byte), the track holds only 1 sector gap for each data sector thus reducing an overhead for a need to support multiple Sync/DAM and ECC blocks and at the same time increasing bit density.

Linux partitioning tools by default start each partition on sector 63 which leads to a bad performance in HDDs that use this 4K sector size due to misalignment to 4K sector from the beginning of the track.

External Links

Current HDD Models that Employ a 4k Sectors

As of June 2011, there are a limited number of HDDs that support "Advanced Format" or 4k sectors as shown below.

All drives in this list have a physical sector size of 4096 bytes, but not all drives correctly report this to the OS. The actual value reported (via new fields in the ATA-8 spec) is shown in the table as the physical reported sector size. As this is the value partitioning tools use for alignment, it is important that it should be 4096 to avoid misalignment issues.

The logical sector size is the sector size used for data transfer. This value multiplied by the number of LBA sectors on the disk gives the disk capacity. Thus a disk with 4096 byte logical sectors will have a lower maximum LBA for the same capacity compared to a drive with 512 byte sectors. Drives with 512 byte logical sectors offer better compatibility with legacy operating systems (roughly those released before 2009) however drives with 4096 byte logical sectors may offer marginally better performance (e.g. more read/write requests may fit into the NCQ buffer.)

Manufacturer Model Capacity Reported sector size (bytes)
Logical Physical
Western Digital WD30EZRX 3.0 TB 512 4096
Western Digital WD30EZRSDTL 3.0 TB
Western Digital WD25EZRSDTL 2.5 TB
Samsung HD204UI 2.0 TB 512 512
Seagate ST1000DL002 1.0 TB 512 4096
Seagate ST2000DL003 2.0 TB 512 512
Seagate ST3000DM001 3.0 TB 512 4096
Western Digital WD20EARS 2.0 TB 512 512
Western Digital WD15EARS 1.5 TB 512 4096
Western Digital WD10EARS 1.0 TB
Western Digital WD10EURS 1.0 TB
Western Digital WD8000AARS 800.0 GB
Western Digital WD6400AARS 640.0 GB
Western Digital WD10TPVT 1.0 TB
Western Digital WD7500BPVT 750.0 GB
Western Digital WD7500KPVT 750.0 GB
Western Digital WD6400BPVT 640.0 GB
Western Digital WD5000BPVT 500.0 GB
Western Digital WD3200BPVT 320.0 GB
Western Digital WD2500BPVT 250.0 GB 512 4096
Western Digital WD1600BPVT 160.0 GB
Note: Readers are encouraged to add to this table.

Aligning Partitions

Check your partitions alignment

Note: This only works with MBR, not GPT.
# fdisk -lu /dev/sda
# Device     Boot      Start   End         Blocks      Id System
# /dev/sda1            2048    46876671    23437312    7  HPFS/NTFS

2048 (default since fdisk 2.17.2) means that your HDD is aligned correctly. Any other value divisible by 8 is good as well.

GPT (Recommended)

When using GPT partition tables, one need only use gdisk to create partitions which are aligned by default. For an example, see SSD#Detailed_Usage_Example.

MBR (Not Recommended)

One can employ fdisk to align partitions to sector 2048 which will ensure that the partitions are aligned to the 4k sector. Interestingly, in sector mode, the default starting point is not 63 or 64 but 2048 in the current version of fdisk (2.17.2) so it is automatically taking care of the 4k sector size!

# fdisk -c -u /dev/sda

Special Consideration for WD Green HDDs

FYI - this section has nothing to do with Advanced Format technology, but this is an appropriate location to share it with users. The WD20EARS (and other sizes include 1.0 and 1.5 TB driver in the series) will attempt to park the read heads once every 8 seconds FOR THE LIFE OF THE HDD which is just horrible! To see if you are affected use the smartctl command (part of smartmontools). If the last column changes rapidly, this section applies to your drive.

# smartctl /dev/sdb -a | grep '^193'
193 Load_Cycle_Count        0x0032   001   001   000    Old_age   Always       -       597115

Disable via hdparm

Use hdparm in /etc/rc.local to disable this 'feature' and likely add life to your hdd:

# echo "hdparm -S 242 /dev/sdX" >> /etc/rc.local

Alternatively, the following shell script can accomplish this automatically:

for DISK in `fdisk -l | grep [12]000.4 | cut -c13-13`; do
    echo hdparm -S 242 /dev/sd$DISK >> /etc/rc.local

Western Digital's solution

According to Western Digital the solution to this problem is different than the one above, see link for more information. Their solution is instead to disable Advanced power management (Uses more power as turns off all low power modes but results in no load/unload cycles) . Use hdparm in /etc/rc.local to disable Advanced power management:

# echo "hdparm -B 255 /dev/sdX" >> /etc/rc.local

Disable atime

Be sure to disable atime by adding the "noatime" or "relatime" (default since kernel 2.6.30) option to each mount in fstab. Without this flag every file access will force a write to disk waking it up.

Disable via changing firware value (persistent)

Warning: The tool used in this process is experimental, use at your own risk!
Note: This method is persistant, you only need to do this once for every drive.

This method will use a utility called idle3ctl to alter the firmware value for the idle3 timer on WD hard drives (similar to wdidle3.exe from WD). The advantage compared to the official utility is you do not need to create a DOS bootdisk first to change the idle3 timer value. Additionally idle3ctl might also work over USB-to-S-ATA bridges (in some cases). Download idle3ctl, extract and compile it. Within the folder that contains the newly compiled binary, execute

 $ sudo ./idle3ctl -g /dev/your_wd_hdd

to get the raw idle3 timer value. You can disable the Intellipark feature completely, with:

 $ sudo ./idle3ctl -d /dev/your_wd_hdd

or set it to a different value (0-255) with (e.g. 10 seconds):

 $ sudo ./idle3ctl -s 100 /dev/your_wd_hd

The range 0-128 is in 0.1s and 129-255 in 30s. For the chages to take effect, the drive needs to go through one powercycle, meaning powering it off and on again (on internal drives, a reboot is not sufficient).

If your WD hard drive is not recognized, you can use the --force option. For more options see:

 $ ./idle3ctl -h

Also make sure to disable atime, #Disable_atime.