Report: nine 2 TB hard drives! - BeHardware
>> Storage >> Hard Disks

Written by Marc Prieur

Published on January 6, 2011


Page 1


The first 2 TB hard drives appeared at the beginning of 2009. Western was the first to launch with the Caviar Green WD20EADS in January, Seagate replying in April with the Barracuda LP. These two drives stood out for their usage of 4 x 500 GB platters and rotation speeds below 7200 rpm: 5900 rpm on the Seagate, with no value given by Western, but approximately 5400 rpm. We carried out our first tests on these drives in July 2009.

Since then, the range has been extended, with all manufacturers now offering products at this capacity. From Hitachi, thereís the 7K2000, a model with 5 x 400 GB platters, but with a rotation speed of 7200 rpm, while the Western Caviar Black and the Seagate Barracuda XT use 500 GB platters. Samsung has two two 5400 rpm drives, the SpinPoint F3 which uses 4 platters and the SpinPoint F4 based on 3 x 667 GB platters.

Western is also marketing the first 2 TB drives, the WD20EARS, divided into 4 KB sectors, a technology weíll come back to on page 2. The most recent version (00MVWB0) has 3 x 667 GB platters, against 4 x 500 GB platters for the 00S8B1s and 00J2GB0s.
2 TB, but what for?
Should such high capacity drives be limited purely to storage? Of course, and especially when it comes to drives that donít have a 7200 rpm rotation, this is their main usage. They offer high storage capacity and a good environmental spec, which makes them attractive solutions for NASí, Home Cinema PCs and even PCs as, for example, storage support for an SSD!

Even though they generally give lower performance, they can of course also be used as system disks; rather the 7200 rpm models, however, than the 5400/5900 rpm ones.

Page 2
Advanced Format

Advanced Format
The sector is the smallest physical unit on a hard drive, and has up until now been sized at 512-bytes, compared to 4 KB per page for SSDs. As early as 1998, IBM identified these 512-byte sectors as prejudicial to the increase in capacities of drives and, as of 2000, the IDEMA (International Disk Drive Equipment and Materials Association), which provides a forum for manufacturers, set up a committee to work on the problem.

512-byte sectors cause problems in terms of the correction of errors. The higher the density and the smaller the sectors, the more problematic they are in terms of ECC algorithms. Currently, each sector comes with 40 bytes of ECC (Error Correcting Code), and all things being equal this would have to be increased in order to retain decent levels of reliability, which would partly annul any gains in capacity given by increased densities. With a 4 KB sector, 100 bytes of ECC are sufficient, instead of 40x8 = 320 bytes. Moreover, in addition to the part given over to data, each sector is made up of the preamble, the data sync and the inter sector gap. These are no larger for 4 KB sectors than for 512 byte sectors and overhead is therefore reduced by 8!

Using a larger sector size allows you to improve storage efficiency on platters. With 512 byte sectors you get a format efficiency of 87% (falling to 81% should ECC space be increased to 80 bytes per sector), but with 4 KB sectors format efficiency is 96%.

What about current BIOS and OS compatibility? No problems and compatibility here is invisible. As of 2005, Western Digital and Seagate put forward a model that has since been adopted across the board, Advanced Format 512e. These drives use sectors that are 4 KB in physical size but emulate 512-byte sectors via their firmware, which guarantees compatibility. Secondly, there will be a new transition to drives which do not use this emulation, removing compatibility with old systems Ė this isnít expected before the end of 2012 and could be pushed back.

In performance terms, thereís a negative impact. Previously, when a drive had to read just 512 bytes, it could do so. Now it has to read 4 KB to isolate the same information. When writing less than 4 KB, or a quantity of data not divisible by 4 KB, the drive has to carry out an RMW (Read-Modify-Write), which means that sector data to be written is read and then rewritten with modifications. Nevertheless, very few files are smaller than 4 KB, and all modern file systems use 4 KB blocks (smallest storage unit), which is a welcome coincidence!

As with SSDs, thereís a problem with the alignment of partitions in Windows XP. XP creates the first partition from the driveís 63rd sector based on 512 byte sectors. Unfortunaltely, this partition is misaligned with respect to a 4 KB sector boundary as it then starts at 7,875 sectors. In fact, every block of the file system is misaligned between two physical sectors, which has an impact on performance due to constant RMWs. The other operating systems donít have this problem and on its EARS series WD is offering a jumper only to be used if a partition is created in XP, which allows you to alter a sectorís address so that the partition is placed as it should be with a default Windows XP alignment.

All manufacturers have agreed to change over to Advanced Format 512e between now and January 2011 at the latest. In this test, this is the case with the Western WD20EARS and the Samsung EcoGreen F4.

Page 3
The test

The test
For this test, we mainly used the same test protocol as the one elaborated for our SSD comparison. 8 drives were tested:

- Hitachi 7K2000 (HDS722020ALA330)
- Samsung EcoGreen F3 (HD203WI)
- Samsung EcoGreen F4 (HD204UI)
- Seagate Barracuda LP (ST32000542AS)
- Seagate Barracuda XT (ST32000641AS)
- Western Digital Caviar Green (WD20EADS)
- Western Digital Caviar Green (WD20EARS, 3 platters)
- Western Digital Caviar Green (WD20EARS, 4 platters)
- Western Digital Caviar Black (WD2001FASS)

The test machine is a Core i7-920 mounted on a Gigabyte GA-X58A-UD3R with 6 GB of DDR3-1333 and a Radeon HD 5870. The synthetic tests are carried out in Windows 7 64-bit with the storage system as a secondary drive, the boot drive being an X25-M 80 GB, all with Intel RST 9.6 drivers in AHCI mode. The practical tests with the test hard drive with the OS are carried out with the Microsoft AHCI drivers from Windows 7.

Note that although the Barracuda XT supports SATA 6 Gbits, the tests were carried out on an Intel chipset and therefore at SATA 3 Gbits. Given the performance of these storage units, which are far from exceeding the interface limitations in terms of sequential speeds (around 250 MB/s), SATA 6 Gbits only adds something when it comes to the cache and therefore only makes a minimal difference.

While we prefer to use IOMeter for SSD testing, here we measured sequential speeds with h2bench so as to be able to gauge how speeds were affected by how far we were into the disk. In addition to this reading we used IOMeter to measure access times and read speeds and then random writes of 4 KB. The effectiveness of NCQ, which allows the processing of concurrent commands, is also measured by testing random accesses, read and write, with 1, 2, 4 and 8 simultaneous commands.

With respect to access times, it's worth highlighting the fact that some test tools only measure it over the 1st TB, even though this fact isnít detailed. This is notably the case with h2bench and HD Tune in their old versions, so make sure youíre using the latest versions!

Next come the practical tests with writes and reads of various groups of files. These groups are made up of:

- A collection of large files: 6.8 GB on average
- Medium: 796 KB on average
- Small: 44 KB on average

The source or the target when reading or writing to the SSD is a RAID of six 150 GB VelociRaptor drives mounted on an ARECA ARC1280ML PCI-Express x8 card with a strip size of 8 KB.

Lastly there are some purely practical tests, namely various operations timed after installation of Windows 7 64-bit on each of the SSDs:

- Start-up of Windows 7
- Installation of Photoshop CS5
- Start-up of Windows 7 + various applications
- Launch of a Crysis level

Page 4
Sequential throughputs

Sequential throughputs
We began with sequential speeds, measured using h2bench.

As expected, the 7200 rpm drives came out on top here, especially those using 500 GB platters, such as the Barracuda XT, which is the fastest, followed by the Caviar Black. The highest reads were measured at 138.6 MB/s and writes at 135.5 MB/s, with averages of 106.3 and 102.1 MB/s. As is always the case with magnetic hard drives, the lowest values are lower than half the highest as speeds diminish proportionally as you advance further into the drive.

The 7K2000 with its 400 GB platters is disappointing and not far ahead of the 5400/5900 rpm drives. Among these slower rotation drives, the EcoGreen F4 is the fastest, ahead of the the WD20EARS by more than 9 MB/s though this drive also uses 667 GB platters. The Barracuda LP is in second place, its rotation speed giving it a fillip.

Page 5
Access time

Access time
We measured average access time with IOMeter, using random 4 KB accesses. So as to view the potential gains linked to NCQ, these accesses were carried out with 1, 2, 4 and 8 simultaneous commands.

Standard access time, obtained with a single command, is quite high on these drives, with the 4 platter WD20EARS holding an unwanted record of 20.5ms! The 3 platter version does much better at 15.3ms, sharing first place among the 5400 rpm drives with the EcoGreen F3, preceeded by the 5900 rpm Barracuda LP. The EcoGreen F4 does a little less well than the F3 and nor does the WD20EADS excel here.

When it comes to the 7200 rpm drives, the Caviar Black does very well at 11.8ms, followed by the 7K2000 at 13.8ms. The Barracuda XT is however very disappointing with an access time of 16.7ms, which is strangely even higher than the LP!

NCQ does a good job in reads. Just for comparison, the Caviar Blackís best score of 7.6ms corresponds to speeds of 524 KB/s with 4 KB accesses. The fastest SSD up to now, the C300 128 GB, managed 143.2 MB/s, which is 280 times better.

The same test was carried out for writes. The numbers here are much better for hard drives because of the inclusion of a cache in writes, which does however make NCQ useless. The Caviar Black does best, followed by the 7K2000. The Barracuda XT is outperformed by three drives with lower rpm: the EcoGreen F3, the WD20EADS and the 3-platter EARS.

Page 6
File copying

File copying
This brings us to file copying. We measured read and writes speeds when copying various groups of files via Robocopy. The files were composed of:

- Large files: average of 6.8 GB
- Medium-sized files: average of 796 KB
- Small files: average of 44 KB

The source or the target when reading or writing to the drive is a RAID of six 150 GB VelociRaptor drives mounted on an ARECA ARC1280ML PCI-Express x8 card with a strip size of 8 KB. The files were read and written on a partition beginning halfway into the drive.

As expected the Barracuda XT and Caviar Black were fastest with reads of large and medium-sized files. Among the 5400/5900 rpm drives, the EcoGreen F4 is fastest with these files but the Barracuda LP came out top with small files.

When it comes to writes, again the Barracuda XT heads the field for large files, followed by the Caviar Black, but surprisingly the EcoGreen F3 does best with medium-sized files in spite of its lower rotation speed! The Advanced Format 512e gives the F4 an advantage over the F3 on large files but loses this advantage for medium-sized and small files. It is on a par with the 3-platter WD20EARS for medium-sized files, but is a good deal superior on small files.

Note the catastrophic performance of the 4-platter model with small files. Our test model had an older and lower performance firmware in comparison to the AFT.

Page 7
Practical tests

Practical tests
After raw performance levels and file copying, we now move on to some slightly more meaningful numbers, the practical tests. While for the other tests, the hard drive was set up as a secondary drive, here it becomes the primary drive. After installing Windows 7 64-bit we time various tasks:

- Windows 7 start-up
We measure how long it takes to start up Windows 7, from initialising the load to the appearance of the Windows desktop.

- Windows 7 start-up + various applications:
This is the time required to start up Windows 7, Adobe Photoshop CS 5, Excel 2010, Word 2010, PowerPoint 2010 and Outlook 2010.

- Launch of a Crysis level:
We measure how long it takes to launch Crysis and a game level launched straight from a command line.

- Installation of Photoshop CS5:
This is the time required to install Photoshop CS5 from the Adobe site download archive to the hard drive. Installation is broken down into two stages, the extraction of files from the archive, then the actual installation.

- Installation of Office 2010:
Here we measure the time required to install Office 2010 from an image on the hard drive.

For information, we have also included the results obtained for a VelociRaptor 150 GB (HDD 10K rpm) and Crucial C300 128 GB (SSD) so as to give a basis for comparison with other storage device ranges.

The Caviar Black is fastest for Windows 7 startup at 25.5s: better than a VelociRaptor! We shouldnít forget that in spite of slightly higher access times overall, Windows 7 occupies proportionally much less space on such a drive, which in turn means that less ground has to be covered when accessing the data. Only an SSD is faster, and this is especially visible when accessing multiple tasks such as Windows startup along with several applications.

Moreover this test shows up the limitations of the EcoGreen F3/F4, WD EADS and EARS and these drives are best avoided if youíre looking for a very responsive system. The Barracuda XT scores top here, followed by the Caviar Black, the 7K2000 and, surprisingly, the Barracuda LP.

The differences here are much lower. The first stage of the installation of Photoshop shows similar results on all the storage devices Ė extraction from an archive is mainly limited by decompression speed, which depends on the CPU and memory. You do however get bigger gains on the second part of the installation process and on the installation of Office 2010. Here, itís the 4-platter WD20EARS that stands out, but rather for its poor performance thus confirming the results obtained for writing small files.

Page 8
Energy consumption, temperature and noise

Energy consumption
Here is the data for energy consumption for each of the drives that we were able to read with a clip-on ammeter, by reading voltages used on the 5V and 12V lines.

In load the 3-platter WD20EARS scores lowest here at just 2.7 watts. The 7200 rpms draw a good deal more, from 6.8 to 7.1 watts. In load, the Caviar Black consumes 10.8 watts while the EcoGreen F4 draws just 5.6 watts.
We then measured the temperature of the drive after 1 hour of intensive use in IOMeter. These measurements were taken with the HDD outside of the casing, with room temperature at 25įC without the fan, each of the drives slightly raised. The temperature was measured in two places using an infrared thermometer, above the middle of the drive and at the hottest point on the left side. We also read the temperature on the internal drive sensor (SMART).

In these conditions the WD Caviar Black gets fairly hot in spite of the low ambient temperature and it would be a good idea to make sure youíre using a sufficiently effective fan if you go for one of these. The Barracuda XT is however fairly cool for a 7200 rpm model. The drives that stay coolest are the 5400/5900 rpm drives, which were all recorded at around the same temperatures.
To take this we placed each drive so as to avoid all vibrations (seeing as they were placed on the desk). A sonometer was placed 10 cm above the drive so as to measure noise pollution, the PC used being without any fan so as to isolate the sound of the drive.

In rotation, the 7200 rpm drives are the noisiest, with the Barracuda XT scoring highest. The 7K2000 is the noisiest during accesses, just 0.1 dBA ahead of the Caviar Black. Quite a bit of noise is generated here and this is something you need to weigh up against the Caviar Blackís reduced access times.

The quietest drive is the 3-platter WD20EARS, closely followed, in rotation, by the EcoGreen F4. The EcoGreen F4 is however much noisier during accesses.

Page 9

From the point of view of pure performance, one hard drive stands out, the Western Digital Caviar Black. Never caught on the wrong foot, it can claim to be faster than the VelociRaptor as a system disk, which would make it an ideal drive if it wasnít so noisy and didnít heat up so much. The 7K2000 shares the same faults but isnít as fast, while the Barracuda XT is better balanced but too expensive.

When it comes to the so-called green drives, the 3-platter Caviar Green WD20EARS and EcoGreen F4 stand out with the help of their 3 x 667 GB platters. The EcoGreen F4 will be fastest in file copying but the WD20EARS will be quieter during accesses. In spite of this, Western gets a red card from us for not changing the name of the drive with this change in density. It's still quite possible when buying a WD20EARS to find that you've got a 4-platter version (WD20EARS-00S8B1) instead of a 3 platter (WD20EARS-00MVWB0), which isn't the case if you go for an EcoGreen F4.

Neither of these drives is however a particularly good choice as system disk, and if thatís what youíre looking for, unless you go for a 7200 rpm drive, the Barracuda LP seems the most appropriate. Of course, no compromise is ideal and our absolute preference would be for an SSD + HDD 5400/5900 rpm pairing.

When youíre going to put 2 TB of data on a single drive however, the most important feature has to be reliability. This is however impossible for us to test, but we do have access to the stats from a large French e-tailer. From these stats the least reliable drive is the Hitachi 7K2000, with a 3.9% returns rate on drives sold in the first half of this year, followed by the Caviar Green WD20EARS at 2.7%, the Caviar Black at 2.5% and the Barracuda LP at 2.3%. The best performer here is the EcoGreen F3 with a returns rate of 1.2%, while the F4, which has only been on sale since October and therefore has an advantage, is at 1.9%. The sales volumes for the Barracuda XT are too low to give us any statistics.

Copyright © 1997-2015 BeHardware. All rights reserved.