Driving Data

Published Date
01 - Nov - 2005
| Last Updated
01 - Nov - 2005
Driving Data
The venerable HDD has undergone changes over the past couple of years. Here's a shootout to help you choose your primary storage solution

The Pentium I machines that hit the Indian market in the early '90s were equipped with hard disks of capacities less than 1 GB, and ran Windows 3.1. Now, a little over a decade later, it's difficult for us to find a gig of free space even on our 80 GB hard disks!

Those who've been associated with computers for the past decade would recollect the 32 GB limit-an indication of how early motherboard manufacturers underestimated the development of storage media.

We have indeed come a long way. The most recent development in the hard disk area is SATA II, a major improvement over the conventional parallel interface (IDE, short for Integrated Drive Electronics), and somewhat of an improvement over its predecessor, SATA I. Speeds have shot up from 150 MBps (in SATA I) to 300 GBps in SATA II. Technologies such as Native Command Queuing have made an entry into the SATA specification, though these require compatible motherboards for optimal performance.

Other than speeds, storage capacities, too, have been on the rise, as you know only too well. Desktop machines now feature hard disks with capacities of 250 GB and even 500 GB. You might criticise the need for such humongous capacities, but some people do need this much space, even on a desktop machine-people like gamers, professionals in animation and graphic imaging, and so on.

A PC game of today can consume two to five GB of disk space: for example, Unreal Tournament 2004 takes up four GB. If you're using a 40 GB disk, you could complete a game, then delete the installation to install and play another one. But we don't like to do that-we like to keep everything we've got! Similarly, most of us dump movies and music onto our drives, and find it psychologically hard to delete a few to create some room.

So increased capacities and speeds it will be, and SATA is the way to go.

The test centre compared fifteen hard disks from five well-known manufacturers. Several of  these were SATA II, except for Samsung's two Parallel ATA drives. SATA II drives are backward-compatible with SATA I, which is a welcome feature since some motherboards do not support the faster version. (See box How We Tested.)

Before we settle down to comparing the drives on the basis of features, let's do away with the odd ones in the lot-the Samsung SP0411N and SV0411N. These are IDE drives with a 2 MB buffer; and moreover, the SV0411N, at 5400 rpm, was way below par. It's not really fair to compare this disk with the others, but we did it anyway-
if the cost per GB turns out good, such disks can serve as large data dumps.

The contest for the top spot was very close indeed! The Hitachi drive could have been the absolute winner only if it had better shock resistance (it coul withstand 55G). The Seagate Barracuda (63G) and the Samsung SATA (63G as well) drives, too, were behind the frontrunner, WD (65G)- only in terms of shock resistance, however. The Hot-Plug feature encourages one to carry disks in and out of the machine, so high shock resistance becomes important-it is a measure of the portability of the hard disk.

The Hitachi drives came with a bootable floppy containing the 'Hitachi Feature Tool', using which we could change the drive settings-from 1.5 Gbps to 3.0 Gbps-and even monitor the drive temperature.

Most switched-mode power supplies (SMPSes) have provision for the Molex power connector, which is the five-pin power connector used for CD-ROMs and IDE hard disks. Provision for a separate SATA power connector is usually absent in SMPSes, so including Molex power support along with a SATA power connector on the drive is a good thing. It would be ideal if manufacturers such as Seagate, Samsung and Maxtor had provided this.

WD's RAID Edition (Caviar RE2) had a few features unique to the maker, namely RAID-TLER and RAFF. (See box Jargon Buster.)

With the increase in capacity, the data density on the platter increases, and the accuracy of the head becomes important, especially in drives such as WD's Raptor (it's a 10,000 rpm disk). RAFF makes sure rotational vibrations are countered, so that the head reads data with minimal errors.

Based on support for both types of power connectors, additional features, and shock resistance, WD edged out the Hitachi drives to dictate terms in the Features section of the comparison.

The WD Raptor delivered outstanding scores in the synthetic tests, followed by other members of the WD family, and the lone Seagate. The Hitachi drives recorded the highest burst speeds in the HD-Tach benchmark. However, there's not much one can deduce from this, since the operating speed of the drive is much less than the burst speed. At the same time, this could have boosted Hitachi's performance in the real-world tests.

How We Tested 
To ensure that all the disks got equal treatment, we maintained a common test bed that would meet the requirements of all the disks. We installed Windows XP on our hard disk-a Seagate Barracuda 120 GB SATA. The test hard disk served, in each case, as a secondary drive. The test bed was powered by an AMD FX53 processor on an Asus A8N SLI Deluxe motherboard, with 1 GB of Corsair DDR RAM and an ATi Radeon X850 XT 256 MB Platinum Edition.

Among the nine listed features, points were only given to the type of power connector, additional features, and shock resistance, as the rest of the features contributed to the Performance and Cost Per GB factors. Most of the additional features-such as NCQ and Hot-Plug-were common to all the drives, and some (such as RAFF, TLER and TCQ) were specific to the manufacturer; only the Parallel ATA drives differed in additional features, so the considerable difference in scores reflected only in the case of the Samsung's parallel ATA drives. The variation in this segment was negligible among the other drives.
Disks that supported both types of power connectors-Molex and SATA-were awarded more points in the Features category.

Our performance tests were classified into the synthetic tests and the real-world tests. The synthetic tests included the HD-Tach and SiSoft Sandra benchmarking tools.
HD-Tach runs on a raw partition. We did not award points to the minimum and maximum read/write speeds, since the data from those readings was considered inconsistent. Points were therefore allotted to average read and write speeds, CPU usage, burst speeds and random access time.
We used SiSoft Sandra's File System benchmark to estimate the Drive Index, Sequential Read, Sequential Write, Random Read, Random Write and Access Time. The score for read and write operations are in MB per second, while Access Time is in milli seconds.
The real-world test included the file transfer and Photoshop image load tests. We created a 700 MB assorted folder containing files of various sizes and types, whereas a 700 MB video clip served as the sequential file. The inter-drive transfer results were obtained by recording the time taken for transferring 700 MB of assorted and sequential data between the test hard disk and our primary hard disk (the one that had the OS on it).
 A RAM drive is a virtual hard disk partition created using software. A pre-defined (712 MB) amount of RAM was allocated for this purpose. The intra-drive test was weighted over the other transfer tests as it tests the effectiveness of the drive's buffer.
 For the Photoshop image load test, we created image files of varying sizes by increasing the resolution of a multi-layered image (in the PNG format). We created two FAT32 partitions on each test hard disk. The path to Photoshop's scratch disk, which acts as a virtual RAM space on a hard disk's partition, was changed to the test sample's first partition, and the images were stored in the second partition of the same test sample. The time taken to open the images was recorded as the observation.

Price Index And Warranty
We compared the drives based on cost per GB, obtained by the ratio of price to formatted capacity. Drives that had a higher capacity at a lower price scored higher on the Price Index. Warranties were either three or five years, and points were awarded accordingly.

Picking The Winners
The scores from Features, Performance, Price Index and Warranty were scaled out of 100 with relevant weightages given to each. The product that scored the highest was awarded the Digit Best Buy Gold, and the one that trailed the Gold winner was adjudged the Digit Best Buy Silver.

Synthetic Tests

HD-Tach reports the burst speed of a drive in addition to other readings such as random and sequential read/write, which is common with the other benchmark tool-SiSoft Sandra. Maxtor's DiamondMax 10 trailed WD's Raptor by a small margin in average read and write speeds.

Note that the maximum and minimum read/write speeds is just an outcome of the graph that HD-Tach plots while benchmarking; the highest peak is the maximum speed and the lowest dip is the minimum speed. Taking the average of these doesn't make sense. It's the real average read and average write speed that should be looked at, rather than the maximum and minimum values.

SiSoft Sandra
When we converted the benchmark scores obtained from SiSoft Sandra to evaluation points, the lone Seagate-the Barracuda 7200.8-was right on top. This 400 GB, 1.5 Gbps drive with an 8 MB buffer delivered scores that edged out the top performers of WD, Hitachi and Maxtor.

The results we obtained using SiSoft Sandra's File System benchmark gave us information about the Sequential Read, Sequential Write, Random Read, Random Write, Access Time and overall Drive Index. While HD-Tach rated Raptor very high for its access time, SiSoft Sandra's result for the same parameter was misleading. We repeated the test but ended up with the same results.

Real-World Tests
The real-world tests are classified into two broad areas-file transfer time and Photoshop image load time, as explained in How We Tested. Real world tests are designed to provide information that you can easily identify with so as to judge the performance of a drive. A 700 MB data transfer between two disks was timed in seconds, and the results are presented in the table of results.

The file transfer test is classified into three sections-inter-drive, RAM drive, and intra-drive. The disk's speed is tested to its limits as it tries to match the speed of the RAM. The speed at which the data transfer occurs wholly depends on the drive's read/write speed, spindle speed and access time.

Technically, the 10,000 rpm Raptor had to perform well in this test, and it did. The Seagate managed to narrowly edge out the Raptor in one of the four recorded timings. In the rest of the transfer tests, both drives were neck-to-neck.

Time to check out the real transfer rates of the drives! Inter-drive file transfer was carried out between a test-bed SATA disk and the test samples. The intra-drive file transfer tests the effectiveness of the buffer of each of the test samples, as the buffer synchronises the data transfer between drives or between two partitions on the same drive.

The inter-drive test results may confuse you-case in point: the scores of the Samsung ATA 133 drives. We need to remind you that the transfer speeds also depend on the test bed hard drive, which happens to be faster than the two Samsung ATA 133 drives. Hence the read speed was more than the write speed for these two drives. We would recommend you rely on the intra-drive speed to measure the performance of these drives. Exceptional performers in the inter-drive test were the Hitachi Deskstar 250 GB and the WD Caviar RE2.

As expected, Samsung's parallel ATA drives were beaten by a big margin in the intra-drive transfer test. The Samsung SP0411C, a SATA drive with a 2 MB cache, performed no better than the Samsung ATA drives! The WD Caviar RE2, Hitachi Deskstar 500 GB and Maxtor's Maxline III topped the charts with superior times in the intra-drive test.

The Photoshop image load test is one of the most important tests, especially for those using software such as Maya, 3D Studio and Photoshop. Photoshop's scratch disk, or the virtual RAM space for the application, was created on the test sample. Images of varying sizes (200 MB, 550 MB, 800 MB, 1 GB and 1.4 GB) were also stored on the test sample, so as to maintain the outcome of the test purely on the basis of the performance of the test sample. We used Photoshop to open these image files from the test hard disk and timed the process for each image.

The contest for the top slot was fierce, between the WD Caviar SE (WD2500KS) and the Hitachi Deskstar 500 GB, with the latter recording better timings. Other drives that performed well were the Maxtor MaxLine III and Hitachi Deskstar 250 GB. If you're looking for a performer at an affordable budget, then the WD2500KS is your best bet!

Jargon Buster 
NCQ (Native Command Queuing): By command queuing, the hard disk accepts multiple commands from the host controller and rearranges them into a completion order. The major portion of the drive's command service time is seek and rotational delay for the drive head.
NCQ is an advancement over LCQ (Legacy Command Queuing). To understand the importance of NCQ we need to know how LCQ functions.
Not all data on a hard disk is stored sequentially. Random data, or data that is part of a folder, may not be placed contiguously on the disk surface.
Imagine we have scattered data A, B, C, D and E on sectors 1, 5, 2, 4 and 3 respectively of the disk. Whenever this block of data is called by the processor, the disk will queue the data as A, B, C, D and E. So, the hard disk head will first approach sector 1 to take data A, then it will move to sector 5 to take data B, skipping data C, which is on the immediate sector. It only takes a few milliseconds for an extra rotation to read data C, but consider a hard disk full of such fragmented data. Thus hundreds (and probably thousands) of rotations are wasted, and this is how LCQ works.
The drive can use rotational optimisations to select the next command to complete, so that the major components of the service time-seek and rotational delay-are minimised. NCQ does this by enabling the hard disk to select the next command, based on what data is closest rotationally to the head's current position.
NCQ is native to SATA drives. However, some early SATA drives may not include this feature.
SATA (Serial Advanced Technology Attachment): Most of us know about the IDE cable, which is used to connect hard disks and optical devices such as a CD-ROM to the motherboard. SATA is no different from the perspective of this primary purpose. It is essentially a serially-communicating interface that not only works faster than the conventional parallel interface, it is also longer in length (up to 1.5 metres) and smaller in width. SATA was introduced with a speed of 150 MBps. Further research led to the development of SATA II, with a transfer rate of 300 MBps.
TLER (Time Limited Error Recovery): Most desktop hard drives are designed considering regular operation, assuming there is no RAID card. All desktop drives include read/write error correction, but they do not issue error messages or respond to commands by the RAID adaptor. If error correction takes time, the RAID controller drops the RAID volume as a non-responsive drive. This problem is most common in ATA drives. TLER was essentially enabled in Western Digital Caviar RE2, as this drive was designed considering RAID operation.
RAFF (Rotary Acceleration Feed Forward): This technology is designed for Western Digital hard drives-the Raptor and the Caviar RE2. The rpm of Raptor drives is very high (10,000 rpm), so rotational vibrations are unavoidable. By implementing RAFF, rotational vibrations can be cancelled to provide better and more accurate performance.
S.M.A.R.T. (Self Monitoring And Reporting Technology): Most hard disks now have SMART inbuilt. This is a protocol used to report the device status to the host machine. A SMART-enabled hard drive constantly monitors itself for any kind of discrepancy, and reports any problem so you can take corrective action. The primary corrective action is taking a backup of your data and then trying various disk utilities tools to check your disk.
Hot-Swap and Hot-Plug: 'Hot-Swap' and 'Hot-Plug' are sometimes used interchangeably. Hot-Swap means you can remove a drive and replace it with another without interrupting the system; an example where this would be used is in a mirrored storage system on a server.
Hot-Plug lets you plug in a new device while the system is running. Care should be taken while removing a device this way. A Hot-Plugged device should be detached only after using some kind of un-mounting option available on the host OS.
Value For Money
Rs 2,500 and Rs 4,000 are attractive figures, but have you thought about how much you're spending per GB of storage? It's a fair notion if you're a moderate user, but the actual price is not an ideal criterion if you demand performance at a fair price.

Back to the paperwork: all the drives have a remarkable warranty period-either three or five years-so the main aspect of discussion is the price index. Let's start with the Samsung hard drives, which we introduce again and again as the reference. The SP0411 and SV0411N belong to the category that most people buy. The price that one pays per GB for this category of hard disk is the highest-you can get double the capacity (80 GB) and much better performance for just an additional 1,000 rupees!

Maxtor's DiamondMax10, with a capacity of 300 GB and priced at Rs 8,000, proved to be the most cost-effective of all the drives, at Rs 28.62 per GB. On the other hand, WD's 74 GB Raptor priced at Rs 8,200 is the most expensive, at a whopping Rs 118.42 per GB.

Generally, we see that the cost per GB of higher capacity drives is much less than that of lower capacity ones, but the Seagate Barracuda (400GB) ripped through the price paradigm. It was not only costlier than its counterparts, the WD4000KD and WD4000YR, but was also more expensive than the Hitachi Deskstar 500 GB by a margin of Rs 3,500.

Considering the warranty, price index and performance, the Hitachi Deskstar 80 GB, Maxtor MaxLine III 300 GB, Western Digital WD2500KS 250 GB and Hitachi Deskstar 250 GB are the ones that make for value for money, in their respective categories. However, if you are hunting for a 160 GB drive, the Samsung HD160JJ is your best bet.

The Consensus
Based on our results from the Performance comparison and the overall analysis, we award the Best Buy Gold to the test sample that secured the highest overall scores.

The overall score is obtained by giving appropriate weightages to individual test segment scores. A higher weightage was given to the intra-drive test and the application load time test in the Performance segment, while additional features and shock resistance were the determinants in the Features segment. Hitachi's Deskstar 500GB recorded the highest performance scores amongst all the drives but it lost out on price index.

We have two Best Buy Gold winners, the Hitachi Deskstar 250 GB (HDT722525DLA380) and the WD 250 GB (WD2500KS). It was difficult to judge the better of the two as warranty, price and capacity were identical. While the Hitachi leads in the performance segment, the WD2500KS leads in  features.

The WD Caviar RE2 was close to Gold with better performance and features than the two Gold winners, but had to settle for Best Buy Silver due to its slightly higher cost per GB.

What Next?
Along with increased capacities, there has been a remarkable decrease in the dimensions of hard disks. The storage density of the platters has increased as well. There had to be a limit to the prevailing technology, and researchers had to think of new ways to develop a high-volume storage medium.

Thus, the need for higher capacity and ace performance is changing the way data is written onto a medium. Currently, data is written onto the disk surface by Longitudinal Recording (LR), which is reaching its limits (refer Towards Terabytes, Digit, October 2005). Disk manufacturers are now moving on to Perpendicular Recording, which promises much higher densities.

We are anxious to get our hands on this technology, and we're sure you will be looking forward to it as well!

Team DigitTeam Digit

All of us are better than one of us.