Fatter Platters (Internal Hard Drive).

By Team Digit Published Date
01 - Aug - 2006
| Last Updated
01 - Aug - 2006
Fatter Platters (Internal Hard Drive).
Quick: what's the slowest device in your system? Chances are you'd say "my CPU," or "my DVD-ROM." If we were to enquire as to the most important device in your PC, chances are you'd say it's the CPU.

Your hard disk is by far the slowest of all your PC components and yet perhaps the most important device inside your PC. Simple reason: your data resides on your hard drive. It's irreplaceable. Warranties mean nothing: what's worth more-a new, empty hard disk, or all the data you'd hoarded and all the programs you'd installed?

Now, not only are hard disks faster and bigger-much bigger-than they were a couple of years ago, prices have fallen so drastically, gigabytes almost cost peanuts. So what do you do with all that storage? Well, with our games going multi-gigabyte and entire movies on our disks, it seems no amount of storage space will ever satiate us.

Remember it's us, the users, and our usage patterns that dictate terms to hard disk manufacturers, not the other way around. We wanted speed, we needed storage, and craved reliability, didn't we? Well, how does that age-old adage go-"be careful what you ask fo  r"?

We received a lot of Serial ATA drives for this test, all except one of these were SATA 2 compliant. We also received a few Parallel ATA drives. Consider ing the price difference between them, if your motherboard supports Serial ATA, simply go for it.

The SATA drives from Seagate, Samsung and Western Digital had pins that could be jumpered to force them to operate in SATA 1 mode (150 MBps). This is useful in case your Serial ATA controller is SATA 1; operating without the jumper could cause compatibility issues. The WD drives came with SATA 2 support; no jumpers. The Hitachi series deserves special mention for the hassle involved with enabling SATA 2 mode. They're capable of 300 MBps operation, but come with 150 MBps mode enabled. Enabling SATA 2 was a real pain; it's done by downloading a utility from Hitachi's Web site that needs to be installed on a floppy (which becomes bootable). Alternatively, an ISO image file (in case you don't have a floppy drive handy, like us), can also be used.

The problem was, our downloaded ISO turned out to be corrupt. In fact, it took three downloads and a writer than could burn at incredibly low speeds to get the ISO working. All's well that ends well and the performance difference was worth the effort for us. A relatively new user is likely to have a lot of problems getting this right. Kudos to the other three manufacturers for designing a simple workaround.

Some of  the SATA drives from WD and Seagate featured a 16 MB cache, which makes these drives more capable- at least theoretically-at dealing with large file transfers. Among the others only the Hitachi 500 GB SATA had a 16 MB buffer. All others had 8 MB buffers, which we feel is more than sufficient for most usage needs.

Most of the newer SATA drives do not have the older 4-pin power connectors, rather they have the newer SATA connectors. Backward compatibility can be a problem with older SMPS's that don't have these connectors. Adapters are thankfully available in abundance, in case your power supply doesn't provide these.

Hitachi 500GB

Seagate has the distinction of having sent in the largest drive in this shootout, their latest 7200.10 series Barracuda, with a capacity of 750 GB. This drive features Perpendicular Recording technology, which allows cramming a previously unheard-of 188 GB on each platter (the drive uses four platters). This drive performed better than all the other drives in many of the tests, but whether this is due to the technology used, or a variety of other factors, it's just too early to tell. Judgment withheld until more drives built around Perpendicular Recording technology emerge, at which time we can do a comparison, pitting a good number of such drives against older drives.

HD Tach
We fired up the HD Tach 8 MB file tests first (followed by the 32 MB ones). While the burst rate scores aren't really significant, we noted that three of the Seagate SATA drives topped the burst transfer rate test. This shows that the Seagate drives are making the most of the additional theoretical bandwidth that SATA 2 provides.

Only four drives cracked the magic 60 MBps in the average read tests: the massive 750 GB Barracuda scored 67 MBps, followed by the 400 and 500 GB WD SATA siblings with scores of 64 and 62.4 respectively. Among the 250 GB drives, the Samsung SP2504C was the topper, with another 60 score. The scores in both the HD Tach tests (that is, the 8 MB and 32 MB file samples), were nearly identical. A couple of drives from Samsung scored low here, with the PATA 40 GB being the only sub-50 MBps drive.

The 32 MB tests saw the same foursome taking the top four spots. The Seagate Barracuda 750 GB showed an inexplicable drop in the Average Write part of the 8 MB tests, showing that maybe-just maybe-a drive so large isn't as suitable for writing lots of relatively smaller files.

CPU utilisation isn't a very significant figure today, simply because 4 and 5 percent of CPU resources is a drop of water in a rainstorm for today's multi-gigahertz and dual core processors.

 How We Tested
For our internal drive shootout, we used an AM2-based system, powered by a 2.6 GHz 5000 processor, 2 GB of DDR2 800 MHz memory in dual-channel configuration, and an XFX 7900GTX graphics card. The Hitachi T7K 250 GB SATA 2 HDD was used as our test bench hard disk. The hard disks to be tested were kept clean from any unnecessary software installs, except for the test files. All drivers and necessary software were installed on the test hard drive.  
All the hard disks were set as master drives (except the SATA drives, which have individual channels). SATA 2 drives were run at the maximum transfer mode, that is, 3 Gbps, while the SATA 1 drives were run at 1.5 Gbps. Two partitions each of 50 GB were made on each of our test drives, except for the smaller drives, where the primary partition was 30 GB, with the remaining size devoted to the second partition.
We installed Windows XP Professional with Service Pack 2, with all the latest drivers installed.
Our synthetic benchmarks consisted of HD Tach and SiSoft Sandra 2007 Engineer. HD Tach is a very trusted low-level benchmark that needs a completely fresh, that is, un-formatted, un-partitioned  hard drive to run. The scores it returns are very close to the actual raw speed that a hard drive can attain, simply because it doesn't access the hard disk through Windows and various other software layers, which slow down the drive and adversely affect scores. Because of this uniqueness of HD Tach, we fired it up first.
We ran both 8 MB and 32 MB file sample tests. The Burst Speed figure is a good indicator of what the bus is capable of, but it loses significance as a real-world figure. This is because it represents a "burst" or instantaneous value. The Average Read and Write figures reflect the real-world performance the drive is capable of, as these are the sustainable data transfer rates.
SiSoft Sandra 2007 is another very popular benchmark that tests various facets of a drive's performance. SiSoft Sandra gives a Drive Index figure-a combined average of the Read and Write tests, which are divided into buffered, random, and sequential read and write tests. Besides these tests, SiSoft Sandra also reports the access time. 
For our real-world tests, we used a single 4 GB .rar file to test sequential read and write performance. For random read/write operations we used a combination of more than 25 different file types. The files were also purposely placed in a multitude of folders, and the size of the files was also greatly varied. There was a Word document occupying exactly 25 KB of disk space, while an AVI file was 400 MB in size! The total file size was 4 GB, the same as that of the single .rar file. The variance in size is a good way to test how a drive performs during real-life data transfers, involving lots of small files, as in copying multiple MP3 files from one drive to another.
For our File Write tests, we copied from the test bed drive to the test hard drives. For the File Read tests, we simply copied the test file back to the test bed drive. Our internal file test was the most strenuous test (as the figures show), where we copied the test files from one partition on the test drive to the other partition.
We installed a fresh copy of Adobe Photoshop CS 2 on our test drives. The scratch disk was also kept on the same partition as the install. The test PSD files were also placed on the same partition, a true test of the mettle of today's drives.
Finally, for our game benchmark, we installed a fresh copy of FarCry with patch 1.32 installed, on the drive to be tested. The first map was loaded, and the time taken to load was noted. We used Map 1, simply because it starts with a cutscene that isn't very graphics-intensive, as opposed to other levels, which start with gameplay: this causes a slight pause, because it takes the graphics card a little longer to render the scene. We wanted the graphics card to play as small a part as possible, so we noted the time just after the starting of the cutscene cinematic.
SiSoft Sandra 2007
SiSoft Sandra's latest version gives some nicely detailed scores for individual hard drives. All SiSoft's scores, except the access time, are measured in megabytes per second.

SiSoft Sandra's Drive Index is an average of a drive's scores throughout the gamut of tests it runs, and the highest scorers were once again from the Seagate and Western Digital stables (in that order)-namely the 750 GB 7200.10 Seagate and the WD4000KS. In the 250 GB and below category we saw the Samsung SP2504C lead yet again.

Hitachi 250 GB

Besides the drive index, SiSoft Sandra also gives figures for Read and Write operations. These are divided into sequential and random figures. Sequential figures measure the drive's ability to read and write data sequentially, that is, to adjacent areas of the disk. This test largely emulates the performance one can expect from a drive while working with large files. Sequential operations are usually done much quicker, and the scores will reflect this.

The Random Read and Write scores indicate how well a drive scores when working with scattered data, such as small or discontinuous files. In these tests the 2 heavyweights from Western Digital (the WD4000KS and the WD5000KS), alongwith with the 750 GB Barracuda posted some of the hightest read scores we've come across. In the smaller drive group, the Samsung SP2504C once again came out on top.

Among the PATA drives, yet another Samsung drive (for the wrong reasons this time), the 40 GB SP0411N came in last, with the WD1200BB joining it in the under-50 scores for the Drive Index. This is understandable, since the Parallel ATA interface isn't (these days) exactly known for speed. The 500 GB and 250 GB Hitachi drives did quite well.

Real-World Tests
We were more interested in the scores the hard drives would put up in the real-world tests, because these represent actual performance through data transfers both through the bus and through the internal drive buffer.

The file read and file write tests check the I/O performance of a hard drive, as well as its ability to properly utilise the bandwidth that the bus (whether SATA or PATA) provides. A drive that performs better at reading data (whether sequential or random) is more suitable for storing data files such as MP3s and AVIs, that is, non-installable files. This is because these types of files are written only once, but may be read several times. You listen to your music daily, but don't necessarily keep copying it to different locations on your drive partitions, do you?

A drive that excels at writing operations (to find drives suitable for each type of operation, check our benchmarks in the table) is more suitable to use for installing Windows and other program files. This is because the OS and programs are constantly performing write operations in their respective folders. Such drives are also very suitable for creating swap files (which are some of the most write-intensive areas of your entire disk).

Samsung 250 GB

The Internal File Transfer involves an intra-drive transfer, where we copy our test files from one partition to another within the same drive. Very suitable to test the performance of a hard drive under stress, simply because data is being read and written simultaneously on the same drive, without the interface between drives playing any significant part. The performance of a hard disk is the sum total of its drive mechanics (including the spindle speed and platter areal density) and I/O buffer. The drive is actually stressed as it has to perform a read and write operation simultaneously, not to mention that the buffer is also being doubly stressed.

The Assorted Internal Read Test was completely dominated by two drives, the Seagate 7200.10 750 GB behemoth, and the slightly smaller (and slower by a hairsbreadth) WD5000KS 500 GB. The only other drive that came anywhere close (read that as a distant third) was the Samsung SP2504C. But this drive was in the smaller capacity group. In all honesty, we were quite disappointed with the Hitachi Deskstar drives, particularly the 7K and T7K series, although no slowcoaches they be; a tad more performance we were hoping to see (if you'll pardon the horrible rhyme!). Among the PATA drives, the WD800BB from Western Digital stood out: it churned out some truly stellar scores, definitely on par with the much faster SATA drives. If you do not have Serial ATA support on your motherboard, this is the hard disk for you!

The single 4 GB .rar file test was once again owned inside-out by the 750 GB heavyweight from Seagate, followed, albeit by a large margin, by its 250 GB sibling. The other drive that was quite impressive here was the 500 GB Hitachi SATA, with a time of 157.6 seconds in the Internal singlefile transfer test.The WD 120 and 160 GB PATA models brought up the rear in this test. If you look closely at the table, you'll see the fastest drive is nearly three times faster than the slowest drive. While we cannot in all fairness compare the latest SATA drive with much older PATA models, this proves that hard drives on the whole are becoming faster with every new release. Each revision adds performance and subtracts from the price (well, most of the time, anyway).

Seagate 250 GB

Next, we ran the gamut of file and application loading benchmarks. Two drives from Seagate really shone while loading Photoshop-the quickest was the Seagate Barracuda 7200.9 series 250 GB (7.3 seconds), followed by the 750 GB (8.1 seconds). The Samsung SP2504C was also very good here, scoring within a second of the leaders.

The mammoth 1 GB Photoshop .PSD file test was interesting because it saw a new leader emerge, a hard drive that had earlier escaped the limelight. The 500 GB Hitachi SATA (the Deskstar HDS725050KLA360) simply blew away all but two drives, the Western Digital siblings, the WD4000KS and WD5000KS. All these three broke the sub 1 minute barrier.

The slowest drive in this test was once again a PATA drive, Samsung's SP0411N, and was, surprisingly, joined by a SATA sibling (the HD160JJ).

Throughout our tests, we also noted which drives ran hotter than the others, as is regular practice. We do not note these scores unless we come across something highly irregular or unusual. It's interesting to note that the blazing-fast Seagate drives also ran the hottest throughout the tests, with the SATA drives in particular generating quite a bit of heat. Temperatures remained well within check however-not a serious problem-but it does add to the heat circulating inside your cabinet, especially when you have multiple drives. The WD drives seemed the coolest in general, with the Samsung and Hitachi drives falling in between.

Phone E-Mail Website
HitachiCyberstar Infocom Ltd 9341057327 lalit.sudrik@cyberstarindia.com www.hgst.com
SamsungSamsung India Electronics P. Ltd 011-41511234 jayakumar.v@samsung.com www.samsunghdd.com
SeagateeSys Distribution Pvt Ltd 011-4181-1694 msinghal@esysmail.com www.seagate.com
Western DigitalVenktron Digital Systems P. Ltd 022-23874427 sales@mediamangroup.comwww.westerndigital.com 

Purchase Decisions
Getting yourself a new hard disk can be a daunting task, especially with all the options out there, both brand-wise and size-wise. As mentioned earlier, Serial ATA is an established technology now, with very tangible benefits over Parallel ATA. Though SATA drives are slightly more expensive than their PATA counterparts, the extra cash spent is well worth the performance gains. SATA cables are also much sleeker; PATA cables are notorious airflow restrictors and dust magnets.

Western Digital 400 FB

Instead of going the traditional "more cash, bigger drive" way, we've tried to distinguish between users on the basis of what they actually need, and not what they can afford. After all, it's your requirements and not your wallet that defines you as a user! We've given you a convenient little box full of goodies (in the form of buying decisions) to make that choice just a little easier!

And The Winners Are…

Serial ATA (250 GB And Below)
In this, the category comprising the smaller SATA drives (250 GB and smaller), we were most impressed by the performance of the 250 GB Samsung SP2504C. It's also very competitively priced at Rs 4,200, and we deem it worthy of the Digit Best Buy Gold award. The Digit Best Buy Silver award goes to the Seagate ST3250824AS, another 250 GB drive. This drive lost out on the Gold because it's just a little more expensive, and also because it trailed the SP2504C in many of the tests by just a bit.

  • RAID 
RAID, or Redundant Array of Independent Disks, is, quite simply, a setup that utilises more than one hard disk to either share or replicate data over hard disks. There are several different ways of implementing RAID and the type of RAID adopted dictates the relative merits and demerits. Motherboards today come with integrated SATA RAID controllers, and feature multiple versions of RAID.
We used one of our fastest drives, the 750 GB Seagate Barracuda, and for the RAID array, we used two Seagate Barracuda 7200.8 series 400 GB drives to show you the performance differences. Although RAID has many avatars, most home users need to know about only three of them.

Also called Striping, this setup is for the speed fanatic. Striping doesn't have any sort of fault tolerance-a failure in any one of the striped drives will cause data loss. That said, drives of today are much less prone to crashes than their predecessors, and for pure speed, RAID 0 rules! Gamers and hardcore enthusiasts swear by their RAID 0 arrays.
RAID 0 needs at least two drives: data is broken down into blocks, and each block of data is written alternately to a separate hard drive. The load is, therefore, spread across a number of channels and drives. Another advantage of RAID 0 is that no parity calculation is involved, since there is no duplicating of data anywhere-so CPU overhead is minimised. RAID 0 is highly recommended for gamers, 2D and 3D professionals, and anyone who wants greater read/write performance from his storage subsystem.

Also called "Mirroring," RAID 1 offers data security via fault tolerance. It needs a minimum of two drives to implement: during a write operation in RAID 1, the same data is written to each of the drives simultaneously, so each drive contains a mirror of the data on the other drives. A read operation can be performed from any of these drives, so if one crashes, data can be accessed from the other.
The only disadvantage is that I/O performance is slower than with a single drive, and the parity computations-and therefore the CPU overheads-are much higher than with a single drive or even with striping. RAID 1 is very useful for users for whom data integrity means everything.

RAID 0 1
A combination of striping and mirroring, this features two or more clusters of RAID 0 arrays collectively configured as a mirrored array. RAID 0 1 requires a minimum of four hard drives to implement. Since this mode contains multiple striped segments, the I/O performance is blistering, while the presence of mirroring adds the much-needed parity to the mix. RAID 0 1 isn't without it's demerits: it's very costly to implement, and besides, the failure of a single drive will in essence create a single, glorified RAID 0 array, and thereby defeat the purpose of parity completely.
Although implementing RAID seems like a costly affair, there are many users who may have a number of hard disks installed in their PCs, or even a couple of older hard disks lying around. With the advancements in today's onboard RAID controllers, it's no longer necessary to have drives of the same capacity, although "RAIDing" unequal-sized drives means you will lose the extra space the larger drive offers. Those with two or more similar-sized hard disks available are perfect candidates for experimenting with RAIDing. Remember, though: make a backup of all your essential data, as setting up a RAID array will destroy any data present on all the disks being used.

Serial ATA (Above 250 GB)
The drive that most impressed throughout all our tests was the Seagate Barracuda 7200.10 750 GB. However, it actually belongs in a class of its own for two reasons-first, it's much larger than the other drives, and second, it uses the Perpendicular Recording technology. So we are not giving it the winners title, even though the scores indicate otherwise.

Western Digital gets all the accolades in this  category: the WD4000KS (400 GB) receives the Digit Best Buy Gold, and the WD5000KS (500 GB), the Digit Best Buy Silver.

Parallel ATA
In case it's got to be PATA for you, look no further than the Hitachi siblings, the Deskstar HDT722525DLAT80 and the Deskstar HDS725050KLAT80 (250 GB and 500 GB respectively). Bear in mind that the 500 GB Hitachi-our Gold award winner-maybe the best of the PATA drives, performance-wise, it is expensive at Rs 20,000.

Western Digital 500 GB

Go for it only if you simply must have the largest and fastest PATA drive out there. For just as much bang without a steep drop in your bank balance, go with our Silver winner-the smaller Hitachi 250 GB, which, at Rs 5,700, offers you the best of both worlds.

Our Conclusion
We tested 26 hard drives this time round. Seven of these were the older PATA drives. With the majority of drives going SATA, it's the end of the road for PATA. All motherboards sold these days sport SATA interfaces, and with the possibility of CD/DVD drives going SATA, this seems fitting.

Much like how SATA debuted some years ago, Perpendicular Recording makes its way into the drive market, with the promise of larger density drives, equating to greater storage that you won't pay a premium for.

With Serial ATA promising as much as 600 MBps in the near future, the storage horizons seem promising indeed. As there is much good that has come before us, let there be more of the same!

Team DigitTeam Digit

All of us are better than one of us.