A universal truth since the beginning of computing has been that you can never get enough storage because you’ll always find something to fill it with. This is true more than ever, with the proliferation of high definition multimedia files, high resolution digital photography, graphically intensive games and even software defined radio. Space demands continue to rise, although hard disk capacity growth has slowed somewhat.
For a long period, we were saddled with only 2Tb and 3Tb drives, and only recently have 4Tb, 5Tb and 6Tb drives emerged. The larger 5Tb and 6Tb drives have an astronomical price-tag to match, and Hitachi even resorts to helium-filling the drives to make it a possibility. They are intended for use by system integrators for high end, data-center applications. It is clear that we are beginning to reach the limits of perpendicular recording technology – the technology that saved us from the 320-500Gb/drive brick-wall. We have been warned not to expect further capacity increases to fall in price as quickly, due to the technical challenges involved in even larger drives.
In all this time, sadly, the ports offered by most motherboards haven’t changed. We see a sum total of about four to eight SATA ports on most consumer motherboards, which effectively limits the number of drives you can have in a machine. Although it is possible to expand this by opting to install a PCI-E SATA controller, many of the cheaper options feature only two ports and are questionable for data integrity and performance, and the higher-end RAID cards cost a lot of money and require fiddling with SFF breakout cables. The only other option is a workstation motherboard, along with the workstation component costs, or to run multiple NASes on a network and live with the power consumption and performance implications.
Luckily, the price of 4Tb drives has now fallen to a point where they are price competitive on a $/Tb basis. Generally, the 4Tb drives have a price tag similar to the 2Tb drives, and slightly (5%) more expensive than the 3Tb drives. For any new system builder looking to maximise the storage available from their SATA ports, it is a reasonable premium to pay. For system upgraders, it gives them an incentive to consolidate the data from multiple drives to free up ports – for more high capacity hard drives or to add auxiliary SSDs.
This makes them quite relevant and interesting to system builders for bulk storage purposes, where the SSDs pick up the load for applications and system drive. Lets take a look at two of the main contenders, the Western Digital Green WD40EZRX and the Seagate Desktop ST4000DM000 which are the two main companies making hard drives, aside from Toshiba which are not as easily obtained.
Western Digital Green WD40EZRX
The first contender is the Western Digital Green WD40EZRX. This is part of their “green” series drives, optimized for cool and quiet operation for desktop applications. It features a SATA III 6Gbit/s connection and a 64Mb buffer. The spindle speed is quoted as “IntelliPower”, but is most likely to be 5400rpm. The number of platters is not specified.
The drive is supplied for OEM installation in a sealed anti-static bag. Unusually, it is plastered with a warranty label on the rear which clearly states that the warranty for the drive is two years.
The drive looks pretty similar to most recent WD drives, nothing too exciting about that.
Seagate Desktop ST4000DM000
The second contender is the Seagate Desktop ST4000DM000. This drive is just called a “Desktop HDD”, and features a 64Mb cache and SATA III 6Gbit/s connection. It is a 4-platter, 8-head configuration with a 5900rpm spindle speed (not in the data sheet). The drive itself is specified with a two year warranty, and alarmingly with a reliability spec claiming a 55Tb/year workload limit and 2,400 power on hours!
The drive was also packed in an anti-static bag for OEM integration, but with no large warranty label.
The drive itself looks like most modern Seagate Barracuda Green/LP drives.
What Happened to the Warranty?
The last time I bought a hard drive, the warranty was a lofty five years for Seagate and three years for WD. The almost co-incidental reduction of warranty between WD and Seagate to just two years seems rather suspicious to me and it’s surprising that I hadn’t heard anything about this before.
When IBM instigated duty cycle limitations on their drives for warranty, that resulted in a backlash. When most drive manufacturers tried to reduce their drive warranties to one year uniformly, that resulted in a backlash which resulted in Seagate offering five years.
But it seems this time, they’re doing it again, apparently to save on costs.
Does this mean that the drives are any less reliable? Possibly, but unlikely. The reason for this is that the drives’ share their mechanics and even their electronics between series – it’s likely that the drives warranted for AV usage and NAS with 24/7 duty cycle and longer warranties are the same drives with slightly different firmware optimization. This slight tweak to the firmware options and warranty are to enforce an artificial market segmentation and cause headaches for those who want to RAID “green” drives. Instead, they will need to buy NAS/Red series drives to have Time-Limited Error Recovery to prevent RAID array breakages in case of any access problems.
It’s a move by drive manufacturers to try and restore profitability to a highly competitive market, and personally, I find it disheartening that they would deliberately “cripple” their products to bare-minimum to compete for price, and sell essentially the same product un-crippled for a premium price, but that’s how the market works. That’s why we now have Black, Blue, Green, Purple, Re, Se, Xe, AV series drives on the WD side, and Desktop, Terascale (formerly Constellation), Surveillance, Video, Enterprise on the Seagate front.
As part of my drive-commissioning process, I like to validate my drives with my platforms to make sure they can store and retrieve data over the entire surface without corruption. Both drives completed a full random fill and three-pass checksum with no failures. While doing so, I also like to run a few benchmarks, so lets see how well the two drives perform.
Due to the loss of several computers due to various hardware failures, the test platform is an old Foxconn P35AX-S with Intel Q6600 CPU. Onboard Intel SATA II connectors were used to test the drives, which would not affect the test results significantly because the drives were not capable of sustained saturation of the SATA II link. The computer is running the latest version of Windows 7 with all patches applied.
To compare the drives, the WD result is on the left, with the Seagate result on the right.
Initial SMART Data
Both drives appeared healthy out of the box. Note that CrystalDiskInfo reports the WD Green having a 5400rpm spindle, and the Seagate Desktop having a 5900rpm spindle. At the conclusion of testing, both drives showed no significant changes in the SMART values, which implies no bad sectors were located and the drives were performing nominally.
Once formatted, you can see that both drives feature exactly the same formatted capacity. This is because the capacity points have been standardized, so it is possible to replace a drive in a RAID array with a competitor’s 4Tb drive and it will have the same number of LBAs. What is apparent is the difference between the binary Tb and the weasel Tb and how the difference grows. A 4Tb drive offering 3.63Tb of storage … might make a few people unhappy.
HDTune Sequential Read
Tests were performed using the Full Test option, which resulted in very long test times of ~8-10 hours.
The HDTune sequential read test shows the WD Green peaking at 144.2MB/s and averaging 112.8MB/s. The Seagate Desktop managed quite a bit better with a peak at 176.7MB/s with an average of 141.7MB/s. This is quite a big gap, and is also reflected in the access time, where the Seagate measured 15.7ms, and the Western Digital measured 17.2ms.
What is notable is that the curves are fairly “smooth” which implies that there are many more density zones on these drives, and the density switches quite a few times across the surface. This implies a sort of “desperation” to cram more bits into the drive. The outermost track of the Seagate drive seems to dip down in speed, which is unusual, but may be implying some possible reliability problems towards the outer surface.
Compared to the Toshiba 3Tb “Green” drives, however, it seems we have made a step backward in performance when considering increased areal density (over USB 3.0, they achieved 164.4MB/s maximum and 126.3MB/s average). Comparing to the Toshiba 3Tb 7200rpm desktop drives makes it clear just how much performance is sacrificed in the name of “green” (they achieved a peak of 190.4MB/s and an average of 151.1MB/s). Maybe the Toshiba 4Tb 7200rpm drive should be on your list if performance is your main concern?
HDTune Sequential Write
A very similar speed trend is seen in the write graphs, although the access time difference narrows somewhat.
HDTune Random Access Read
In the case of a random access read workload, the Seagate bests the WD in virtually every metric, delivering a higher number of I/O operations than the WD and while offering a lower worst case access time.
HDTune Random Access Write
When it comes to writes, the WD wins fairly consistently when it comes to the number of IOPs, but fails pretty badly for consistency at larger blocks with many “long delays” when it comes to making writes. Quite interesting to see the discrepancy here, so it’s not a complete victory for Seagate but pretty close.
Testing with CrystalDiskMark seems to show a narrower gap between the two drives, but the Seagate still bests the WD in all cases except in 4k read where WD wins both for single and 32-queue depth.
Looking at the results from ATTO seems to show the Seagate hitting its full performance by the time we hit 8kB accesses, whereas the WD doesn’t do so until about 16kB accesses. The Seagate bests the WD in pretty much all cases.
While H2testW is normally used to test USB memory keys, you can also use it to test hard drives. No failures were detected in both cases, but you can see the WD takes almost 10 hours to do a single fill or readback, whereas the Seagate only takes about 8 hours. The performance difference amounts to a two hour difference for a full read! This is why performance might outweigh the claimed energy savings, and energy efficiency isn’t a straightforward thing to measure.
What about the Energy?
When the whole “green drive” idea became fashionable, I was quite dubious of the idea because of the performance hit. It was a good thing in terms of reducing temperature and power supply requirements, however, but it seems that buying a green drive to save money might be a fallacy depending on your workload. In the case where the performance loss of the drive results in a longer run-time, the increased use of energy of the whole computer could easily outweigh the small savings in the drive.
Unfortunately, as I don’t have any dual-rail measurement gear, I couldn’t get power readings from both drives. However, if we take the values from the specification sheets:
Power Consumption (W) Idle R/W Sleep WD 4Tb Green 3.3 4.5 0.4 Seagate Desktop 4Tb 5.0 7.5 0.75
So, it seems that the WD drive is quite a bit lower in power consumption. Lets do some math to calculate the energy required to read the whole drive:
4.5w * 9h 50m 53s / 3600 = 44.31625Wh
7.5w * 7h 51m 25s / 3600 = 58.927 Wh
In this workload, it seems the WD is more energy efficient for the drive alone.
But what’s the difference? It’s 14.61Wh. Since the WD takes 1.991 hours longer to do the workload, if it keeps a computer which consumes just 7.337w on for 1.991hours, then the energy would have broken even.
But that’s the problem, the rest of the computer consumes a lot more than 7.337w. Now can you see why energy efficiency claims are far from straightforward?
Lets take a different case where your drives are idling away for 5 years, 24/7. What’s the difference in energy, and how much does that cost? It’s a bit more simple arithmetic.
(5w – 3.3w) * 365 days * 5 years * 24 hours / 1000
= 74.46kWh per drive
At a cost of AU$0.28/kWh for electricity, the hit to your wallet is AU$20.85 over the life of the drive. You probably earn more than this per hour. If you had to wait for your slower drive for an hour (over the course of the review, it has been several hours), then you’d have been financially losing. False economy, possibly, although it depends highly on your workload.
In short, I really don’t think the power consumption differences should be a primary factor in your purchasing decision, however, any energy saved is a positive for the environment, but it’s not clear that opting for the slower, more miserly drive is a win overall especially for heavy workloads.
Given the higher expected areal density of the 4Tb drives, one would expect higher performance. While compared to other green drives, it seems to be fairly similar, the performance is poor compared to non-green drives. The performance was quantified, and the Seagate takes a pretty convincing victory, likely because of the increased spindle speed of 5900rpm vs 5400rpm.