Rumors about a missing hardware link between the Sega Saturn and Dreamcast have swirled mysteriously for many years. Now, thanks to a series of interviews published by Japan’s Beep21 gaming magazine (machine translation), the truth is out. An ex-Hitachi, ex-Sega, and ex-Sony engineer has confirmed that he proposed and developed a Saturn graphics accelerator, codenamed TRIP, based on Hitachi’s SH‑3 processor.

The Sega Saturn console launched in late 1994 / early 1995, as Sega’s fifth-generation machine, going up against Sony’s first-generation PlayStation. Sega became concerned about leaks pointing to great PlayStation (PS One, PSX) performance. So, a few months before the launch, it reconfigured its Saturn with an extra Video Display Processor (VDP). Thus, the machine ended up with ‘processor soup’ made up of dual Hitachi SH-2 (28 MHz) RISC processors, two VDPs, plus co-processors dedicated to 3D geometry, sound, and I/O.

With its previous-gen console, the Genesis (AKA Mega Drive), Sega didn’t hold back from iterating with the Mega-CD, 32X, Sega Channel network, and more. But rumors about an accelerator for the Saturn came and went with no hardware ever released, nor revealed, until now.

TRIP accelerator with SH-3 processor

The missing link between the Saturn (1994) and Dreamcast (1998) was an accelerator for the former, internally known as the TRIP. Beep21 heard about this elusive project from an engineer called Junichi Naoi.

According to the source story and interview, Naoi was a Hitachi engineer who was in the team that developed the SH-1, SH-2, and other CPUs and ASICs. For the SH-2, he implemented several Sega-requested enhancements into the design. However, in 1994, Naoi would join Sega and help the firm develop cost-optimized revisions of the console over its lifecycle.

In 1996, a couple of years after Saturn first hit the market, and a couple of years before the Dreamcast would splash down, Sega felt that its current console was losing its shine. Not only was the PlayStation 1 doing amazingly well, but 3D accelerators were starting to establish PCs as very capable modern gaming machines.

Thus, the TRIP project was born at Sega. The proposed accelerator module for the Saturn would include a Hitachi SH-3. This chip was expected to significantly boost graphics and geometry calculations on the Saturn. According to the report/interview, the TRIP was judged to be good enough to propose that future Dreamcast classics Virtua Fighter 3 and Shenmue be launched alongside it.

Development work on TRIP and those launch titles went ahead, but the project was abruptly cancelled. We don’t know much about why, yet. Reasons could have included development wrinkles, costs, timing, or a combination of such issues. Nevertheless, a prototype TRIP board was made, and Shenmue assets were rebuilt for it before it was culled.

The engineer would Sega, and corporate knowledge of TRIP evaporate

Naoi went on to join Sega’s arcade development division, but left for Sony to contribute to the PS3 era. As one of the few contributors to the short-lived, quietly cancelled TRIP project, knowledge of the Saturn accelerator among fellow Sega alumni basically evaporated.

Sega’s Dreamcast arrived in 1998 with yet another generation of the Super Hitachi (SH) chip, the SH-4. Overall, the new console's architecture was far simpler, with this single fast (400 MHz) RISC CPU, and an NEC PowerVR2 GPU, augmented by a Yamaha sound chip.

Losing essential data to a hard drive failure can be one of the most frustrating experiences of your life. But, there might be a way to identify if a drive will actually fail or if it will last forever. Secure Data Recovery, according to Blocks and Files, discovered the average failed hard drive lasted around two years and ten months before dying.

This number was taken from the average operation time of numerous failed hard drives that the data recovery specialists received from customers. The specialists excluded drives that failed from non-predictable events and only kept drives that failed due to normal read and write operations. The targeted drives were made by six manufacturers, including Western Digital, Hitachi, Seagate, Toshiba, Samsung, and Maxtor.

Secure Data Recovery also found that the average expected lifespan of a failed hard drive highly depended on the drive manufacturer. According to a power-on-hour chart it created, there is almost a 50% difference in power-on time between the best and worst-performing brands.

Toshiba performed the best, with an average of 34,799 power-on hours recorded between 151 bad drives. Conversely, Hitatchi - a WD sub-brand, was far worse, with an average recorded lifespan of 18,632 hours accumulated from 211 bad drives.

Results from the four other drive manufacturers sit in between Toshiba and Hitachi, with Western Digital coming out with 25,676 power on hours on average from 936 disks. Seagate amassed 32,298 hours from 559 drives. Samsung accumulated 19,224 hours based on 123 drives, and Maxtor registered 29,771 hours based on 27 drives.

Secure Data Recovery also recorded the number of bad sectors on each drive. Again, Hitachi was by far the worst-performing manufacturer, with 3,348 bad sectors across the same 211 drives. But the best-performing drives in this category come from Maxtor, with just 228 bad sectors found. Thankfully, the rest of the six manufacturers were not as bad as Hitachi, with Toshiba turning up with 1,884 bad sectors in total, Samsung with 529, Seagate with 2,671, and Western Digital with 628.

Despite the significant variance in power-on hours and sector errors between drive manufacturers, we can't take these results at face value. The results would be far more accurate if Secure Data Recover had an even distribution of failed drives from all six manufacturers.

As a result, lower disk counts of Hitachi, Toshiba, Samsung, and especially Maxtor have to be taken with a grain of salt. Seagate and Western Digital, on the other hand, have the most recorded drives, so their numbers will be the most accurate. But overall, it does seem like Hitachi and Seagate are the most unreliable manufacturers of them all. 

If you have a hard drive from any of these six manufacturers, the average life expectancy is just under three years if your drive has a propensity for failure. However, if the drive doesn't die during that period, there's a good chance it will remain healthy for years and years to come. If you want to know more about HDD reliability, check out our previous coverage from Backblaze.

Now that even cell phones have the graphical oomph to render things in gorgeous 1080p, the 4k leap will be the next big milestone. Tom’s has spoken quite about it before, particularly what it’ll take to render all the pretty modern games with almost 8 million pixels, and while it’s a hefty drain on system resources even for a desktop, the tripartite Japan Display partnership is already trying to cram all that detail into a tablet.

Engadget reports that the coalition between Sony, Toshiba and Hitachi have announced that they have a fully functional 12.1-nch 4k TFT package. It’s a stunning achievement, and while one might think it’d be a huge drain on battery power and need a killer set of hardware to push all that additional detail, the Japan Display group maintains the module is power-efficient and fairly thin, given the real complexity of it.

The full report is linked here, and it’s not too hard to make sense of with the help of Google. Some other things to note, the viewing angle and contrast ration are a fair bit lower than what is standard these days, especially for portable electronics, sitting at 160 degrees and 1000:1 respectively. Curiously, the screen brightness is also remarkably high at 500 nits. Crunching the numbers for the power draw of the panel itself and looking at standard battery sizes results in something in the neighborhood of twelve hours, though that’s only for the display.

If this comes to market within the next year, I’d be shocked, but it’s definitely something that’s fun to think about as we toss around bigger and bigger numbers and look to up the prettiness factor on the electronics we use every day.

Follow us @tomshardware, on Facebook and on Google+.

HGST's 1 TB Mobile Drive, Spinning At 7200 RPM

One-terabyte notebook hard drives aren't exactly new. You can already find the Hitachi Travelstar 5K1000, Toshiba's MQ01ABD100, the same company's MK1059GSM, and Samsung's Spinpoint M8 HN-M101MBB. However, those four 1 TB disks share one thing in common: they trade in their speed for higher capacity. None of them spin faster than 5400 RPM. In contrast, smaller 750 GB drives can be found in both 5400 and 7200 RPM flavors.

HGST (formerly Hitachi Global Storage Technologies), now a subsidiary of Western Digital, set out to change this with its Travelstar 7K1000 HTS721010A9E630, a 1 TB notebook-oriented drive sporting two platters with 500 GB each, and rotating at 7200 RPM. It features a 16 MB data cache and a 6 Gb/s SATA interface. The manufacturer specifies a wide range of applications for the Travelstar 7K1000: notebooks, desktop PCs, compact gaming PCs, video devices, and mobile storage.

Combining high-density platters with a fast spindle, we're eager to see if the Travelstar 7K1000 can snatch the performance crown away from Seagate's Momentus XT, a 750 GB, 7200 RPM drive. The HGST drive steps into the ring with an immediate advantage, though: its price. You'll find the HTS721010A9E630 for as little as $80 or so online, while the smaller hybrid hard drive goes for about $100. A 1 TB version of Seagate's Momentus XT, now referred to as the Laptop SSHD, comes armed with 64 MB of data cache, 8 GB of NAND flash, and a $120 price tag on Newegg. We don't have our hands on that one yet, but are looking forward to running some more hybrid hard drive-friendly benchmarks once its lands so we can properly compare the performance of both technologies.

So, is a lower price the HGST drive's only leg up, or does its performance rival the mechanical competition, too?

Technical Data And Test Configuration

Technical Data

Test Configuration

Results: Throughput And Interface Bandwidth

Right out of the gate, HGST's Travelstar 7K1000 HTS721010A9E630 takes the lead in h2benchw with the highest sequential read and write rates of 102.9 MB/s (average). The 750 GB Hitachi Travelstar 7K750 and WD Scorpio Black WD7500BPKT follow close behind at 98.1 and 98.0 MB/s, respectively.

The other 1 TB drives trail behind those three top models. The pack of mid-range drives is led by Toshiba's MQ01ABD100, which achieves 88.9 MB/s.

The Travelstar 7K1000 also emerges victorious in Iometer, achieving almost the same numbers seen in h2benchw. While the streaming read benchmark merely confirms the three fastest models, the sequential write test surprises us: WD's Scorpio Black WD7500BPKT comes in last.

In our interface bandwidth test, which is really just a measure of how fast data can be moved from the hard drive's cache through the SATA connection, HGST's Travelstar 7K1000 HTS721010A9E630 and Hitachi's Travelstar 5K1000 both achieve about 400 MB/s. This is a reflection of its 6 Gb/s SATA support. Samsung's Spinpoint M8 HN-M101MBB is ranked third, but only achieves 244.7 MB/s due to its 3 Gb/s interface.

Results: Access Time And I/O Performance

Not surprisingly, the notebook drives with faster spindles tend to lead the access time chart. There are exceptions, though, like WD's Scorpio Blue WD10JPVT, which comes in fifth place. The newer Travelstar 7K1000 HTS721010A9E630 takes the top spot, boasting 14.99 ms and 15.69 ms read and write accesses, respectively. The 7200 RPM Seagate Momentus and WD Scorpio Black WD7500BPKT are close behind.

Knowing full well that mechanical hard drives (much less those optimized for mobile applications) are unsuitable for I/O-intensive workloads, particularly in a world of affordable SSDs, we still run four stock Iometer profiles to determine the I/O throughput of each disk.

Spinning at 7200 RPM only slightly helps the Travelstar 7K1000 in our quartet of benchmarks, which simulate access patterns for databases, file servers, Web servers, and workstations. The drive consistently achieves a spot in the top third of our field, ranging from fourth to sixth place. Seagate's Momentus and Western Digital's Scorpio Black WD7500BPKT score highest.

Results: 4 KB Random Reads And Writes

As we established on the previous page, the I/O performance of hard drives is hopelessly inferior to SSDs. This also applies to purely 4 KB chunks of data written and read randomly. But hard drive vendors shouldn't be using that as a reason to stop improving the random access times of their platter-based devices. After all, these small transfers continue to be a major component of everyday storage.

Once again, the Travelstar 7K1000 HTS721010A9E630 does well, though not well enough to win. It scores in the top third of the field. With that said, the performance differences between the 7200 RPM drives are actually pretty small.

Results: PCMark 7

The synthetic benchmark PCMark 7 places HGST's Travelstar 7K1000 HTS721010A9E630 in second place, behind the winning Seagate Momentus XT, which leads by almost 1000 points.

HGST's new drive only wins in a few components of this benchmark suite: video post-processing with Windows Movie Maker, and Windows Media Center performance. The Scorpio Black WD7500BPKT and Hitachi Travelstar 7K750 perform similarly to the Travelstar 7K1000.

Results: Power Draw And Efficiency

There's nothing particularly exciting to report from our power consumption numbers. If we look at the raw results on their own, we see a fairly normal range.

The most power-friendly drives are predominantly the 5400 RPM disks. Naturally, a slower motor draws uses less energy. The Travelstar 7K1000 HTS721010A9E630 does not surprise us, winding up in the middle of the field, with one exception. During HD video playback under VLC, the Travelstar 7K1000 draws a mere 0.8 W, which is the lowest reading among all drives.

Benchmarks: Efficiency

When we calculate efficiency, we put each drive's power consumption in relation to performance parameters like IOPS or streaming write throughput. The clear winner in the IOPS per watt discipline is Western Digital's Scorpio Blue WD10JPVT. HGST's newer Travelstar 7K1000 HTS721010A9E630 doesn't fare particularly badly, but it's also not outstanding. The picture changes when we look at streaming write efficiency. In that metric, the drive takes a first-place crown. No other disk achieves higher write performance per watt.

A Well-Balanced Drive We Can't Wait To Compare To Seagate's SSHD

By adding the Travelstar 7K1000 HTS721010A9E630 to its model range, HGST delivers a very fast 2.5" drive, which helps address the perpetual problem facing mobile user with just one bay for storage: do you drop in a smaller, pricier SSD, or stick with a conventional disk? By no means is this 1 TB repository a stand-in for flash. However, its 7200 RPM spindle at least brings down access times compared to the field of 5400 RPM, 1TB contenders. Of course, it also helps that this large repository is selling in the $80 range.

What are the Travelstar 7K1000's specific strengths? Based on what we saw in a field of 14 hard drives, the disk achieved the highest throughput, the lowest access time, and a very high interface bandwidth, which is only matched by its older sibling, Hitachi's Travelstar 5K1000. Given that it's aimed at general-purpose applications, the drive isn't as adept in, say, database applications or specifically low-power environments. Our benchmarks reflect this. Although it isn't a slouch in the 4 GB random read or write tests, it doesn't stand out either. Nor does it overwhelm us with its power consumption numbers.

Based on the standard benchmark suite we ran, HGST's Travelstar 7K1000 earns a thumbs-up. If you're in the market for a 2.5" disk that's balanced to take capacity, performance, and price into equal consideration, this one seems like a perfect candidate.

We'd say that the ball is on Seagate's and Toshiba's court to come up with something to go up against this 1 TB, 7200 RPM notebook-oriented hard drive. However, you're probably wondering about the new 1 TB Laptop SSHD from Seagate, formerly referred to as the Momentus XT, right? Well, the newest model sports a 64 MB data cache and 8 GB of NAND flash. But it spins at 5400 RPM. So, in applications where data isn't yet cached, the Travelstar is likely faster. Information stored in the solid-state space should be gobs faster than HGST's mechanical disk, though. Although the 1 TB model currently sells for about $40 more than the 7K1000 we reviewed today, the addition of flash could make the Laptop SSHD worthwhile, and we look forward to finding out.

A Tropical Predator Retires; Farewell Barracuda

It’s time to bid farewell to Seagate’s Barracuda. The branding, that is. Not the actual 3.5" desktop hard drive family, which will henceforth be known as Desktop HDD with .15, .14, and .12 generational suffixes. 

Seagate is looking to make a splash with its flagship in the newly-named line-up, the Desktop HDD 4TB, also known by the ST4000DM000 model number. Up until now, Seagate's desktop portfolio didn't stretch as high as 4 TB. Only the business-oriented Constellation ES.3 line included a 4 TB repository. But now that Seagate is playing in the super-high-capacity space, it currently enjoys a commanding price-per-gigabyte advantage. The ST4000DM000 is going for as little as $180 on Newegg right now. That's four and a half cents per gig.

The Desktop HDD 4TB (ST4000DM000) spreads its multiple terabytes across four platters and spins at 5900 RPM. The 1 TB and 3 TB Barracuda 7200.14 drives employ the same platter size, but instead spin at 7200 RPM. As a result, we expect the new Desktop HDD.15 to be somewhat slower, and Seagate's technical specifications reflect this. The company says its Barracuda drives are capable of up to 210 MB/s, whereas the Desktop HDD.15 maxes out at 180 MB/s.

Aside from the extra capacity and slower spindle, nothing else really changes from the Barracuda 7200.14 to the Desktop HDD.15. The drives still sport a SATA 6Gb/s interface, feature 64 MB of data cache once you hit the 1 TB model, and are built for 300,000 load/unload cycles. Our benchmarks should give you a good indication of how the Desktop HDD.15 ST4000DM000 performs, and how it stacks up to the Barracuda 7200.14.

Technical Specifications And Benchmark System

Technical Specifications

Benchmark System

Results: Data Throughput And Interface Bandwidth

As we expected, Seagate's Desktop HDD.15 ST4000DM000 isn’t as fast as the Barracuda 7200.14. The difference isn’t really that large, though, and an average 132.2 MB/s sequential read speed and 131.5 MB/s sequential write speed is enough to put the new drive in second place for both benchmarks. This also means the 4 TB Desktop HDD beats several 7200 RPM competitors, like Western Digital's Black WD4001FAEX and the Hitachi Deskstar 7K4000.

The Desktop HDD.15 ST4000DM000 can't quite hit the 180 MB/s transfer rate that Seagate promises in our h2benchw benchmark. Its 167 MB/s are still very respectable, though.

Of course, because we're using the same SATA 6Gb/s interface as the one used on the Barracuda drives, we end up with similar performance across the wire (even though this is primarily a measure of the drive's cache).

Results: 4 KB Random Read And Write, Access Time, And I/O Performance

Random reads and writes are not strengths of mechanical storage. For exciting numbers in these metrics, you want an SSD.

But the Desktop HDD.15 fares particularly poorly compared to the rest of the field due to its slower spindle. The 7200 RPM drives establish a solid advantage in this benchmark.

The access times posted by Seagate's Desktop HDD.15 aren't as good as what we see from the Barracuda 7200.14. This is a result of the 4 TB drive's 5900 RPM motor. Our measurements aren't bad, though, registering 17.7 ms for reads and 19.89 ms for writes.

The Iometer benchmark patterns for common database, file server, Web server, and workstation workloads aren’t the Desktop HDD.15’s strong suits, either. Seagate's 4 TB disk places third or fourth to last in all four tests, whereas the Barracuda 7200.14 consistently makes it into the top three.

Results: PCMark 7

The Desktop HDD.15 ST4000DM000’s PCMark 7 performance is thoroughly mediocre. Its performance is alright, but it never manages to break out of the lower-middle field. Two Hitachi hard drives, the 7K3000 and 7K4000, win this one.

Results: Temperature, Power Consumption, And Efficiency

After the Desktop HDD.15’s strong, but certainly not impressive performance in our read and write tests, the 4 TB disk finally gets to face a few tests where it can excel.

For starters, its operating temperature is only 32 degrees Celsius, which means it's not even warm to the touch. Western Digital's Red WD30EFRX is even cooler at 31 degrees Celsius. The 7200 RPM Seagate Barracuda has an operating temperature of 37 degrees Celsius, which is still well within an acceptable range.

Power Consumption

The Desktop HDD.15 and Western Digital Red WD30EFRX do battle for the honor of turning in the lowest power consumption results for a 3.5" hard drive. Even though the WD drive barely bests Seagate in the end, a 3.6 W measurement at idle and 5.9 W reading at peak write speed are both excellent numbers.

Efficiency

Dividing power consumption by read and write performance gives us an efficiency result for each disk. This is a good way to figure out which usage scenario the drives are best-suited to.

Seagate's Desktop HDD.15 ST4000DM000 plays in the big leagues once again, and it’s only beaten by the Western Digital Red WD30EFRX when we calculate the efficiency in streaming writes and database operations.

Seagate's 4 TB Desktop HDD.15 Is Great For User Storage

Seagate’s new 3.5” Desktop HDD.15 ST4000DM000 changes things up a bit in the company's desktop hard drive family compared to its predecessor, the Barracuda 7200.14. Most notably, the spindle speed is down from 7200 to 5900 RPM. The Desktop HDD.15 is the first and only model in this line that operates at this lower speed.

The company is aiming its new drive at the same applications and usage scenarios as the previous generation's Barracuda disks: desktop and all-in-one PCs, home servers, gaming PCs, RAID arrays, external direct-attached storage devices (DAS), and network-attached storage appliances (NAS). You're of course free to use Seagate's Desktop HDD.15 in any of those environments, though our benchmarks show that it fares better in some applications than in others.

A slower spindle immediately casts some doubt on the Desktop HDD.15's viability as a system drive. And once we run our tests on the disk, the potential for less performance is confirmed as fact. The test results come back average at  best, particularly in transfer speeds and I/O throughput. Conversely, Seagate's Desktop HDD.15 shines when we look at surface temperatures and power consumption, both of which are low.

The Desktop HDD.15 is Seagate’s first desktop drive boasting a massive 4 TB capacity. When combine that big storage space with conservative thermals and efficiency-friendly power use, you get an ideal target for user data on the desktop. Hook it up next to a 128 or 256 GB SSD and enjoy the best of both worlds. It's also right at home in external direct-attached storage devices (DAS) and NAS appliances.

The company formerly known as Hitachi Global Storage Technologies, HGST, said that it has broken the 20-nm barrier associated with platter-based storage by combining self-assembling molecules and nanoimprinting. The result is the creation of large areas of dense patterns of magnetic islands only 10-nm wide. These islands are only about 50 atoms wide, and some 100,000 times thinner than a human hair.

HGST, now a Western Digital subsidiary, said on Thursday that the bit density of its new 10-nm pattern is double that of today's disk drives, and currently shows excellent initial read/write and data retention in tests. Once expanded to an entire disk, the nanoimprinting process is expected to create more than a trillion discrete magnetic islands.

"Self-assembling molecules use hybrid polymers, called block copolymers, composed of segments that repel each other," the company explained. "Coated as a thin film on a properly prepared surface, the segments line up into perfect rows.  The size of the polymer segments determines the row spacing."

Once the polymer patterns are created, a chip-industry process called "line doubling" makes the tiny features even smaller, creating two separate lines where one existed before, the company said. The patterns are then converted into templates for nanoimprinting, a precision stamping process that transfers the nanometer-scale pattern onto a chip or disk substrate.

"A key challenge proved to be preparing the original surface so the block copolymers form their patterns in the radial and circular paths necessary for rotating disk storage," the company added. "HGST is the first to combine self-assembling molecules, line doubling and nanoimprinting to make rectangular features as small as 10 nanometers in such a circular arrangement."

HGST's new nanolithography achievements arrives at a critical time for HDDs as solid state drives become more prominent in the PC sector. Unfortunately, they come with high cost for consumers despite their speed, especially in larger capacities. As it stands now, the ideal setup is to have a small SSD for the OS, and a large HDD for media storage and other data.

Mechanical drives, on the other hand, continue to be the cheaper route while also offering tons of space for a growing amount of stored user content like movies, TV shows and music. On a cloud-based level, this achievement could have a huge impact, allowing companies to store more data both internally and externally. Cloud-based storage services could boom, or the could bomb, depending on HDD consumer-based prices.

"We made our ultra-small features without using any conventional photolithography," said HGST's Tom Albrecht said. "With the proper chemistry and surface preparations, we believe this work is extendible to ever-smaller dimensions."

HGST is aiming to have the process ready for wide-scale commercial production by the end of the current decade.

_{Contact Us for News Tips, Corrections and Feedback}

Mini-ITX Gaming: Small, Fast, And Inexpensive

We’ve seen how AMD’s Llano-based APUs stack up against Intel’s Sandy Bridge-based Pentiums in Better with Time? The A8-3870 And Pentium G630, One Year Later. That story generated quite a big of feedback, much of it asking how Intel's Ivy Bridge architecture might fare. Today, we’re putting together a new build with an Ivy Bridge-based Pentium at its core.

Of course, simply building an entry-level gaming PC on a budget is a pretty tired topic, so we chose to tackle a more formidable challenge. Could we fit a budget-oriented configuration inside a mini-ITX chassis? Would it still accommodate an optical drive for us to install all of our favorite titles? Might there be room for the hard drive needed to house those games? Perhaps most important, is there room in a cheap mini-ITX case for a discrete graphics card able to deliver smooth, stutter-free frame rates? Surely, we couldn't expect something so specific to also look good, right?

Falcon Northwest showed us what a boutique builder can do with months of R&D and aspirations of supporting high-end hardware in Meet The Tiki: Core i7-3770K And GeForce GTX 680 In A Mini-ITX Box? This isn't the same thing though, our goal here is to tackle small, attractive, and inexpensive. Although that seems almost impossible, we promise you it's doable. You just need to track down the right parts. A high degree of manual dexterity helps, too.

Finding A Good Deal On A Mini-ITX Case And Power Supply

I lost a lot of hair trying to find the right mini-ITX chassis and power supply. There simply isn't much out there to choose from, much less with a bundled PSU around the $60 price range. Our power supply choices were between the TFX form factor and a picoPSU, so we had to choose between output and size. We calculated that we'd need no less than 120 W, which is actually quite a lot for a picoPSU, especially given the limited selection in that product segment. If we went that route, our choices would have cost about $140 for a case, the picoPSU, and a notebook power brick. Too expensive, we decided.

What remained were cases with bundled PSUs. Generally, they lack the level of quality we're willing to accept, they're larger than what we want, or they come with older, much less efficient power supplies. After a mission of online shopping and calling around to various vendors, we finally discovered a gem of an enclosure featuring an integrated TFX power supply and selling for about $60. Could it be the chassis we were looking for all along?

Good Looks: A Case And PSU

The Deal/Steal: Chieftec FI-01W

This case looks great, whether it's lying flat in a desktop orientation or standing up as a tower, especially given its low price tag.

When this story went live in Europe, our enclosure of choice was called the FI-01W. But Chieftec doesn't sell it here. In the U.S., the same case shows up as Winsis' WI-10. It's really hard to track down, though. So, although all of our pictures are of the Chieftec chassis, we're using Winsis' WI-02 instead. Internally, the WI-02 is identical. It even includes a 200 W power supply. For as little as $45 online, we're happy to recommend the WI-02 as an alternative.

Like other Chieftec enclosures, the FI-01W (and, by extension, Winsis' WI-02) is built like a tank, albeit a tiny one. This can be a blessing as well as a curse. Be ready for a twisted neck and scraped fingers during the build process because this chassis has absolutely no give and promises only cramped working conditions.

Let’s first take a look at how the Chieftec FI-01W stacks up on paper.

This is where we would have nitpicked the the Chieftec FI-01W’s smaller problems, except that we didn’t find any.

The only thing that we didn’t love about the FI-01W was its lack of a door for the optical drive. While the case comes with a removable cover, a simple door mechanism would have made the case perfect (though we recognize that a quality mechanism would have also made it more expensive). At the equivalent of $60 in Europe, or as low as $45 here in the U.S., we're fine with the chassis as-is.

We'll dive into power supply performance once we finish the build.

A Reasonable Price: CPU And Motherboard

Mini-ITX and Intel: When The Price Is Right

We decided to build this rig using an Intel Ivy Bridge-based CPU. Keeping cost in mind, a 3.2 GHz dual-core Pentium G2120 emerged as our only real option.

As long as two cores are enough, the Pentium offers great performance. And while AMD sells quad-core parts for about the same price, they don't fare as well when it comes to cooling and power consumption, both of which are important considerations in such a compact space. This time around, it just had to be the little 55 W Ivy Bridge-based Pentium.

The Right Motherboard: MSI B75IA-E33

MSI’s B75-based B75IA-E33 is a newer board that offers everything we could want from a mini-ITX platform in this price range. Although the form factor limits any motherboard's available features quite a bit, MSI manages to give its B75IA-E33 a logical and usable layout. Even the UEFI menus are arranged nicely. We found this to be a perfect fit for both the build and budget we had in mind.

While there aren't many bundled accessories to speak of, MSI's B75IA-E33 includes all the essentials. The unexpectedly large manual is easy to navigate, though the text labels in the schematics could have been larger.

Cooling: Third-Party Or Intel's Bundled Heat Sink?

Inexpensive Cooling Without A Whisper

The noise generated by Intel's bundled heat sink and fan combination is a matter of personal taste. Some folks can live with it, while others can't. Count me, personally, in that second group. I just had to buy a third-party cooling solution. The $30 Xigmatek Praeton presents an inexpensive step up from Intel's stock hardware.

Almost all mini-ITX motherboards have their audio connectors along the bottom. The corresponding cable in Chieftec’s case is so short that it wouldn't have reached if Xigmatek's cooler were any taller. An addition quarter-inch of cable length would have helped, but a significantly longer cable is what this form factor really needs. 

In the picture above, the you can see the HD Audio header cable stretched diagonally across the CPU fan. It doesn’t actually touch the blades, so it never becomes a functional problem. It's just ugly.

Finger Exercises: The Hard Drive And SSD

A Removable Drive Cage Simplifies Installation

The drive cage is easy to pull out of the chassis without necessitating that you work in the confines of a 6.45 L box. That's good news for those of us with not-so-nimble fingers.

The plastic faceplate is removable as well. However, be careful to bend the plastic clips gently, or they may end up in the trash and not back in the case.

Middle: Joint

Right: Small Lever

Installing the 3.5” Drive

You simply slide the 3.5” hard drive into place, and then fasten it with a quick-release mechanism. This makes installation easy. However, it doesn't decouple the disk from the chassis in any way.

Fortunately, even though our storage device wasn't isolated, we couldn't hear the 300 GB Velociraptor spinning away. The fact that Chieftec's enclosure is built like a piece of WWII armor, and not like an aluminum can, really helps mitigate vibration.

Installing the 2.5” Drive

The SSD is attached to the chassis by good old-fashioned screws, and getting to them is no problem with the removable drive cage.

SSD boot drive? Check. Hard drive for user data? Check. Now it’s time to deal with that slim optical drive.

Flat Like A Pancake: The Slim Optical Drive

Installing the Slim Optical Drive

Let’s face it, no gaming PC is totally complete without an optical drive. Our mini-ITX-based build is no exception.

The installation is easy enough, but unless your drive is one of the few with a full-size SATA connector, you'll need to buy one additional part in order to get it working.

A Necessary Part: Slimline to SATA Adapter

If you want our advice, don’t buy an expensive slim optical drive with full-size SATA connectors. Instead, grab a cheap drive and a slimline-to-SATA adapter, which you should be able to find for less than $10 online.

You might run into a problem where the adapter sticks out too far and runs into the CPU cooler. Fortunately, the Praeton's fan can be moved enough to accommodate the adapter's extra length.

Brace yourselves, the motherboard installation is next. We recommend that you install the board first, then connect SATA cables to the hard drive and SSD, and finally, reinsert the drive cage. Assembling the case in any other order almost guarantees scraped fingers.

Tight Spaces: The Motherboard Installation

Wanted: Nimble Fingers

Again, installing the motherboard really isn't difficult, so long as you do it before putting the drive cage back in.

As always, remember to snap in the bundled I/O shield before screwing the motherboard down. That’s probably obvious to the average Tom’s Hardware reader, but still important to mention. We've even been caught up in a fancy build and missed this little detail.

The I/O panel snaps right into its designated opening. Sadly, this deserves mention because such a good fit doesn’t happen nearly as often as it should.

Fasten the motherboard to the tray using four pre-installed spacers.

Packing everything that needs to go into the case makes for an extremely tight fit. The optical drive’s slimline-to-SATA adapter does push into the CPU fan, but as we mentioned, the cooler's retention brackets are flexible enough to keep this from being a problem.

Before we move on to the graphics card, we have a couple more pieces of advice regarding the motherboard installation.

First, plugging in the front-panel connectors can be tricky, so we recommend connecting them to the motherboard before placing it inside the case. Otherwise, breaking out tweezers is going to be your only realistic option.

And while cable routing usually comes down to personal preference, we had to compromise a clean-looking build for secure connections when it came to dealing with components like the optical drive's adapter and the HD Audio cable. It probably goes without saying in a mini-ITX PC, but there's really not a lot of empty space left inside the case at this point.

Pushing Pixels: Sapphire's Radeon HD 7750

Leaving Intel’s HD Graphics In The Dust

We've run plenty of benchmarks that show Intel's HD Graphics engine simply won't cut it for gaming. The Pentium G2120 does its job much better paired to a discrete graphics card.

Chieftec's case makes selecting a graphics card easy, since it only leaves room for one model: a low-profile, single-slot Radeon HD 7750. Sapphire, PowerColor, and Club 3D all sell versions of the card. But since we already have Sapphire's in the lab, our choice is easy yet again.

Because Intel's Pentium G2120 and Sapphire's Radeon HD 7750 are in roughly the same performance class, neither is likely to bottleneck the other, making them a solid combination. Overall, you should see good performance right up until the platform reaches its limit. Then performance will fall apart completely.

The graphics card installation mechanism is simple, but effective: a lever holds the card in place. Not like there's much room for it to move anyway. In fact, the graphics card fan is almost exactly on top of the power supply fan. This is not the most elegant solution we’ve encountered, but it does give the card just enough space to breathe.

We covered the Radeon HD 7750's performance at length in AMD Radeon HD 7770 And 7750 Review: Familiar Speed, Less Power. And while it might not be the very fastest card available, Catalyst driver updates have made it substantially faster. It’s a good alternative to the older Radeon HD 6770, which is still found in many mainstream machines. Indeed, the Radeon HD 7750 is more than enough for gaming on an HDTV, and it sure beats the graphics power offered by today’s gaming consoles.

Now, let's get back to the Chieftec case's included power supply.

Power Consumption

Modest Numbers: PSU and Power Consumption

We calculated a total system power consumption for this build at about 120 W, and we weren’t too far off. The finished system ends up drawing a maximum of 112 W from the wall socket.

A closer look at the 12 V rail reveals that our build’s power consumption falls almost completely into the PSU's optimal range. This conservative power draw is in spite of the absence of an 80 PLUS certification. For the sake of comparison, we also tried an 80 PLUS Gold-rated 450 W supply. It drew six watts more at idle and two watts less at full load. Certificate or not, this system’s standby power consumption is less than 0.5 W.

Our build's 90 W maximum power consumption under normal gaming conditions is in a class of its own. We haven’t been able to even approach this number using an AMD APU paired with discrete graphics. That also means Intel’s solution requires less cooling, which is certainly a boon in such a cramped form factor.

Our decision to use MSI's B75IA-E33 motherboard turned out to be a really good one for low power consumption. We also benchmarked a competing Z77 Express-based board and found it drawing nine watts more. With Eco mode turned on, our system's power consumption remains just north of 30 W at idle, despite its mechanical hard drive, 8 GB of memory, and discrete graphics card.

Small Package, Reasonable Price, And Good Performance

It was fun building this svelte system worth around $500, and it performs really well given our space and cost restrictions. Chieftec's FI-01W proved to be a great little enclosure priced reasonably...in Europe. Here in the U.S., Winsis' WI-02 does the job just as nicely. Around $45 for a respectable-looking case and a decent power supply should spark some interest.

We really like the quick-release mechanism for the drives, and appreciate the ease of working with a removable drive cage. Six liters of internal volume aren’t a lot compared to the ATX enclosures most of us are accustomed to. But that's all we needed to attach cables, connectors, and drives securely (and without skinning any of our fingers). Heavy-duty build quality is also nice; there’s no bending the Chieftec chassis out of shape. Really, the only low point for us is the front-panel audio connector, which should have been a couple of inches longer. Any other critique is simply endemic of the mini-ITX form factor, or the result of paring-back that needs to be done on a product in this price range.

Compared to some of the quad-core chips from AMD, Intel's Ivy Bridge-based dual-core Pentium seems a bit expensive. But it combines low power consumption with consistently fast performance, and we needed both of those. MSI's B75IA-E33 motherboard similarly lends modest power use and an incredibly efficient layout.

If this story has you thinking about building a similar system, we can’t blame you. The finished product performs really well.

Want to go for it? Really? Well, here's a list of the components we used in today's project. For a few bucks more, you could even swap out the 1 TB hard drive for a fast 128 GB SSD, upgrade the cooling, or incorporate Blu-ray playback.

Western Digital's Performance-Oriented 4 TB Hard Drive

Western Digital recently added a new model to its Black family of performance-oriented hard drives. The new disk, technically referred to as WD4001FAEX, sports 4 TB of space. It spins at 7,200 RPM, employs a 6 Gb/s SATA interface, and includes 64 MB of cache.

Although the company won't divulge how many platters this behemoth uses to hit its capacity ceiling, our benchmark/power data suggests the magic number is five. You can’t tell by simply looking at it; the drive is 26.1 mm-thick, just like many other 3 and 4 TB repositories.

As a general rule, the more capacity a hard drive hosts, the longer it takes to benchmark. Western Digital's WD4001FAEX is certainly no exception. It took us 19 hours of testing to get a complete picture of its performance. Fortunately, the company kept this in mind when it sent out samples, providing us with some much-needed sugar. It's too bad that you don't get chocolate as part of the disk's standard bundle. When you spend $330, a bare 4 TB hard drive is all you get.

Technical Specifications

Benchmark System

Benchmark Results: Data Throughput And Interface Bandwidth

Western Digital's WD4001FAEX and Hitachi's Deskstar 7K4000 are very similar performers. This shows in the sequential read and write benchmarks, where they are locked in a close race for second and third place. The Hitachi drive posts somewhat higher maximums, but its average speeds are very close to those of the Western Digital Black. In the real world, we'd like to think these two would be indistinguishable.

With that said, no drive is able to come close to Seagate's Barracuda, and its read and write speed averages that exceed 150 MB/s.

The WD4001FAEX's read and write speed always stays above 80 MB/s. The drive assures good, consistent data throughput.

Benchmark Results: 4 KB Random Reads/Writes And Streaming Reads/Writes

The random read/write performance of hard drives is usually dictated largely by spindle speed. However, our Iometer benchmarks demonstrate a handful of exceptions that blur the line between 7,200 and 5,400 RPM disks. The WD4001FAEX turns out to be one of the slower 7,200 RPM models. Only Seagate's Barracuda fares worse.

Of course, when you compare these results to those on the previous page, it's clear that Seagate is optimizing most heavily for sequential transfers. Western Digital does as well, to a less degree. The outcome is a compromise in random performance. We think that's a smart trade-off, considering conventional drives aren't well-suited to random I/O anyway, and it's probable that an enthusiast using a 7,200 RPM disk with 4 TB of capacity would probably also have an SSD under the hood handling more performance-sensitive duties.

The WD4001FAEX does moderately well in our streaming read and write tests. Its performance is neither really good nor really bad. Seagate's Barracuda leads our line-up by a significant margin, clearly outmaneuvering the competition.

Benchmark Results: Seek Times And I/O Performance

Western Digital's WD4001FAEX is one of the faster drives in our read seek time benchmark. At 16.18 milliseconds, it places within a tight cluster of the six fastest drives separated by just 1.5 milliseconds.

Write seek time is one of the Black drive's strong suits. It takes a second-place finish at 6.95 milliseconds.

As we might have hypothesized after seeing the previous page's 4 KB random read/write benchmark results, the WD4001FAEX is a middle-of-the-pack performer in our database, file server, Web server, and workstation workloads.

Of course, in a world with SSDs in it, 20 IOPS up or down means very little. The folks running applications dependent on I/O throughput already know solid-state storage is the way to go. So, a lackluster showing in these tests is hardly what we're going to agonize over.

Benchmark Results: PCMark Application Performance

In our Windows-based application benchmarks, Western Digital's WD4001FAEX shows up right where it left off on the previous page: in the middle of the field. A couple of the sub-tests do show it in a better-than-average light, but the overall Storage suite score puts it under two Hitachi and two Seagate disks.

Temperatures And Power Consumption

The 4 TB Black drive brings up the rear when it comes to measuring temperatures, posting 41 degrees Celsius and tying Hitachi's Deskstar 7K4000 as the warmest disk in our line-up. Then again, 41 degrees is still acceptable, particularly in a desktop machine with plenty of airflow.

In contrast, though, Western Digital's WD30EFRX (from the Red family) manages a frosty 31 degrees, which is why it's marketed as the right drive for network-attached storage appliances, which are usually loaded with other drives also generating heat.

If you're looking specifically for low power consumption, then the WD4001FAEX is certainly not the way to go. Western Digital doesn't even try to hide the fact that its five-platter 7,200 RPM flagship is a bit of a power hog. The company's data sheet lists an idle power number of 8.1 W, which is actually worse than the 7.9 W we recorded.

The 4 TB Black model shows us 8.1 W of power use during HD video playback and 9.3 W during a maximum write speed benchmark, doing nothing to redeem the high idle consumption result.

As a result, when we look at performance per watt, Western Digital's new 4 TB monster falls to the bottom of our chart.

Western Digital's WD4001FAEX: Probably Not For Home Use

Sometimes, benchmark results frame a product as decidedly average. After looking at all of our numbers, Western Digital's WD4001FAEX certainly seems to qualify as just another 4 TB drive; nothing particularly noteworthy. But it's important to consider the Black drive's intended usage before jumping to a final decision on it.

We've already seen that the WD4001FAEX's performance is generally middle-of-the-road. That's not necessarily a bad thing, though. It simply means no one attribute stands out compared to the other 3 and 4 TB drives out there. The Black drive manages to place third in our desktop performance suite, which is weighted most heavily to throughput, but also includes PCMark and I/O results. Only Seagate's Barracuda and Hitachi's Deskstar 7K4000 finish higher, and only one of those two drives is a 4 TB model.

So, who should buy Western Digital's WD4001FAEX? The company says its drive is meant for PC enthusiasts and professionals. But we can’t quite agree because Seagate's Barracuda is, quite simply, faster.

However, Western Digital’s five-year warranty coverage tells us a slightly different story. Priced at $330, the 4 TB Black drive is a good deal for companies that want to put as many terabytes as possible into a server for data storage (rather than high accessibility). It should work well in that type of environment, in spite of fairly high operating temperatures and power consumption. If you're a home user with a five- or seven-bay NAS, we'd be inclined to stick with Western Digital's Red family instead.

While we were running benchmarks on Western Digital's WD4001FAEX and eating the chocolate in our review kit, we started thinking about what it'd take to make a hard drive more exciting in 2013. A combination of 4 TB on four platters, for example, would probably be better for power consumption (fewer platters to spin), temperatures (less friction), and performance (higher areal density), endearing itself to home use. Or maybe Western Digital could make a 4 TB version of its Red family for networked appliances?

Even more interesting would be a big leap in technology that'd enable a 4 TB drive using three platters. Will it happen? Some day, almost assuredly. Only time and the hard drive industry can tell us when. It has been a long time since we've seen anything to get excited about in the hard drive space, though, especially when it comes to per-platter capacity.

FX-8350 Brings AMD Back To The Table

System Builder Marathon, December 2012: The Articles

Here are links to each of the four articles in this quarter’s System Builder Marathon (we’ll update them as each story is published). And remember, these systems are all being given away at the end of the marathon.

To enter the giveaway, please fill out this SurveyGizmo form, and be sure to read the complete rules before entering!

Day 1: The $500 Gaming PC
Day 2: The $1,000 Enthusiast PC
Day 3: The $2,000 Performance PC
Day 4: Performance And Value, Dissected

Introduction

I read through Chris' analysis of AMD's Vishera-based FX-8350 very carefully (AMD FX-8350 Review: Does Piledriver Fix Bulldozer's Flaws?) and was happy to see that the company at least had a viable alternative to Intel's Core i5-3570K on its hands.

Naturally, I wanted to know how an FX-based System Builder Marathon machine would compare to the box I built last quarter, which housed...a Core i5-3570K. So, I bought similar components this time around, except for the platform, and set off to figure out how our new benchmark suite would treat the competing architectures.

We know the FX-8350's performance can be inconsistent compared to the Core i5 due to its modular design. However, we've run enough tests to know that lightly-threaded workloads generally favor Intel, while more demanding applications go AMD's way. Although the Piledriver design does help the FX in games, we're expecting Intel to retain its advantage in that discipline.

Since we ordered the parts for this build, some prices are up and others are down. Fortunately, the current $1,009 price tag is very close to the $1,000 target. Keep in mind that the previous configuration went $57 over budget, and the graphics card alone is down $50 since last quarter.

CPU, Motherboard, And Cooler

CPU: AMD FX-8350

We all know that AMD recently updated its FX processor family with a new architecture called Piledriver. Improvements to the design allow the company to increase the operating frequency of its fastest models, while ducking in under the same thermal ceilings. AMD also launched its Piledriver-based chips at lower price points than the previous-generation CPUs.

Although those evolutionary changes aren't significant enough to challenge Intel's highest-end processors, they do pick a fight with the mid-range Core i5s, particularly in heavily-threaded tests. And it just so happens that we're adding more demanding benchmarks this quarter, too.

Read Customer Reviews of AMD's FX-8350

This is the first processor from AMD we've used in a mid-range build for a long time, and we look forward to seeing what it can do. Currently, the chip sells for $220, which is 10% more than AMD says it should sell for.

Motherboard: Gigabyte GA-970A-D3

Read Customer Reviews of Gigabyte's GA-970A-D3

If we want to keep this build under budget, we can't afford to spend too much on a motherboard. The good news is that Gigabyte's GA-970A-D3 is feature-complete enough for us, and we've heard it does fairly well in overclocking. We also have to like the fact that it's selling for $70 on Newegg.

This board's only limiting factor is a single PCIe x16 slot. That second connector is limited to x4 signaling. But, since we're only using one graphics card today anyway, we're not going to worry too much about slot layout.  

CPU Cooler: Xigmatek Loki SD963

Read Customer Reviews of Xigmatek's Loki SD963

Xigmatek’s Loki has proven itself as a great budget-oriented cooler, and we'd like to see if it can handle the 125 W FX-8350 as well as it cools Intel's 77 W Core i5-3570K.

Selling for $25, there's not a lot of worthy competition for this model.

Video Card, Power Supply, And Case

Video Card: Gigabyte GeForce GTX 670 GV-N670OC-2GD

We're giving Gigabyte's GV-N670OC-2GD another try, since we used it in last quarter's System Builder Marathon and our goal is to compare platform performance without altering other variables. Of course, no two graphics cards overclock in exactly the same way, so we're hoping to squeeze similar frequencies out of this card when it comes to pushing our new build to its limit.

Read Customer Reviews of Gigabyte's GV-N670OC-2GD

Gigabyte's GV-N670OC-2GD earned our Recommended Buy award in Seven GeForce GTX 670 Cards, Benchmarked And Reviewed as a result of its quiet operation and excellent performance. The good news is that the price of this GeForce GTX 670 card dropped from $400 to $350 since our last build.

Power Supply: Corsair CX600 V2

Read Customer Reviews of Corsair's CX600 V2

Corsair’s CX600 is a great power supply; we discovered that last quarter when we used it for this same machine.

A low price should be indicative of fairly entry-level features. It's not modular, and it's limited to a pair of six/eight-pin power cables for our GeForce GTX 670 graphics card. Nevertheless, it should prove ample for our needs. It also fits nicely in our budget at $70.

PC Case: HEC Blitz

Read Customer Reviews of HEC's Blitz

At $50, HEC's Blitz is not exactly the lowest-budget case we could have purchased. Though, in our defense, its price is higher now than when we ordered it.

Even still, it's a roomy enclosure that comes with two 120 mm fans. We're expecting it to get the job done. No, it's not particularly drop-dead gorgeous, but it's not brash or juvenile, either.

Memory, Hard Drives, And Optical Drive

Memory: Mushkin Enhanced Blackline 8 GB (2 x 4 GB) DDR3-1600 Memory Kit

A lot of what we experiment with comes from your feedback, and comments in the forums suggest that the Tom's Hardware audience considers 8 GB of memory the minimum for a $1,000 build. You ask, we deliver. 

For a scant $50, Mushkin's 8 GB DDR3-1600 memory kit gives us the capacity you want, the data rate our processor can use, and low-enough timings to make us happy. Once again, this is the same RAM we used in last quarter's SBM, so we should be able to draw a fair comparison with it.

Read Customer Reviews of Mushkin's Enhanced Blackline 8 GB DDR3-1600 Kit

SSD: OCZ Agility 3 AGT-25SAT3-60G 60 GB

Read Customer Reviews of OCZ's Agility 3 60 GB SSD

Same story here. We chose OCZ's Agility 3 last time, so we're using it again this quarter.

This small boot drive is fast enough to accelerate the responsiveness of Windows and a couple of applications. We turn around and install games and benchmarks on a larger mechanical hard disk. At $65, it's a pretty cheap add-on, even though we're really starting to eye those 128 GB SSDs in the $100 range for the future.

Hard Drive: Hitachi GST Deskstar 7K1000.C 1 TB

Read Customer Reviews of Hitachis GST's Deskstar 1 TB Hard Drive

Hitachi's Deskstar offers 1 TB of storage, a 7200 RPM rotational speed, and 32 MB of data cache for $90. Although it only employs a 3 Gb/s SATA interface, you'll never see anything even close to a bottleneck from that specification. So, we're plenty satisfied with this drive as secondary storage for user data.

Optical Drive: LG GH24NS90 DVD-ROM

Read Customer Reviews of LG's GH24NS90

Optical drive prices seem to fluctuate frequently on Newegg. While LG's GH24NS90 was the cheapest model when we placed our order, the price is now at $19. Still, that's not bad, and our choice should serve as a solid solution for disc-based storage.

System Assembly And Overclocking

We didn't run into any real trouble assembling this quarter's system. Although we sometimes bang our heads against inexpensive cases and their flimsy structures, HEC's Blitz distinguishes itself with great build quality and lots of interior space.

Probably the worst thing we can say about this case is that it doesn't seem to have a 2.5" drive mount. So, we fixed our SSD to the side of a 5.25" drive bay. Otherwise, our only hardware-oriented complaint is a hard-to-reach screw hole up on top of Gigabyte's GA-970A-D3 motherboard, which is obscured by one of the memory slot clips.

When it comes to BIOS and software support, everything worked straight out of the box. No BIOS update was needed with a Bulldozer-based chip just to get our Piledriver-based FX-8350 up and running. In an age of GUI-driven UEFIs, the GA-970A-D3's old-school BIOS seems a little out of place. But with all of the functionality we need easily available, we have no complaints.

Overclocking

This is the first time anyone at Tom's Hardware has tried his hand at overclocking a retail FX-8350. And, after reading Chris' experience taking his sample from AMD up above 5 GHz, I was looking forward to something similar. It turns out that I was being far too ambitious, though. Xigmatek's Loki doesn't have the headroom to keep the 125 W processor cool beyond its stock clock rates. Beyond performance, thermals are probably AMD's biggest disadvantage in this comparison. We really would need to spend a lot more on cooling to achieve any sort of meaningful overclock.

Regardless of the processor or northbridge voltages we used, we couldn't exceed 4.63 GHz. "Fair enough," I first thought. "If I disable Turbo Core and lock the chip in at 4.6 GHz, I should still see a reasonable speed-up." But a Prime95-induced load quickly demonstrated instability as the FX-8350 shot up over 80 degrees.

It seems as though I had underestimated the FX's ability to generate copious heat, and failed to budget enough for cooling. Even at the stock 1.35 V setting, and with the clock rate dialed in to the processor's peak Turbo Core frequency of 4.3 GHz, Prime95 caused the chip to falter. Simply nudging clock rate, without touching the voltage, results in a significant temperature increase. For example, operating at 4 GHz yields a maximum 60-degree reading, but 4.2 GHz sees that number jump to 70 degrees. Interestingly, I didn't see any throttling, as Chris did when his sample crested 70 degrees. Here's the thing, though: while his Tj. Max was reported as 70 degrees, the retail processors are capped at 90, though the chip is clearly unstable well before it gets that hot.

The best I could achieve with this build's heat sink was 4.33 GHz, forced by dropping the voltage to 1.3375 V, turning off Turbo Core, and increasing the multiplier. Prime95 didn't crash, and the temperature stayed under 75 degrees. We're hesitant to call this a bad sample when the cooler is seemingly barely adequate. Should we choose an FX in the future, we'll need to cut back elsewhere on our budget to leave more room for a higher-end air or closed-loop liquid solution.

The GeForce GTX 670 didn't overclock as well as we would have liked, either. A 60 MHz core speed-up results in a stable 1137 MHz, which doesn't sound too bad until you look back to last quarter and see I was able to sustain 1243 MHz. 

Conversely, the current card's memory overclocks incredibly well, hitting a 7012 MT/s data rate. Compare that to last quarter's 6500 MT/s ceiling.

Test System And Benchmarks

Windows 8 is here and we've updated the entire benchmark suite. That's the good news. The bad news is that, in the process of adopting new workloads, new versions, and more applications, we don't have comparison data for some of the tests. Last quarter's numbers are included where applicable, and of course next quarter, we'll have an even more comprehensive suite to present.

And here are the benchmark details:

Benchmark Results: Synthetics

3DMark 11 is one of the benchmarks that we're recording scores from differently. Instead of capturing the Extreme, Performance, and Entry preset data, we're giving you the Graphics, Physics, and Overall scores using Futuremark's Performance preset. This gives us a clearer picture of how the different subsystems work together, rather than three aggregate numbers.

3DMark typically favors Intel's processors. Moreover, our GeForce GTX 670 didn't overclock as well this quarter. That should cover the differences we see between the old and new machines. Future comparisons will be more enlightening, since we'll have numbers for the CPU (the Physics suite) and GPU on both systems.

We're adding the Productivity suite to our PCMark scores.

Intel processors typically do best in this benchmark, which is derived from components that ship with Windows, but don't often take advantage of more than four cores.

The storage results are similar, which is no surprise considering both platforms use the same OCZ solid-state drive. 

AMD's FX CPU places well when it comes to Whetstone and multimedia benchmarks, but Intel leads in Dhrystone and memory bandwidth.

The Cinebench (based on Maxon's Cinema 4D) and Fritz benchmarks are new to our suite. So, we don't have any comparison data to show you from last quarter. Next time around we'll be able to make a good comparison in these heavily-threaded metrics, though.

Benchmark Results: Media Encoding

Although of our version numbers are updated, the workloads we're using remain identical, and are comparable. While the encoding applications have been updated, the workloads remain identical and are comparable.

Lame and iTunes are used to convert audio files from one format to another, and neither of them are threaded. Not surprisingly, then, the Core i5-3570K has little trouble besting this quarter's FX-8350, which isn't as fast per-core.

Conversely, HandBrake and TotalCode Studio (formerly known as MainConcept) are heavily-threaded, illustrating the FX's advantage.

At stock speeds, the FX-8350 is the faster CPU. But because it gets so hot, and consequently doesn't overclock well, it doesn't enjoy much of a speed-up when we tune it. The Core i5, on the other hand, is able to accelerate into a lead in both applications thanks to its jump to 4.6 GHz.

Benchmark Results: Rendering And Productivity

One of the new tests in our System Builder Marathon suite is Blender. Although we don't have comparison data from the Core i5-based machine, the trend tells us that this threaded benchmark would likely favor the FX-8350 at stock settings and then shift toward the overclocked Core i5.

The FX-8350 system is slower than we might have expected in 3ds Max, though. This one is threaded as well, so we at least would have thought it'd best the stock Core i5. However, repeated iterations show our results are accurate.

The new system does well at its stock clock rates compared to the previous build in ABBYY's FineReader OCR application, but again succumbs to the overclocked Core i5.

We have no comparison data for Visual Studio, unfortunately. However, AMD FX-8350 Review: Does Piledriver Fix Bulldozer's Flaws? showed us that the the FX-8350 should edge out Intel's Core i5-3570K at stock settings. Again, an aggressive overclock would likely turn things in favor of the Intel processor eventually, though.

Benchmark Results: Adobe CS6

We're also adding a number of applications from Adobe's Creative Suite 6, in addition to the processor-bound Photoshop test that you're accustomed to from past System Builder Marathons.

On top of that one, we have an OpenCL-accelerated Photoshop test able to help us derive graphics card compute performance. That's something we've been missing in the past, and it'll now be part of our evaluation criteria.

The classic Photoshop test is the only one we have comparison data for, and we can see the FX-8350 doing exceptionally well against Intel's Core i5. It even picks up some additional performance from our dialed-in overclock, subtle though it may be.

Benchmark Results: Compression Tools

Our new suite sees us using a more demanding compression workload (larger than 1.3 GB of files). Moreover, the introduction of WinZip 17 means we can finally apply OpenCL acceleration to any platform with support. 

Because we're compressing a different set of files, none of the benchmarks from last quarter's SBM can be compared, unfortunately. We look forward to testing Intel's Core i5-3570K in a future match-up, though.

Benchmark Results: Battlefield 3 And The Elder Scrolls V: Skyrim

Battlefield 3

Battlefield 3's single-player campaign is typically limited by graphics performance. We do see small variations at stock settings, but because the same exact Gigabyte GeForce GTX 670 is used in both builds, the differences are pretty small at stock frequencies.

Last quarter's more generous graphics card overclock gives that setup a quantifiable advantage once both machines are tuned.

Skyrim

Skyrim represents the other extreme; its performance is largely platform-bound.

Using the High quality preset, Intel's Core i5-3570K has a big impact on performance, and it trounces the FX-8350 when it's overclocked. But it also wins at stock settings, too. Fortunately for AMD, even the lowest results we recorded average greater than 80 FPS. That's plenty for a smooth gaming experience.

Stepping up to the Ultra preset, last quarter's machine maintains a compelling advantage up through 1920x1080. For some reason, though, the AMD-based box is able to secure a slight lead when it gets overclocked.

Benchmark Results: F1 2012

F1 2012

DiRT Showdown is being replaced by F1 2012, which is newer and circumvents some of the performance issues we've seen in the past from DiRT.

Unfortunately, again, we don't have any comparison data from last quarter's configuration. But, based on previous experience with older versions of titles based on this same engine, we expect processor performance to play a role in determining performance up until the higher resolutions and detail settings are used.

Power And Temperature

Idle power consumption is pretty close when you look at both boxes. Moreover, power use under a graphics load is fairly similar. No surprise there, right? We're using the exact same card in both machines, after all.

The real differentiator comes from swapping back and forth between processors under load. AMD's FX uses about 150 W more than Intel's Core i5.

This is amazing. Why? The FX-8350 is actually undervolted when we overclock it. And yet, it draws significantly more power from the wall.

Thermal readings require some context, since voltages, heat sinks, and fans all have an impact on how far we're able to push each platform. However, we observe that the GPU idle and load numbers are very similar between both builds (as we'd expect). The Core i5 idles much cooler than AMD's FX. The Core i5 is also cooler at its stock settings. Of course, when we overclock, though, both get pushed as far as possible at their respective clock rates. And so, they end up roughly as hot.

Analyzing Our FX-8350 Experiment

Although we cannot compare the results from our entire benchmark suite (due to the addition of several tests and the modification of other workloads), we're still curious to see how the comparable numbers match up.

On average, the FX-8350 and Core i5-3570K do pretty well at their stock settings, the Intel-based box about 10% quicker. This will likely change as we fold more heavily-threaded tests into the Marathon, starting this quarter. Naturally, you'll want to look closest at the benchmarks that matter to you specifically when you evaluate performance, since each architecture excels in a different way.

When it comes to overclocking though, Intel extends its lead with significantly lower power consumption and much better performance. If we were measuring efficiency, that'd be a home run. Yes, Xigmatek's Loki is insufficient for overclocking the 125 W FX-8350. But let's be realistic. If we wanted to squeeze better performance out of AMD's chip, we'd need to spend more money on cooling, and power consumption would rise even faster as higher voltages paved the way for more aggressive clock rates. It'd be a great experiment, and we might even play around with it in the future, but it's clear that Intel's Core i5-3570K remains the better choice for overclockers in this price range.

Here's another idea: maybe the cheaper FX-8320 would make more sense matched up to a pricier (and more capable) cooler? While we consider what might have been, or what might be next quarter, prepare yourself for Thomas' $2,000 build and then the value comparison. We'll get to see how the $500, $1,000, and $2,000 systems compare when it comes to performance, power, and, well, overall value.

Four New 2.5" Hard Drives, Benchmarked

Over the past few years, 2.5" hard disk capacities have grown by leaps and bounds. And although they're still typically slower than 3.5" drives, performance in the 2.5" form factor continues to increase due to higher data density. Today, even 2.5" drives positioned as quiet energy-savers manage sequential transfer rates over 85 MB/s, while more performance-oriented disks sport transfer rates just shy of 100 MB/s.

We're adding four new drives to our 2012 Mobile Hard Drive charts: Hitachi's Travelstar 7K750 and Travelstar 5K1000, Toshiba's MQ01ABD100, and Western Digital's Scorpio Blue WD5000LPVT. Among these four models, only the Travelstar 7K750 is performance-oriented. The Travelstar 5K1000, Toshiba's disk, and the Western Digital drive are “green” offerings that focus on low noise and low power draw.

A 2.5" Hard Disk Market Survey

Of the five major hard disk brands (Hitachi, Samsung, Seagate, Toshiba, and Western Digital), there are really only three manufacturers. Hitachi’s drive division was acquired by Western Digital, and Samsung’s by Seagate. While Hitachi and Samsung disks only live on as brand names, so far the acquisitions aren't impacting the model line-ups. You can still choose among a wide variety of 2.5" repositories, including mammoth 1 TB models.

Although terabyte-sized monsters, such as the Toshiba MK1059GSM, have been available for more than two years, innovation in the 2.5" disk market has not halted. The evidence can be seen in one of our own little pet peeves: disk height. If you ordered a 1 TB drive two years ago to beef up your notebook, you might have been in for a surprise. Odds are good that the component wouldn't have fit in your notebook’s drive bay! Because the first generation of 1 TB drives typically used three 333 GB platters, their height necessarily increased to 12.5 mm, which is too tall for most mobile platforms.

500 GB Platters Allow Standard-Height 1 TB Drives

As the data density of platters increases, those problems are now surmountable. Most mobile 1 TB disks have transitioned to two 500 GB platters at a Z-height of 9.5 mm. In today's story, Toshiba's single-platter MQ01ABD100 is the density leader at 744 Gb per square inch, followed by Hitachi's Travelstar 5K1000 at 694 Gb per square inch and Travelstar 7K750 at 502 Gb per square inch.

Just because drive height compatibility issues are improving doesn't mean you should stop double-checking spec sheets before making purchases, though. Case in point, Western Digital still sells a 12.5 mm version of its 1 TB Scorpio Blue (WD10TPVT) alongside a 9.5 mm Scorpio Blue (WD10JPVT) with the same capacity.

There are other kinds of exceptions as well, like disks with a PATA interface. While parallel ATA has generally fallen out of favor, some manufacturers still offer models with this deprecated connector. For instance, Western Digital’s Scorpio Blue line-up still contains a handful of PATA-based disks. Fortunately, most mobile disks now sport at least a 3 Gb/s SATA interface. A few, such as the Hitachi Travelstar 5K1000 and Seagate Momentus XT, even support 6 Gb/s data rates.

Advanced Format is Already Ubiquitous

Achieving a 9.5 mm Z-height relies on greater data density. And while this is largely accomplished through improvements like smaller read/write heads and vertical recording, it is also partly attributable to Advanced Format (AF). AF sectors are 4 KB in size, or eight times larger than the classic 512-byte sectors. By making each sector larger, there are fewer gaps between them and less space wasted on error correction. Depending on the vendor you ask, the transition to AF is responsible for anywhere from a 7 to 11% capacity increase.

Lower-capacity disks inherently don't benefit from AF as much, and they could conceivably be built using the older legacy format if vendors wanted to do that. However, there's also a cost benefit to adopting AF. To top it off, IDEMA (the hard disk manufacturer trade group) decreed that, as of January 1, 2011, all 2.5“ and 3.5“ disks sold through the distribution channel must be AF-compliant.

AF is not just about Z-height reduction or cost savings, either. It also helped overcome the so-called 2 TB limit in 3.5" disks, whereas 32-bit-long values for partition size and logical block addresses (LBAs) would have made disks larger than 2 TB impossible.

Hitachi Travelstar 7K750 (HTS727575A9E364)

Hitachi’s portfolio of internal 2.5” hard disks is split into two subcategories: mobile drives and high-performance mobile drives. Hard disks belonging to the former category rotate at 5400 RPM, while those in the latter segment spin at 7200 RPM, enabling higher performance.

The $115 Hitachi 750 GB Travelstar 7K750 is one of those high-performance drives aimed at notebooks, compact desktop PCs, all-in-ones, and multimedia devices. Of course, you could always toss it into an external USB- or eSATA-based enclosure and use it as an external repository, too.

As older models are slowly phased out and newer versions are introduced, Hitachi’s line-up contains a mixture of drives with 3 Gb/s and 6 Gb/s SATA interfaces, and even a few products that still employ 512-byte sectors.

The new 2.5“ Travelstar 7K750 model features 16 MB of cache, a 3 Gb/s SATA interface, and AF sectors (the physical sector size is 4 KB, but 512-byte sectors are emulated). Data is stored on two 375 GB platters with 502 Gb per square inch data density. Since the disk holds just two platters, it's able to fit within a 9.5 mm Z-height, and should consequently slide into any notebook.

Performance

The Travelstar 7K750 is a very fast mobile disk, averaging 98 MB/s sequential read and write rates, putting it on a par with Western Digital's Scorpio Black WD7500BPKT (the chart leader). Western Digital's drive only wins by a slim margin after averaging the maximum, average, and minimum data rates of the two competitors. Other 7200 RPM specimens, such as the Seagate Momentus and Momentus XT, trail the 7K750 and Scorpio Black by 6 MB/s.

Hampered by 5400 RPM spindles, the three other disks in our update, Hitachi's Travelstar 5K1000, Toshiba's MQ010A, and Western Digital's Scorpio Blue, naturally trail the 7200 RPM 7K750 by a significant margin.

With average access times of 16.4 ms (read) and 15.5 ms (write), the Hitachi disk lands in the middle of the pack. Occasionally, though, its power draw is slightly higher than average. For example, idle power sits at 0.8 W, while the maximum value among its competitors is 0.6 W. But average power use seems pretty normal, and a 59.6-point performance-per-watt rating is better than the other three drives. Western Digital's Scorpio Blue WD10JPVT tops the performance-per-watt chart with an outstanding score of 78.1.

Other Travelstar 7K750 Models

Apart from the model we tested, Hitachi offers two other variants. One model has hardware encryption support, called BDE (Bulk Data Encryption), and another model boasting enhanced availability (EA) is rated for for 24/7 operation.

The EA models are intended for mission-critical systems, such as blade servers, routers, RAID configurations, and surveillance systems. Unfortunately, MTBF data is not available for any of the three drives.

For reference, though, the standard model number ends in 4, while the BDE-enabled disk's model number ends in 1. The EA variant has the letter E instead of S as the third character in the model sequence (S for standard and E for enhanced).

Hitachi Travelstar 5K1000 (HTS541010A9E680)

Contrary to the Travelstar 7K750, the 5K1000 is not deemed a high-performance drive, but merely a mobile drive. A data cache only 8 MB large and a comparatively slow 5400 RPM spindle speed reflect this categorization. Modest power consumption and quiet operation are where this disk is supposed to excel, not performance. Fittingly, Hitachi is targeting the 5K1000 at notebooks, gaming consoles, DVRs, and external drive enclosures.

The Travelstar 5K1000's specs are still fairly modern, though. The disk features a 6 Gb/s SATA interface, which the “high-performance" 7K750 lacks (not that it matters; neither drive comes anywhere close to even saturating a 3 Gb/s connection). It also boasts 1 TB of capacity.

A street price of $110 seems fair, even though the 1 TB Toshiba MQ01ABD100 is slightly less expensive at $94. The two disks are priced competitively enough for customers to make a choice based on merit, not just price. Just remember that there are several other 1 TB 2.5” disks available, such as Samsung's Spinpoint M8 HN-M101MBB, Toshiba's M1059GSM, and Western Digital's Scorpio Blue WD10JPVT.

The Hitachi Travelstar 5K1000's 1 TB of storage is divided between two 500 GB platters, resulting in a data density of 694 Gb per square inch. Those two platters enable a Z-height of 9.5 mm, paving the way for a wide range of viable applications since it should fit into any 2.5” drive bay.

Performance

The Hitachi Travelstar 5K1000’s focus on capacity and energy savings is evident in our benchmarks, where it places last among the four new candidates. Given that it's only narrowly beaten by Toshiba's MQ01ABD100, we still consider the 5K1000's 88 MB/s read and write performance fairly good. Let us be clear, this is not at all a slow disk; it even beats the Seagate Momentus XT and Hitachi Travelstar 7K500 (both 7200 RPM-based products) in the h2benchw benchmark.

While all four 2.5” hard disks in today's update present similar read access times, we clocked the Hitachi Travelstar 5K1000 at a better-than-average 16.8 ms, with an average write access time of 18.4 ms. Hitachi specifies the 5K1000 as an energy-saver, and testing shows that it only draws 0.6 W at idle, while consuming just 2.2 W in a write stress test. Those are both excellent numbers.

Other Travelstar 5K1000 Models

Like the Hitachi 7K750, the Hitachi 5K1000 also comes in two other versions: BDE (Bulk Data Encryption), and enhanced availability (EA). Again, though, MTBF ratings for any of the models are not available from Hitachi.

The standard model number of this disk ends in 0. The BDE-enabled model number still ends in 1, while the EA variant again has the letter E instead of the letter S as the third character in its model sequence. An EA model with BDE does not appear to be available for the 5K1000, either.

Toshiba MQ01ABDxxx (MQ01ABD100)

We've seen Toshiba focus on physically small drives for quite some time, and the company offers more models in this segment than any other hard drive vendor. Searching Toshiba's website for 2.5" client-oriented drives turns up an astounding 41 product families. So, we tried sorting the disks according to five use cases: computing, consumer electronics, multimedia/video, automotive, and industrial. Even then, we still had a difficult time finding the right product among the mass of candidates. Computing alone is so general that it doesn't really help us narrow down the field at all.

The MQ01ABDxxx family appears in three of the five applications: computing, consumer electronics, and multimedia. According to Toshiba, its MQ01ABDxxx is optimized for quiet, energy-efficient operation, and should be well-suited for use in mobile PCs, desktop PCs, game consoles, and DVRs.

With a spindle speed of 5400 RPM and a data cache just 8 MB-large, maximizing transfer rate wasn’t a major design criteria. The MQ01ABDxxx-series sports a 3 Gb/s SATA interface complemented by Advanced Format support, which presents 512-byte sectors to the operating system as it internally uses 4 KB sectors.

We're testing the flagship MQ01ABD100, which has a capacity of 1 TB and costs about $94 online. Although it's not the only 9.5 mm-high 1 TB drive we have on-hand, cramming that much data onto two platters still is a major technical achievement. As we already pointed out, a 12.5 mm three-platter disk simply doesn't fit into the hard drive bays of most notebooks.

The Toshiba MQ01ABD100 packs a whopping 744 Gb into one square inch of magnetic media, the highest data density in our test. Meanwhile the 1 TB Hitachi Travelstar 5K1000 has a slightly lower, yet still impressive, data density at 694 Gb per square inch.

Performance

Despite the fact that it's primarily positioned as a quiet, power-saving storage product, the Toshiba MQ01ABD100 demonstrates impressive results in our h2benchw benchmark. With an average read/write rate of 89 MB/s, it pulls ahead of the Hitachi Travelstar 5K1000 by a slim margin. Other drives capable of similar performance include Seagate's Momentus XT and Western Digital's Scorpio Blue WD10JPVT. Not unexpectedly, this Toshiba MQ01ABD100 loses in a direct comparison to 7200 RPM disks like Hitachi's Travelstar 7K750.

Toshiba specifies an access time rating of 12 ms, but our measurements come in significantly higher than that at 17.8 ms (read) and 21.3 ms (write). Being the slowest of our four new test candidates, we wonder if there is some correlation with this disk's high data density? Even still, although its access times look really slow, that's probably not particularly relevant to the target markets this drive is intended to satisfy.

An average of 0.5 W in our idle power draw test is beaten only by Western Digital's Scorpio Blue and Seagate Momentus XT. The maximum power draw is just 2.9 W, which confirms the MQ01ABD100 as a “green” drive.

Western Digital Scorpio Blue (WD5000LPVT)

Western Digital's Scorpio Blue hard drive family slogan is "Big capacity for portable computing." This pretty much sums up what the line-up is all about. Fitting plenty of storage capacity into a small form factor that consumes as little power as possible while generating minimal noise is the Scorpio Blue's objective, not achieving blistering performance. Therefore, all of the models in this family have 5400 RPM spindles. The high-density drives should still achieve decent data rates, though.

Western Digital pushes the envelope when it comes to data density, packing up to 500 GB on a single platter. Beyond the 500 GB model on our test bench today, the Scorpio Blue series is also available in capacities of 250, 320, 640, 750, and 1000 GB. The 1 TB flagship (WD10JPVT) we tested a while back does its job in a 9.5 mm-high form factor, and thus fits into any notebook drive bay. But beware; Western Digital still manufactures a three-platter 1 TB Scorpio Blue called the WD10TPVT, which is 12.5 mm-high.

This time around, however, we chose to take a look at a very unique version of the 500 GB Scorpio Blue. It's a 6.8 mm-high, single-platter WD5000LPVT that should sport lower power consumption than most, if not all, two-platter disks. Indeed, the WD5000LPVT's datasheet, along with its 320 and 250 GB derivatives, reveals very low figures for power use.

Performance

Aggressive performance is not a primary design goal of the Western Digital Scorpio Blue WD5000LPVT. Nevertheless, the disk does offer a surprisingly high data rate. Its average sequential read and write operations clock in at 90.1 MB/s, surpassing the other two 5400 RPM contenders getting benchmarked for the first time: Hitachi's Travelstar 5K1000 and Toshiba's MQ01ABD100. The WD5000LPVT even matches two 7200 RPM models, Seagate's Momentus and Momentus XT.

The WD5000LPVT's access times are fairly typical: 16.3 ms (read) and 17.3 ms (write). We could only partially confirm Western Digital's specified low power consumption values. According to our measurements, the disk draws 0.4 W at idle; only Seagate's Momentus XT draws less power at 0.3 W. However, during a write stress test, the WD5000LPVT;s draw shot up to 3.3 W, which is significantly higher than the 1.5 W claimed in this drive's datasheet.

Comparison Table And Test Configuration

Comparison Table

Test Configuration

Benchmark Results: Throughput And Interface Bandwidth

The 7200 RPM Travelstar 7K750 is the only new disk in this update claiming to be a performance-oriented 2.5" drive. Not surprisingly, it delivers solid results in the sequential read and write benchmarks, registering an average of 98 MB/s each. This is only slightly exceeded by Western Digital's Scorpio Black WD7500BPKT.

With spindle speeds of 5400 RPM, the other three disks predictably trail the 7K750. But the Western Digital Scorpio Blue WD5000LPVT, Toshiba MQ01ABD100, and Hitachi Travelstar 5K1000 aren’t slow by any means. The slowest of the three trails Hitachi's 7K750 by a mere 10 MB/s, on average.

Hitachi's Travelstar 5K1000 is the only one of the four new hard disks sporting a 6 Gb/s SATA interface. In theory, this would give the drive a significant performance advantage over its competition. In reality, though, this measurement only really conveys performance between small data caches and the interface itself. Should a hit occur, you'd get a nice speed-up. But because these drives only have 8 or 16 MB buffers, this chart blows Hitachi's lead way out of proportion.

Benchmark Results: Streaming Reads/Writes And 4 KB Random Reads/Writes

Hitachi's Travelstar 7K750 handles streaming workloads best. However, it's trailed quite closely by Western Digital's Scorpio Blue WD5000LPVT and Toshiba's MQ01ABD100. Although the Hitachi Travelstar 5K1000 falls further behind, it's still no slouch when it comes to streaming.

Typically, disks with high spindle speeds and low access times shine in random read/write tests. However, the three 5400 RPM disks take the lead compared to our lone 7200 RPM sample in Iometer's random read/write test.

Western Digital's Scorpio Blue WD5000LPVT is the fastest, followed by mid-range performances from Hitachi's Travelstar 5K1000 and Toshiba's MQ01ABD100, with the Travelstar 7K750 finishing last, despite its 7200 RPM spindle.

Benchmark Results: Access Times And I/O Performance

h2benchw is a low-level metric that directly accesses the hard disk, ignoring the operating system’s sector translation. This test is relevant because all four candidates use 4 KB sectors internally, but present them as 512-byte sectors to the operating system. We aren't presented with any surprises, however. The Hitachi Travelstar 7K750 leads the pack, and the Toshiba MQ01ABD100 trails behind.

None of the four 2.5" disks fare particularly well in the database, Web server, or workstation benchmarks. Had we tested the enhanced availability versions of the two Hitachi disks, we would have hoped for better results.

In the end, Western Digital's Scorpio Blue WD5000LPVT winds up winning all four Iometer benchmarks.

Benchmark Results: PCMark 7

Due to their design as quiet, energy-efficient disks, the Hitachi Travelstar 5K1000, Toshiba MQ01ABD100, and Western Digital Scorpio Blue WD5000LPVT do not stand out in PCMark 7. Meanwhile, the 7200 RPM Hitachi Travelstar 7K750 only manages to eke out a small advantage over the other three disks.

Power And Performance Per Watt

Among the three power-optimized drives, Hitachi's Travelstar 5K1000 draws the least power, followed by Toshiba's MQ01ABD100 and Western Digital's Scorpio Blue WD5000LPVT. The Hitachi Travelstar 7K750, which emphasizes performance, scores worst among the four candidates.

The Travelstar 7K750 leads the streaming performance-per-watt comparison due to its solid sequential performance. The exceptionally power-frugal Toshiba MQ01ABD100 trails closely, in spite of being slower.

In the database-oriented performance-per-watt benchmark, Western Digital's Scorpio Blue WD5000LPVT pulls ahead of its three competitors.

Three 5400 RPM Drives Serve Up Solid Results

As expected, the performance-oriented Hitachi Travelstar 7K750 leads the pack of our four new test candidates. It's a fast laptop-oriented disk to be sure, achieving the highest average sequential read and write rates we've ever measured from a 2.5" hard drive. However, it can‘t establish a significant lead in PCMark 7 or our I/O-based metrics. In some tests, it actually trails competing models, even though the other three samples are 5400 RPM drives. Comparing the Travelstar 7K750 to all previously-tested disks, it emerges as a solid performer and winds up in the upper third of our charts. Not bad.

The fact that we come away thinking Hitachi's 7K750 is only a so-so performer is colored by solid showings from its "green" competition. Even though the Hitachi Travelstar 5K1000, Toshiba MQ01ABD100, and Western Digital Scorpio Blue WD5000LPVT aren't optimized for performance, they still nip at the 7200 RPM drive's heels. As an added bonus, all three 5400 RPM disks offer low power consumption. The Hitachi Travelstar 5K1000 and Western Digital Scorpio Blue WD5000LPVT even beat the 7200 RPM drive in our performance-per-watt index.

Overall, the Western Digital Scorpio Blue WD5000LPVT gets quite close to the Hitachi Travelstar 7K75, while  the Hitachi Travelstar 5K1000 and Toshiba MQ01ABD100 fall behind. But this doesn’t mean they are slow; even the slowest disk in this test achieves almost 90 MB/s in our sequential read and write benchmarks.

As we wrap up, we want to again shine a spotlight on Advanced Format, which facilitates higher data density and helps finesse 1 TB of capacity into a 9.5 mm Z-height, necessary for broad notebook compatibility. Although higher data density helps as well, this has an undesirable side effect. Mainly, seek times continue going up. This is because it's becoming more and more difficult to position and stabilize microscopic read/write heads over the correct track. This might not be a big deal for slow-turning repositories for user data, but it could sabotage attempts to build high-performance disks in the future.

The AFP news agency reports that Hitachi has discovered a way to store digital information on slivers of quartz glass. This data can seemingly exist forever, enduring extreme temperatures and hostile conditions without degrading... at least until the sun begins to die and expend to consume the earth, that is.

According to the report, the new tech stores data in binary form by creating dots inside thin sheets of quartz glass which can be read using your standard optical microscope. The current prototype is roughly 0.8-inches square and 0.08-inches thick, and consists of four layers of dots which can hold up to 40 MB per square inch -- approximately the density of a music CD. More data layers can easily be added.

Hitachi researcher Kazuyoshi Torii said that quartz glass is highly stable and resilient material, used to make beakers and other instruments for laboratory use. Due to the medium, the chip is waterproof, resistant to many chemicals and unaffected by radio waves. Even more, it can be exposed directly to high temperature flames and heated to 1,832 Fahrenheit for at least two hours without being damaged.

"We believe data will survive unless this hard glass is broken," said senior researcher Takao Watanabe.

The discovery arrives as consumers push their data to the cloud. Physical media is an acceptable temporary backup, but not ideal for the long-term. Even NAND flash-based solutions won't last forever, thus pushing consumers to find ways of storing their digital photos and other personal items online. This too isn't exactly ideal, but the data is backed up by multiple servers, thus serving as a more secure solution than a simple HDD or DVD.

"The volume of data being created every day is exploding, but in terms of keeping it for later generations, we haven't necessarily improved since the days we inscribed things on stones," Torii added. "The possibility of losing information may actually have increased."

The article doesn't state how these chips will be "read" outside using a microscope although it sounds like some kind of microscope-based reader is on the horizon. There's also no indication as to when this storage solution will become commercially available, but researchers indicated that government agencies, museums and religious organizations will likely be first in line.

_{Contact Us for News Tips, Corrections and Feedback}

Outside The Box: Three USB 3.0-Based Hard Drives

Even in this day and age of blazing-fast SSDs and inexpensive Blu-ray burners, external hard disks remain the top choice for blending mobility, capacity, and performance. There is no way for conventional magnetic storage to outperform solid-state storage, and optical storage is just so much less convenient. So, yeah, mobile hard drives are a lot slower than SSDs and, in many ways, they're less robust than Blu-ray discs. However, they do offer other advantages like comparatively high capacities at lower and the built-in compatibility associated with USB. Just plug them into any semi-modern PC or Mac, and their respective operating systems automatically load the requisite drivers.

External 2.5" disks are perceived as small and light-weight, though they give up a lot of space compared to 3.5" models. But smaller drives offer several advantages that help even the playing field a bit. Aside from their pocket-friendly size, many 2.5" drives are able to draw all of the power they need through the USB interface. Say farewell to bulky wall warts and black bricks. 

Fortunately, 2.5" drives have evolved tremendously, picking up capacity along the way. They can't yet match the multiple terabytes crammed into 3.5" disks. However, one of the models we're reviewing packs in an enormous 2 TB, which is half of the capacity of the largest 3.5" external drive. Moreover, it does this in less than one-quarter of the 3.5" solution's volume. The other two contenders on our bench both weigh in with 750 GB, which is a little more typical in the 2.5" space.

Where 2.5" drives really do fall short, however, are in performance benchmarks. Even complemented by identical USB 3.0 controllers, 2.5” disks are slower than the 3.5" options out there due to the physics of angular velocity and how it affects the amount of data passing under the heads. But as we've shown in the past, the fastest 2.5" models achieve up to 80 MB/s or so, serving up enough sequential throughput for everyday use. Not all drives are created equal, though, and for good reason: the road warrior who needs to transport lots of data needs a storage device that behaves differently than a network admin's super-fast backup target.

We rounded up three external drives geared toward different types of users, and we want to know how they perform both in synthetic and real-world tests. Adata's DashDrive Durable HD710, Hitachi's Touro Mobile Pro, and Western Digital's My Passport are all on our test bed and ready to go.

Adata DashDrive Durable HD710 (750 GB)

Adata's USB 3.0-based DashDrive Durable HD710 is available in capacities of 500 GB, 750 GB, and 1 TB. You can find the external case highlighted by yellow, blue, or black, and there's no price difference between the colors. We just so happened to test a blue 750 GB model, which sells for just over $100 on Amazon. We liked that the drive's USB 3.0 cable is integrated, and thus cannot be forgotten or, worse, lost.

The DashDrive Durable HD710 is ruggedized, and Adata isn’t shy about advertising that fact. Its massive case sports diagonally-printed stripes that make us think of highway warning signs. According to its manufacturer, the HD710 conforms to the international standard IEC 529 IPX7, surviving a 30-minute immersion in 1 m (3 ft) of water. It can also survive a drop test, according to MIL-STD-810G 516.5, which is equivalent to a fall from a table to the floor.

Somewhat disappointingly for such a purportedly tough drive, its warranty runs out after a mere three years.

Robustness is what sets the Adata drive apart from the other two products in our story. It's suitable for taxing operating conditions that might render other mechanical devices unusable. Inside, Adata employs a 5400 RPM hard drive, which assures modest (but not particularly noteworthy) performance. Its sequential read and write rates of 77.8 MB/s earn the DashDrive a last-place finish amongst our three test candidates. Then again, that's not too far down the ladder. Western Digital's My Passport achieves a slightly higher result that hovers around 83 MB/s.

In contrast to its competitors, Adata's DashDrive Durable HD710 does not come with software. Its bundle is limited to a quick-start guide and the USB 3.0 cable built right into the external enclosure. Two tools, OStoGo and UFDtoGo, can be downloaded from the Adata website. But they're relatively mundane offerings: OStoGo copies the files off of a Windows 7 DVD on the external drive, and then modifies them in such a way that you can install Windows from the HD710. UFDtoGo synchronizes personal data (like email and browser favorites) between a PC and Adata's product.

Hitachi Touro Mobile Pro (750 GB, HTOLMNA7501BBB)

We previously tested the 500 GB Hitachi Touro Mobile Pro, and it proved to be one of the fastest 2.5” drives we've seen. In fact, it claimed a ranking in the top third of our USB 3.0 Storage Charts. The 750 GB version (HTOLMNA7501BBB), which sells for less than $100, is slightly faster still.

The piano black Touro Mobile Pro sports a USB 3.0 interface like its smaller sibling, and it also boasts a 7200 RPM drive inside. But that's where similarities to the 500 GB model end. The 750 GB version is the fastest external 2.5” drive we have ever tested. Its maximum sequential read rate of 125.6 MB/s and its maximum write speed of 125.5 MB/s make it the new speed king, pushing the previous winner, Samsung's S2, down one spot, even though the S2 is still better with small files.

It's also worth noting that, while Samsung's S2 loses out to Hitachi's Touro Mobile Pro in synthetics, it remains faster in real-world metrics, such as the copy test.

Hitachi's drive is positioned as a professional product for backing up data. Although it can be used to save information from the cloud, the Touro Mobile Pro is also capable of putting data into the cloud through a bundled application called Hitachi Backup. The software includes 3 GB of online storage space, and is upgradeable to 250 GB for $49 per year. The box also contains a quick-start guide and a USB 3.0 cable.

Hitachi Backup may not be feature-rich, but this Flash-based application is easy to use. By default, it copies the logged-in user's data to the external drive or the cloud either at pre-defined times or on demand. You can specify different folders for the two backup targets, and also define the number of revisions to keep. The latest backup is available in a folder named “Current,” and older backups are found in “Archived.” With a slider, you can specify the bandwidth usage of the cloud backup function as “Low,” “Medium,” or “High”.

Western Digital My Passport (2 TB, WDBY8L0020BBK)

Adata's drive is the toughest one, Hitachi's is the fastest, and Western Digital's submission turned out to be the largest, setting a new capacity record for external 2.5” drives with 2 TB of space.

In order to fit 2 TB on a 2.5” drive, Western Digital's disk consists of four 500 GB platters. This leads to a slightly thicker case (0.83”). But, all things considered, the chassis is still smaller than the ruggedized Adata drive, and only marginally heavier, weighing in at 0.5 lbs. Like its storage capacity, the My Passport's price is hefty at $200. When you divide capacity into cost, though, this model is actually pretty comparable to the competition.

If the 2 TB model is too rich for your blood, Western Digital also sells 500 GB, 750 GB, and 1 TB versions of the My Passport.

Western Digital's drive features a 5400 RPM disk drive able to achieve maximum sequential read rates of 83.5 MB/s and maximum sequential writes as high as 83.3 MB/s. That's slightly faster than Adata's DashDrive Durable HD710, but it can't come close to Hitachi's Touro Mobile Pro.

The bundled software, SmartWare 1.5.4.5, is an all-around tool for backing up, restoring, and erasing data. It features a timer function that can turn off the drive, along with a diagnostic utility. Each software feature is accessed via a separate tab, aiding ease of use. 

Although auditioning the software didn't reveal any amazing innovations, there is a handful of features we liked. For instance, SmartWare can execute a backup after an idle timer triggers the action, which should minimize the impact on productivity after someone steps away from their machine. The backup tool supports versioning; it keeps a user-configurable number of older file versions.

Another integrated tool, WD Security, allows you to enter a drive password. The My Passport only appears as an accessible drive after the password is entered.

Technical Data And Test Configuration

Technical Data

Test Configuration

Benchmark Results: Throughput And Interface Bandwidth

Driven by its 7200 RPM disk, Hitachi's Touro Mobile Pro achieves a throughput of more than 125 MB/s, which earns it a top spot in this largely theoretical measurement. The previous speed king, Samsung's S2, is pushed to second place.

The other two new drives, Western Digital's My Passport and Adata's DashDrive Durable HD710, sport 5400 RPM disks and end up finishing fairly close to each other in the performance chart by pushing 83 MB/s and 78 MB/s, respectively.

The Hitachi Touro Mobile Pro is the strongest contender in this line-up when it comes to streaming reads and writes, though Iometer is less forgiving than h2benchw. The Adata DashDrive Durable HD710 and the Western Digital My Passport still cannot compare, though, as they're similarly beaten-up by Iometer.

Benchmark Results: Access Time And I/O Performance

Among today's three contenders, access times vary by only seven milliseconds. That's small enough to not affect the test results by much.

There is a notable performance difference between the Hitachi drive, on one hand, and the Adata and Western Digital drives, on the other. The Iometer benchmarks demonstrate this. In the file server and Web server workloads, Hitachi's Touro Mobile Pro outclasses its competition. Adata's DashDrive Durable HD710 and Western Digital's My Passport occupy a lower performance echelon, finishing close together and challenging each other for second place among our three candidates.

Despite the Touro's strong showing in these two charts, however, the results are only marginally relevant. External USB 3.0-based disk drives aren't designed for I/O-heavy tasks. So, it's less likely that you'd replicate these numbers than the sequential benchmark figures on the previous page. 

Real-world results would be even better. On the next page, we'll dig into what these drives are really meant to do: copying files.

Benchmark Results: Real-World Copying

JPEG Images

We're really able to see the benefit of adopting USB 3.0-connected drives when it comes to transferring large files. But the performance picture isn't perfect.

Although Hitachi's 750 GB Touro Mobile Pro is a strong performer in the read test, it gets beaten by Western Digital's 2 TB My Passport for the first time in our round-up. It also succumbs to Samsung's S2 and a 500 GB version of itself. It also loses to the Samsung S2 in our write test.

As a test control, we have a couple of USB 2.0-based drives at the bottom of the chart. Their performance is very similar because they're being constrained by the bus' maximum throughput. Truly, USB 3.0 is instrumental in popping the cork on external storage performance.

MP3 Files

When we start moving compressed MP3 audio, the results appear largely similar. Western Digital's My Passport scores another win over Hitachi's Touro Mobile Pro, which takes second place in read performance. Hitachi reasserts itself in the write workload, though not by as large of a margin as the streaming Iometer numbers would have indicated.

Picking The Right External 2.5" Hard Drive

The real star of the show today is USB 3.0, which allows the Adata DashDrive Durable HD710, Hitachi Touro Mobile Pro, and Western Digital My Passport to demonstrate their respective advantages, while allowing us to pinpoint their weaknesses without interference from a USB bottleneck. Technically, you could pick any one of these three external disks and be satisfied with it. However, each has specific strengths that endear it to a particular type of customer.

The Fastest: We cannot hand Hitachi's Touro Mobile Pro an unconditional victory when it comes to performance. It loses out to Western Digital's My Passport in the real-world read tasks we'd expect these drives to be tasked with. Moreover, Samsung's S2 demonstrates more consistently-good results overall.

Hitachi does particularly well in our synthetic metrics, where it takes a number of first-place finishes. Sporting a maximum sequential read/write rate of 125 MB/s in h2benchw, you'd expect the Touro Mobile Pro to decimate its competition in picture and music file transfers. Although it is the fastest drive of the three we tested, its advantage shrinks quite a bit. In this round-up, at least, Hitachi scores a victory, though we'd hardly call it commanding.

The Toughest: If you're shopping for an external drive able to take a beating, the number of viable options drops precipitously. Fortunately, Adata's DashDrive Durable HD710 does qualify. This ruggedized option is able to drop from a tablet to the floor or spend half an hour at the bottom of a garden fountain without compromising your data. In spite of its specially-designed chassis, the DashDrive Durable HD710 is only slightly larger than non-ruggedized drives, and barely heavier, weighing in at 0.5 lbs. Its performance is average, making it respectably quick. Even though we would have liked to see a greater-than three-year warranty on this particular model, Adata can at least rest easy knowing it offers one more year of coverage than either Hitachi or Western Digital.

The Largest: Diminutive physical dimensions do not give away the fact that Western Digital's My Passport crams in an impressive 2 TB of capacity, representing a new record for 2.5” drives. If you need to maximize this ever-important variable, the My Passport is a winner. Just be ready to pay about $200 for the drive. That's only 10 cents per gigabyte, which isn't bad overall.

The good news is that Western Digital doesn't necessarily compromise performance in its aim to serve up lots of storage space. Its 5400 RPM drive is able to outmaneuver Hitachi's 7200 RPM model in our real-world read tests. It finishes further down the stack in our write tests, and is more similar to Adata's DashDrive Durable HD710 in the synthetics. You'd likely have a hard time telling the difference during day-to-day use, though.

So, Who Wins?

Each of these three drives satisfies a purpose. And, in a way, none of them directly compete against the others as a result of their different balances between performance, capacity, and durability. As a result, we'd have an equally difficult time declaring a winner as a loser; each serves its purpose well in its specialty. The key is knowing what you're looking for in a drive before you sit down to buy one. Perhaps our comparison will help guide such a purchasing decision.

NAS-Oriented Hard Drives, With A Twist

When power users and small offices populate their network-attached storage devices with hard drives, they typically lean on desktop-oriented disks in protective RAID arrays, if only because they're less expensive. Really, though, those drives are well-suited for life in a NAS, with other mechanical devices clacking around. For example, Seagate’s Barracuda hard drives aren’t certified for the continuous operation networked-storage imposes. There are a handful of exceptions, of course, like Hitachi's Deskstar, which the company says can handle the demands of 24x7 availability. But, at the end of the day, they're still desktop disks intended for client workloads.

Nearline Hard Drives

In order to achieve more robust reliability in a multi-drive array, you really need to look at purpose-built nearline hard drives. They're optimized, both at the firmware and hardware levels, to serve up higher endurance and reliability.

The usual criteria for evaluating hard drives, such as speed and price per gigabyte, are still important in the design of nearline storage. But there are other added considerations taken into account as well like reliability, a low thermal profile, and moderate power consumption. The mechanical components of nearline drives are improved to be more robust than their desktop counterparts as they cope with the increased heat of continuous operation. Manufacturers build the drives to higher, more stringent, standards, and step up their testing and quality control.

Those enhancements aren't free, of course. Nearline disks are more expensive than comparable desktop drives. They do tend to come with a longer warranty, though.

Enterprise and nearline drives like Hitachi's Ultrastar, Seagate's Constellation, and Western Digital's RE4 families feature up to five platters and spin at 7200 RPM more. In order to operate in an environment with many other mechanical storage devices, they sport rotational vibration (RV) sensors that allow them to optimize head position to avoid longer seek times due to vibrations from other drives. Typically, desktop drives are rated for 2400 power-on-hours per year, with Seagate's Barracuda 7200.14 standing in as our example. Nearline drives, on the other hand, are good for up to 8760 hours per year, the 24x7 availability equivalent.

A NAS-Oriented Alternative To Nearline: Western Digital Red

Western Digital's Red family is positioned in between the desktop and nearline drives. They're meant neither for typical desktop usage nor for large 19-inch rack-mounted servers. Instead, they’re being aimed at home office and small office network-attached storage appliances.

The resulting piece of hardware includes a blend of features from desktop and nearline drives. For instance, the 5400 RPM spindle speed is typical of entry-level desktop drives, while 24x7 operation is taken from the nearline playbook. The Western Digital Red is available in capacities of 1, 2, and 3 TB. The two larger-capacity models include SATA 6Gb/s connectivity, and they're the offerings we have in the lab today.

Meet Western Digital's Red

The Western Digital Red family joins the Blue (desktop drives), Green (efficiency-optimized desktop drives), and Black (performance-oriented desktop drives) line-ups. The three models in the Red series are intended for appliances with between one and five drive bays. Western Digital says that this specific range is optimized in its firmware.

As mentioned, the Red drives boast power-on ratings of 8760 hours per year and a mean time between failure (MTBF) rating of 1 000 000 hours, on par with the pricier nearline drives out there. They don’t employ rotational vibration sensors, which is another reason they're being positioned as solutions in smaller storage systems, and not big racks loaded down with spinning disks that weather more severe vibrations.

Western Digital's Red drives are composed of anywhere from one to three platters with a capacity of 1 TB each. The company doesn’t explicitly call out how fast their spindles turn. Instead, it reference its IntelliPower technology, designed designed to balance RPMs, transfer rate, and caching algorithms to increase efficiency and performance. According to our benchmarks, though, the Red drives are spinning at 5400 RPM, which is understandably slower than nearline drives that command a premium.

In turn, a combination of high data density and low rotational speed should at least impart lower power consumption and operating temperatures.

Performance

The 2 TB (WD20EFRX) and 3 TB (WD30EFRX) models demonstrate a sequential read speed of 112 MB/s. Generally, this is middle-of-the-road. However, it’s actually pretty fast compared to other 5400 RPM drives, such as Hitachi's Deskstar 5K4000, which achieves 102.9 MB/s, and Samsung's EcoGreen F4 HD204UI that achieves 100.7 MB/s.

Western Digital's Red drives really shine when it comes to power consumption and thermals. Consuming less than 4 W at idle and a maximum of 5.4 W during streaming write operations, the Reds are solidly in front of their competition. The operating temperature maintains a comparatively-frosty 31 degrees Celsius, which is about four degrees less than the other drives we benchmarked.

Technical Specifications And Test Configuration

Technical Specifications

Test Configuration

Benchmark Results: Data Throughput And Interface Bandwidth

The 2 TB (WD20EFRX) and 3 TB (WD30EFRX) models manage to achieve a sequential read speed of 112 MB/s. This is faster than the other 5400 RPM drives. They even beat the 7200 RPM Seagate Constellation ES.

Benchmark Results: Read/Write Access Time And I/O Performance

The Western Digital Red drives bring up the rear when it comes to access time measurements. They’ve clearly been optimized for quiet operation and avoiding load actuator movements at the cost of this metric, which would probably be more important on the desktop.

The Red series isn’t a good choice for high I/O NAS environments.

Benchmark Results: Streaming Reads/Writes, 4 KB Random Reads/Writes

The Western Digital Red drives generally deliver average to above average performance in streaming tasks. These are fairly common in a SOHO environment.

Once again, Western Digital's Red drives demonstrate slow I/O performance. Just like in Iometer’s database, file server, Web server, and workstation benchmarks, the two drives can’t hold their own against the competition.

Benchmark Results: PCMark 7

Both Red drives stand their ground in PCMark 7's application benchmarks, falling in right behind the 7200 RPM drives. Once again, this shows that the drives are optimized more for acoustics and power rather than all-out performance.

Power Consumption And Temperature

High data density (at 1 TB per platter) and low speed (5400 RPM) make these two Red drives winners in all of our power consumption benchmarks.

The Red drives consequently deliver the best performance per watt ratio.

Temperature

No other drive even comes close to the Red drives' 31 degrees Celsius operating temperature.

Red Gets A Recommendation In NAS Appliances

Western Digital's Red family is a well-rounded offering aimed at home- and SOHO-based NAS appliances. The hard drives’ strengths include low operating temperatures, conservative power consumption, and modest acoustics at a 5400 RPM spindle speed. If you're running a pedestal-based server in the same room you work and need to keep storage- and cooling-related noise to a minimum, these drives are definitely ideal. Not only are the drives themselves quiet, but their low operating temperature allows the NAS to operate for longer without spinning its own fans up.

The Red drives aren’t as attractive once you step outside of Western Digital's target usage model of small business network-attached storage and into more demanding professional applications. A slow spindle speed and the lack of a rotational vibration sensor are evidence that these disks weren't intended for use in larger servers. Also, relatively modest performance in Iometer’s database, file server, Web server, and workstation metrics, as well as our 4 KB random read and write benchmarks, make it clear that these drives aren't meant for environments where I/O performance is important.

We can wholeheartedly recommend the Western Digital Red drives in their intended application: small business NAS appliances. I/O generally isn't emphasized in that scenario, and the drives do demonstrate good streaming and sequential read/write performance (at 110 MB/s). This leads to fast transfers of large files, a common task in network-attached workloads.

The drives run cool, are quiet, and exhibit low operating power consumption. In addition, they come with some of the features usually found on enterprise-class hard drives, such as 24x7 availability and high reliability (an MTBF of 1 000 000 hours). We're only disappointed that Western Digital's warranty isn't as enterprise-oriented, limited to just three years.

Nine Thunderbolt Devices Feel The Need For Speed

Delivering up to 1 GB/s data of throughput and the ability to communicate over PCI Express externally, it's no surprise that Thunderbolt has the attention of enthusiasts today. Those same savvy buyers see that the technology still bears a prohibitively high price premium, though, and decry its lack of value. If you aren't yet familiar with Thunderbolt and want a low-level lesson on it, check out Everything You Need To Know About Thunderbolt.

Nevertheless, several Thunderbolt-based external storage solutions are now available with the potential to nearly double the peak performance of today's most popular interfaces. The question is: are vendors able to populate these enclosures with the right hardware to enable blazing-fast transfers?

Finding Its Place

We certainly understand if you're reluctant to spend extra money on a new and unproven connectivity standard that looks like it's being set up to replace USB. But take solace in Intel's insistence (for now) that Thunderbolt is meant to complement USB 3.0 and not replace it. Although the two interfaces appear at odds, they're decidedly ideal in very different ways and for equally disparate applications.

USB 3.0 does, for example, enable impressive performance across a bus rated for 5 Gb/s. However, its inefficiencies knock what, on paper, appears to be a 500 MB/s interface down far below that theoretical maximum. As an interface for storage, Thunderbolt's 10 Gb/s (1 GB/s) peak has no trouble nudging USB 3.0 from its pole position.

It's perhaps more accurate to think of Thunderbolt as the spiritual successor to the FireWire standard, aimed at users with high-end applications unable to compromise performance—and willing to pay top dollar for it.

And like FireWire 800, which was exorbitantly-priced when it debuted, today's Thunderbolt-based adapters cost more than $150 and require expensive cabling that isn't included. Intel says that it's committed to bringing down the cost of second-generation Thunderbolt controllers and cables, hopefully allowing the technology to dangle within reach of value-conscious customers sooner rather than later.

A large number of Thunderbolt-related product announcements at this year's Computex give us some indication that a flood of products should be emerging soon. But there are already dozens of Thunderbolt-based storage devices available. We gathered up nine and put them to the test.

Elgato Thunderbolt SSD 240 GB

Elgato is well known in the Mac world for its capture card products, and is now jumping into the storage market with the Thunderbolt SSD.

One significant limitation of this drive is that it hosts a single Thunderbolt port (a side effect of its Intel Port Ridge controller), which means you have to put the Thunderbolt SSD at the end of a daisy chain or use it on its own. This is an issue if you only have one Thunderbolt port and are already using a mini-DisplayPort adapter to drive a monitor; you're forced to choose between one device or the other. There is no way around that issue, as we haven't seen any mini-DisplayPort adapters with Thunderbolt pass-through.

The Thunderbolt SSD's thick aluminum case makes it rather hefty, but it's actually fairly simple inside. Aside from Intel's DSL2210 controller, you find an ASMedia ASM1061 SATA 6Gb/s controller on the PCB and an older SanDisk 3Gb/s 240 GB Ultra SSD. Based on SandForce's first-gen technology, SanDisk’s Ultra-series drives are functionally similar to OCZ's Vertex 2.

Elgato’s decision to use a SATA 3Gb/s-class SSD really impacts the device's benchmark results. As you can see, sequential read and write speeds top out at ~250 MB/s. But that's not all. Testing with incompressible data (SandForce's Achilles' heel) drops sequential write throughput to ~90 MB/s, making the Thunderbolt SSD ill-suited for working with multimedia files. Random reads and writes aren’t impressive either, unfortunately. SandForce’s first-generation controller logic performs rather poorly in those two disciplines, explaining why we see random access numbers peaking near 100 MB/s.

Upgrading the Thunderbolt SSD's internal drive voids its warranty. If you want to throw a faster SSD in there anyway, be aware that Elgato employs a tamper-resistant tar-sealed hex cap that discourages opening the case. In its stock configuration, this is a good example of a drive that lacks the composition to fully exploit the potential of Thunderbolt. And yet, after the cost of SanDisk's roughly $130 drive (in the 120 GB model), you're still paying close to $300 more to get 3 Gb/s performance in Thunderbolt trim.

G-Technologies G-RAID Thunderbolt 8 GB

G-Technologies is a subsidiary of Hitachi Global Storage Technologies, which was acquired by Western Digital a few months ago. We're told there are no plans to discontinue Hitachi GST's products. But, should that occur, Western Digital will continue to honor warranty coverage, alleviating concerns over buying a G-Technologies device.

The G-RAID Thunderbolt is housed in a silver-colored MacBook Pro-like aluminum shell, with two ports located on the back for daisy chaining, along with a power switch and fan exhaust port. The combination of an active cooler fan and an aluminum housing seems fairly effective for dissipating the heat from Intel's Thunderbolt controller and the installed hard drives. At idle and during our streaming write tests, the unit remained cool, and we're happy to say its fan runs more quietly than the mechanical disks.

The two internal 4 TB 3.5” Hitachi Deskstar 7K4000 hard drives occupy three-quarters of the length of the unit, and both repositories are positioned near the front grill. Not all G-RAID Thunderbolt units feature the configuration we're reviewing today. G-Technology also offers 4 TB (2 x 2 GB) and 6 TB (2 x 3 GB) versions employing Hitachi’s Deskstar 7K2000 and 7K3000 family, respectively. All three capacities are rated for similar performance, so your only real decision is how much storage space you need.

Upgrading the installed hard drives on your own voids the warranty, so it may be wise to err on the side of caution and buy bigger if you anticipate a rapidly-growing library of music or videos eating up capacity quickly. There is no way to sneak around this restriction, as a tamper-resistant sticker hides one of the screws required to disassemble the chassis.  

Unlike the Elgato unit, G-Technology uses Intel's CV82524EF/L Thunderbolt controller, equipped with four channels, to facilitate support for daisy chaining. Also unlike the Elgato unit, the G-RAID Thunderbolt forgoes ASMedia's storage controller in favor Marvell's 88SE9182 SATA 6Gb/s solution.

There is no hardware-based RAID controller on-board. So, you have to configure RAID via software either in Windows or using OS X's Disk Utility. That's an important note, since G-Technology incorrectly claims that this device supports hardware-based RAID 0 on its product webpage.

Configured manually to operate in RAID 0 mode, the G-RAID Thunderbolt reaches sequential read speeds of ~325 MB/s, regardless of queue depth. Sequential writes, on the other hand, start out at ~270 MB/s and jump to 320 MB/s once there are sixteen concurrent operations.

Sequential write performance hangs steady at 155 MB/s in RAID 1, and sequential reads climb from 50 MB/s to ~160 MB/s.

LaCie Little Big Disk 240 GB

LaCie’s Little Big Disk (LBD) is available in four trims: 1 TB (2 x 500 GB, 5400 RPM), 1 TB (2 x 500 GB, 7200 RPM), 2 TB (2 x 1 TB, 5400 RPM), and 240 GB (2 x 120 GB, SSD). The 240 GB model, which is what LaCie submitted for review, is unquestionably the most interesting. It's the first Thunderbolt-based device we have seen with SSDs in RAID. That's perfect for pushing the interface's performance, providing the company picks the right drives.

LaCie touts the Little Big Disk as a "portable powerhouse," and SSDs drive that message home quite effectively. However, we do end up with critical feedback to pass along. Contrary to LaCie’s product photos, the Little Big Disk requires a separate AC adapter for daisy chaining because Thunderbolt, in its current form, does not supply enough power to run everything off of bus power.

Our enthusiasm is further dampened by the fact that LaCie chose to employ Intel’s 120 GB SSD 320. Don't get us wrong, we've found the SSD 320 to be incredibly reliable. It's just not a great-performing drive. You'd get better benchmarks results from any number of more modern 6 Gb/s-capable SSDs. But before you consider a do-it-yourself replacement, bear in mind that you'd be voiding LaCie's warranty. Four screws on the back of the drive are resin-coated, making it easy to tell if the chassis is opened.

All of LaCie's Little Big Disk models employ the same Intel CV82524EF/L Thunderbolt and Marvell 88SE9182 SATA controllers.

Queue depth appears to have no impact on performance, which is good news if you don't have one of those rigorous storage workloads needed to extract peak performance from many SSD-based products.

In RAID 0, sequential reads even out at ~470 MB/s, while writes peak around 250 MB/s. Using relatively low-capacity SSDs makes it hard to imagine a scenario where RAID 1 would make sense, but we generated those numbers anyway, if only as an interesting comparison to what RAID 0 does for throughput. And, as expected, when we limit the Little Big Disk to the capabilities of a single SSD 320, its sequential reads max out at ~250 MB/s, while sequential writes fall to ~125 MB/s.

Unfortunately, the SSD 320's random performance results are fairly modest, so the numbers we see aren't as spectacular as what you might otherwise expect from a modern SSD. The 4 KB random reads slowly climb from ~20 MB/s to ~130 MB/s, with no correlation to RAID configuration. We see that 4 KB random writes max out at 100 MB/s in RAID 0; RAID 1 achieves up to ~50 MB/s.

LaCie 2big 6 TB

LaCie’s 2big external RAID enclosure, featuring hot-swappable bays, is a much better option for enthusiasts looking for a little flexibility without jeopardizing warranty coverage. The 2big isn't a clear winner over LaCie's Little Big Disk in every way, though. It's larger and less portable, whereas the Little Big Disk leverages 2.5" drives to keep its form factor much more compact. But that shouldn't be a deterrent; the 2big accepts higher-capacity 3.5" storage and is deliberately designed to be stationary. If you need mobility, LaCie has more purpose-built solutions (like the Little Big Disk). 

As with pretty much all of the Thunderbolt-based products we're looking at today, the 2big isn't an inexpensive piece of hardware. LaCie's 4 TB model (9000191) sells for $600, or ~$0.15 per GB. Jumping to the 6 TB model (9000192) increases the price to $750, but reduces the cost per-gigabyte to ~$0.09. Fortunately, both available versions can be found online for a bit less.

Regardless of the capacity point, LaCie utilizes Seagate's 7200.14 Barracuda (ST3000DM001) family inside its 2big models. That's a respectable choice, as the 7200.14 happens to be a decent middle-of-the-road performer.

The 2big employs the same Intel CV82524EF/L and Marvell 88SE9182 configuration as the Little Big Disk, but it integrates beefier power circuitry to support 3.5" disk drives.

Expect to see sequential read results somewhere in the neighborhood of 400 MB/s from the 2 big in a RAID 0 configuration. Writes level off around ~290 MB/s.

Protecting your data with RAID 1 nearly halves those performance numbers.

Promise Pegasus R6 12 TB

Sporting six hot-swappable bays, Promise’s Pegasus R6 is the top dog when it comes to performance-oriented Thunderbolt-based storage solutions. Intel and its partners commonly use this device to demonstrate what the technology is capable of achieving, and the R6's high price makes it even more exclusive. The entry-level 6 TB (6 x 1 TB) model will set you back $1800, and the 12 TB version commands a $2500 price tag.

You get what you pay for, though, and the Pegasus R6 demonstrates its suitability in business-class environments with a serial port for uninterruptible power supply support at the back of the chassis.

Promise sent us its 12 TB Pegasus R6, employing six 2 TB Hitachi 72K300 hard drives (the 6 TB variant is equipped with Hitachi’s 7K1000.D series). The drive bays are well-designed, featuring sturdy aluminum face plates and large release buttons that make it easy to swap storage out in the event of a disk failure.

The picture above, turned upside-down for better clarity, shows the interior of the Pegasus R6, where we find a 250 W power supply driving a motherboard with PMC-Sierra’s PM8011 PCIe SAS 6Gb/s RAID controller. The PM8011 is actually an eight-lane controller. But because Thunderbolt only enables a PCIe x4 uplink, the other four lanes aren't used.

It's interesting that Promise populates the SAS controller with SATA-based hard drives, though not entirely surprising given the premium SAS-capable disks would have added. You can look at the daughtercard's connectors, though, and see the full SAS interface that accommodates SATA disks, too. SAS 6Gb/s and SATA 6Gb/s both enable similar data rates, so there's no concern over compromised performance. We think that Promise's decision to arm its enclosure with SATA storage was the right one. However, should the company wish to offer a version of its R6 with nearline SAS drives in the future, that could become a possibility.

The R6 supports RAID 0, 1E, 5, 6, 10, and 50. However, we're narrowing our focus to the three configurations we feel are most attractive on a device like this: RAID 0, RAID 1E, and RAID 5.

The R6 really stretches its legs when we measure sequential read performance in RAID 0. It falls just 50 MB/s short of the 1 GB/s barrier at a queue depth of 16! If redundancy is more important to you than blistering speed, RAID 1E operates effectively as three RAID 1 arrays, which is why sequential read performance falls to ~315 MB/s. RAID 5 is a fair compromise, enabling block-level striping with distributed parity. We see sequential reads top out around ~780 MB/s in that arrangement.

This is the only device we're reviewing with performance characteristics that look a little different under OS X than Windows, and there's a reason why. Originally, Thunderbolt allowed device I/O and display signaling to share bandwidth over both bi-directional channels. However, in the summer of 2011, Apple and Intel decided to put device I/O on its own channel in order to preserve the display's signal integrity under heavy workloads.

Prior to that, we were able to achieve close to 920 MB/s in RAID 0 on our 15” MacBook Pro. However, since we "upgraded" to a 13.3” MBP, our new results seem to suggest a ceiling around ~800 MB/s. Promise tells us it sees slightly better performance on PCs, but the delta should only be a few percent at most. Our numbers require more Mac-specific testing, since we don’t know whether it was the update or our system to blame.

Promise Pegasus R4 8 TB

The four-bay version of Promise's Pegasus is more affordable, offered in 4 and 6 TB configurations. You'll pay $1150 for the 4 TB R4 (versus $1800 for the entry-level 6 TB R6). Aside from its lower price and two-fewer drive bays, there's very little else distinguishing the R4 and R6 families. Promise even uses the same Hitachi Deskstar hard drives in them both.

Crack open the R4 and you find the same motherboard and power supply seen in the R6.

Eliminating two drive bays from the R6 does impact performance, though. In RAID 0, the R4's sequential reads top out at ~635 MB/s, representing a closet-to 33% drop from the R6. Sequential writes in RAID 0 only drop about 18%, from 635 MB/s to 535 MB/s.

RAID 1E and RAID 5 performance are mostly unchanged. The only exception is that sequential reads in RAID 5 peak at ~460 MB/s, about a 70% drop from the R6’s 780 MB/s.

Preview: Promise Pegasus R4 (SSD Version)

Promise plans to release an SSD-based version of the R4, but has not yet decided on a specific drive model to use yet. We are told that Intel's 240 GB SSD 520 is a strong contender due to the company’s stellar reputation for reliability and the zippy performance delivered by SandForce's second-generation controller.

Promise does not anticipate selling an SSD-based R6, stating the added cost of six SSDs would would push the price of the R6 beyond what the market will bear. We don't have an estimate for what the R4 with solid-state storage will cost, either, so we were forced to create our own estimate. Using the current 4 TB R4 as a baseline, we assume the 1 TB Hitachi 7K1000.D Deskstar hard drives are about $100 each, with the chassis and components totaling about $750. A single 240 GB Intel SSD 520 runs just north of $300, so a four-SSD variant of the current R4 could conceivably weigh in under $2000.

What this quad-SSD R4 variant might offer will depend on what you compare it to. Next to the R6, its benefits are likely to be limited. In theory, sequential performance should speed up by 5-10% (50-100 MB/s), while random I/O improves by an order of magnitude. Achieving better random throughput is important for certain applications that a hard drive-based solution simply cannot address. However, LaCie's Little Big Disk is more effective with its two SSD 320s than Promise's R4 armed with four SSD 520s when it comes to random performance.

The SSD-based R4's throughput matches what you can achieve using a hard drive-based R6. However, that performance gain costs lots of capacity. Even sporting 240 GB drives, the R4 tops out under 1 TB of space. That's nowhere near enough room for data-dense multimedia applications that chew up multiple terabytes.

Perhaps the characteristics of solid-state storage make this particular R4 more attractive, though. If you have production equipment operating in a vehicle, for example, or are constantly moving drives from one place to another, SSDs prove far more resilient against physical shock. They also run a lot cooler and use a lot less power.

Seagate GoFlex Desk Thunderbolt 3 TB

Seagate recently announced its Backup Plus, a refresh of its external GoFlex storage line-up. As described by Seagate, the company's new Backup Plus drives feature the Seagate Dashboard, a one-click backup interface supporting both Macs and PCs that also enables sharing and saving photos to and from popular social networking sites.

In concert with the Backup Plus introduction, we're also seeing a refresh of Seagate's product bundles, including the 3 TB GoFlex Desk Thunderbolt that includes a Backup Plus hard drive. Aside from new software features, however, very little changes. Seagate is still using its same generation of Barracuda 7200.14 disks. 

Don't feel like you have to wait for one of the new bundles to come out before making a purchase, though. If you look at the prices, it's a little cheaper buy the Thunderbolt adapter after grabbing one of Seagate's older offerings compatible with USB 3.0, for instance. In fact, you don't really even need a GoFlex hard drive to use with the Thunderbolt adapter. Because Seagate employs a standard SATA connection between the two devices, you could even drop a bare disk on the dock and it'd work. To prove our point, we even installed a Blu-ray drive onto the adapter.

Internally, the GoFlex Desk Thunderbolt Adapter consists of a motherboard hosting ASMedia's ASM1061 SATA controller and a PCI Express x1 slot. Intel's CV82524EF/L Thunderbolt controller is soldered onto an add-in card that drops onto the small board. Unfortunately, you can't yank out the x1 card and plug it into your PC. We tried, failed, and then asked. That's just not how it works.

Because we're measuring the performance of a single hard drive, the GoFlex Desk Thunderbolt's sequential read and write performance levels off at ~190 MB/s, regardless of queue depth.

Seagate GoFlex Ultra-Portable Thunderbolt 1 TB

Similar to Elgato's Thunderbolt SSD, Seagate's GoFlex Ultra-Portable Thunderbolt Adapter employs an Intel DSL2210 controller, meaning it can only be used at the end of a daisy chain or on its own. The good news is that it only costs $100, making the portable adapter a lot less expensive than the desktop version.

The Ultra-Portable Thunderbolt bundle is also in the middle of a product refresh. But it is worth noting that Seagate’s modular product formula simplifies an upgrade down the road. You can buy the storage now (in the form of a USB 3.0-equipped drive) and upgrade to Thunderbolt later when the technology stands to benefit you more.

Based on our performance numbers, it is frankly difficult to imagine buying the GoFlex Ultra-Portable Thunderbolt Adapter as an upgrade. Sequential read and write speeds top out around 110 MB/s, which is already roughly on par with the version sporting USB 3.0.

We can more easily imagine using this adapter in conjunction with a fast SSD, though. If you were attaching a solid-state drive to a desktop, you'd simply connect it internally to a SATA port. But most mobile systems don't give you that option. Thus, the GoFlex Ultra-Portable Thunderbolt adapter becomes an easy way to add a very fast 2.5” SATA-based device to a Thunderbolt-enabled notebook, while achieving near-native performance.

Western Digital My Book Thunderbolt Duo 6 TB

Like LaCie’s 2big, Western Digital’s My Book Thunderbolt Duo gives you the option to replace internal storage without invalidating its warranty.

We have some seen confusion surrounding Western Digital's cooling design, though. Some users complain about its plastic enclosure, claiming it runs too hot. But company representatives say that's by design. A small fan draws in cool air from the bottom of the unit, and convection serves to cool the internal components before venting out the top. This helps explain why the My Book Thunderbolt Duo's exhaust temperatures are routinely 10-15^o F higher than LaCie’s 2big, which also enjoys the benefit of an aluminum enclosure for dissipating heat. Its mechanism explained, the only drawback to Western Digital's design is that you can't stack drives on top of each other (or other components on top of a My Book).

Assurances from Western Digital aside, we stress-tested these drives with more than 100 hours of intensive write operations to try getting them to pop, and didn't see a single failure. Although they don't have the 2big's sexy aluminum enclosure, a plastic housing is less expensive to manufacture, reflected in the My Book Thunderbolt Duo's price tag.

Internally, Western Digital's implementation differs from LaCie's in that it employs ASMedia's ASM1061 SATA controller.

Sequential read and write performance tops out just over 240 MB/s in RAID 0 and 125 MB/s in RAID 1.

Test Setup And Benchmarks

When it comes to Thunderbolt and storage performance, A/V professionals should be particularly happy with the interface. As you've seen in our product spotlights, sequential throughput is quite often exceptional. By emphasizing the speed at which 128 KB blocks of data can be pushed through the interface, we got an accurate picture of how each submission is able to do its job.

It's not our intention to shun random-access performance entirely. However, none of these solutions are going to deliver impressive IOPS rates. If you don't believe us, check out the second page of Everything You Need To Know About Thunderbolt. A typical 3.5” rotating hard drive  is good for somewhere between 200 and 300 IOPS in random read/write workloads, which is less than 1 MB/s when you're dealing with 4 KB blocks. Even when you stripe six 3.5” desktop-class drives, the result is less than 10 MB/s of random reads.

That's not a problem with SSDs, which are capable of up to 80 000 IOPS, in the case of a Vertex 3. That's why we only presented random benchmark results for the SSD-based Thunderbolt solutions.

Thunderbolt is operating system-agnostic. But because Apple had a one-year head-start with the technology, more of its platforms are already equipped with the requisite controller hardware to support Thunderbolt. We ran our benchmarks on both Macs and PCs, achieving identical performance except where we noted otherwise. There are some PC-oriented problems tied to hot-plugging and daisy-chaining, but they don't affect performance, and Intel claims that it's working to help resolve them.

Installing vendor-specific SATA controller drivers delivered a ~1-3% performance boost compared to using Windows 7's generic ACHI driver.

Benchmark Results: Single File Transfer

Because Thunderbolt is so well-suited to the data-intensive needs of A/V professionals, we thought it fitting to measure the speed at which a 31 GB Blu-ray movie rip could be written to each device.

When it comes to raw sequential performance, a striped array communicating over Thunderbolt is formidable indeed. Promise's Pegasus R4 and R6 lead the pack, both units finishing the write task in 1:45. If you're willing to give up a lot of capacity in the name of solid-state storage, four SSD 520s in the R4 drop that completion time to 1:28.

LaCie’s 2big trails by only 10%, but it impressively outperforms the company's own SSD-equipped Little Big Disk. That shouldn't come as a surprise, though. According to Iometer, the 2big's hard drives outmaneuver the SSD 320s by ~15-25 MB/s in sequential read operations. There's a good lesson in there somewhere. Don't expect that, just because a device is armed with SSDs that it'll automatically move data around faster than an enclosure with hard drives.

Indeed, Western Digital's hard drive-based My Book Thunderbolt Duo matches the performance of LaCie's Little Big Disk, and the G-Technologies G-RAID Thunderbolt 8 TB follows behind by only a few seconds.

Even the performance of a single-drive Thunderbolt-based device should impress the folks who aren't interested in a multi-disk RAID enclosure, best illustrated by Seagate’s GoFlex Desk Thunderbolt Adapter’s small victory over the USB 3.0 version.

It's particularly notable that even a Thunderbolt link with multiple peripherals daisy chained onto it has sufficient bandwidth to enable full performance from downstream devices. Meanwhile, USB 3.0 slows down for every component added to a hub.

A quick breakdown of the single-drive observations:

• With a 3.5” hard drive, Thunderbolt proves faster than USB 3.0 by a small margin.
• With a 3.5” hard drive, the benefit of Thunderbolt compared to USB 3.0 increases as you add devices to the same interface.
• With a slower 2.5” hard drive, Thunderbolt performs about on par with USB 3.0. A daisy chainable device is needed to demonstrate an advantage.
• Regardless of the hard drive, Thunderbolt performs faster than FireWire 800 and USB 2.0 by huge margins.
• Do not assume an SSD in an external enclosure is going to be faster than hard drives.

Benchmark Results: Folder Transfer

We focused most of our attention on the performance of sequential transfers. However, not all sequential tests are the same. You might move a large move file 128 KB at a time. But a folder filled with different-sized files can still be moved sequentially using smaller blocks, too. The performance you see from such a mixture is not necessarily consistent with what we just saw writing a huge Blu-ray rip to each Thunderbolt device.

Testing our RAID 0-based enclosures, our rankings are largely the same as the previous page. The only exception is LaCie’s Little Big Disk, which outperforms the 2big, G-RAID Thunderbolt, and WD My Book Thunderbolt Duo.

Thunderbolt's benefits become clearer using single-drive devices with 3.5" disks. Moving a single file to the GoFlex Desk resulted in similar performance from Thunderbolt and USB 3.0. However, USB doesn't support command queuing. As a result, Thunderbolt takes a 22% lead when transferring bulk files. This is also an advantage when it comes to editing video, as data caching behaves similarly.

We see Elgato’s Thunderbolt SSD surge back because it's no longer getting hammered by incompressible data. Consequently, it matches the performance of Seagate's GoFlex Ultra-Portable (Thunderbolt or USB 3.0).

Thunderbolt: Faster Than USB 3.0; Three Winners Emerge

We know that Thunderbolt's maximum sequential throughput is close to the 1 GB/s mark, and Promise's Pegasus R6 has the unique distinction of saturating this new interface using six 3.5" hard drives. But it'd be wrong to assume that it takes $2000 to get the most out of Thunderbolt. Unlike USB, Thunderbolt was designed so that multiple devices can operate in parallel and still achieve their peak performance. You can get that same 1 GB/s from three or four peripherals doing 250 to 300 MB/s each.

This critically important advantage doesn't get enough attention. With USB, the interface bottlenecks performance, not the devices themselves. As an example, if you burn a DVD on a USB 2.0-based writer and write to a thumb drive, those conflicting operations could yield a useless coaster. Although USB 3.0 facilitates a lot more headroom, the technology is similar, and bottlenecks remain probable.

Thunderbolt alleviates interface limitations. Yeah, one LaCie 2big can "only" hit 350 MB/s in sequential reads. But a second 2big in the same chain can hit 350 MB/s, too. You can use those two drives, plus something like BlackMagic’s Intensity Extreme (a Thunderbolt-based external capture device used by professionals to edit 1080p content) and not have any of those three devices stepping on each others' toes. If that matters to your application, then spending money on high-end hardware probably isn't your top concern.

For everyone else, though, Thunderbolt does still face a big pricing problem. Even one 4 TB 2big at almost $600 reflects heavy mark-up. If you have room for a quartet of 1 TB disks inside your case, they'd only run about $360.

Those fat premiums will hopefully slim down over time. But even at a more attractive price, several of the devices in our round-up are haunted by other issues that call into question their utility. Elgato’s Thunderbolt SSD, for example, suffers from poor throughput because it employs an SSD based on SandForce's first-gen controller technology. That's fine for moving personal data around. But it's far less adept at handling encoded media files. Surely, a $420 price tag for 120 GB of capacity doesn't help. Seagate’s GoFlex Ultra-Portable Thunderbolt Adapter is only able to operate as an endpoint device. You can use it in a daisy chain, but you have to give up the potential for outputting to a display with a mini DisplayPort adapter.

Western Digital’s My Book Thunderbolt Duo competes directly with LaCie's 2big, but is hurt by slower drives. The thing is, the My Book Thunderbolt Duo costs $50 less than the 2big, which is a reasonable trade-off, so long as you're also able to accept the plastic case. They both offer notable value, earning LaCie's 2big 6 TB and Western Digital’s My Book Thunderbolt Duo 6 TB our 2012 Recommended Buy Award.

If you'd rather manage just a single Thunderbolt-based device (count us in), you can't ignore the incredible performance enabled by Promise's Pegasus R4 and R6. The SSD version of the R4 isn't final yet, but we're certainly impressed with our early preview.

Performance and value are still vital criteria, though. The least-expensive Pegasus R4 costs 33% less than the R6, and sacrifices the same margin of sequential read performance. However, both units serve up identical sequential write speeds, making the R4 a better value. That also earns Promise's Pegasus R4 our 2012 Recommended Buy Award.

According to a press release, Seagate will have its Pulsar.2 enterprise SSDs with a 12Gb/s SAS on display. The company said that the devices target cloud services and other enterprise applications "throughout all tiers of data storage." Seagate stressed that the interface will be initially integrated with SSDs to take advantage of "the high performance benefits of SSDs", but the technology is also compatible for use in higher-capacity enterprise hard drives. There was no announcement when 12 Gb/s HDDs will become available, but this seems to be a matter of when, not if.

Seagate noted that its 12 Gb/s implementation is "fully compatible with controllers and other technologies from PMC-Sierra and LSI." The company expects the 12Gb/s SAS standard to be finalized by the end of this year. First drives will be available sometime in 2013, while volume availability is predicted for H2 2013.

HGST announced last week that it will be demonstrating 12 Gb/s drives at the SCSI Trade Association’s Technical Showcase.

Hitachi Global Storage Technology (HGST), which was recently acquired by Western Digital, claims to be first with a demonstration of the first 12Gb/s SAS SSD. The company will be unveiling the device at the SCSI Trade Association Technology Showcase, which take place in Santa Clara on May 9.

As the name indicates, the drive is theoretically capable of transferring data at 12 Gb/s - twice the rate of current 6 Gb/s drives - which translates to a total bandwidth of 4.8 GB/s.

“We have successfully achieved interoperability between our 12Gb/s SAS drive and 12Gb/s SAS HBAs and expanders from both LSI and PMC-Sierra,” said Brendan Collins, vice president of HGST product marketing, in a prepared statement. “Meeting these interoperability milestones is critical when preparing the industry for the adoption of a new interface standard."

HGST said that it expects market adoption of 12 Gb SAS drives in 2013.

Three And 4 TB Hard Drives For Your Digital Lifestyle

As we were preparing to update our Hard Drive Charts for 2012, we started sending out requests for high-capacity disks. We were taken by surprise when we saw that Hitachi Global Storage Technologies provided not one, but two different 4 TB samples.

Although Seagate is also selling its 4 TB external GoFlex, Hitachi is the only vendor with an internal drive available at retail. Western Digital hasn't started offering a 4 TB internal disk yet. Samsung sold its hard drive business to Seagate. And Toshiba’s 3.5” drives are more enterprise-oriented.

Nevertheless, there are plenty of 3 TB disks we can use for comparison in a more comprehensive round-up of today's high-capacity storage devices.

Why Manufacture Two 4 TB Drives?

The way Hitachi names its products is pretty self-explanatory. The Deskstar 7K4000 is a 4 TB drive that spins at 7200 RPM, while the Deskstar 5K4000 operates at 5400 RPM. They both reach their capacity using five-platter designs, meaning that these drives are mechanically more complex than most 2 TB hard drives, which typically only employ three platters. However, Hitachi has been building drives with higher platter counts for a while, giving the company quite a bit of experience at managing the challenges more platters introduce.

But does it really make sense to build two different 4 TB drives? And what practical difference is there between them in real-world testing, knowing that performance-sensitive enthusiasts would rather spend money on an SSD for their Windows installation plus an additional storage hard drive rather than just one 4 TB monster selling for a premium?

Hitachi created the Deskstar 7K4000 with 64 MB cache for the power user segment demanding lots of capacity and a minimum level of performance for workloads like video editing, semi-portable storage solutions, and video surveillance servers. The 4 TB capacity point is several times larger than the biggest SSDs, which incidentally cost several times more than Hitachi's 4 TB hard drives.

The Deskstar 5K4000 has 32 MB cache and is designed for data storage where large amounts of information need to be available on file servers, NAS solutions, and near-line applications. Low power consumption is touted as a benefit of the 5K4000 that helps manage heat and cooling in larger deployments, too.

Performance and Cost

The Deskstar 5K4000 is currently available for $300 in the U.S. (260 Euros in Europe), while the 7K4000 costs a little under $350 (340 Euros) if you can find it (availability seems to be an issue for Hitachi still).

You'll want to put those prices into perspective. Hitachi's own 3 TB 5K3000 starts at $220, and many 2 TB hard drives sell for $120. So, you could almost get three 2 TB disks for the price of one 4 TB drive. A compelling cost per gigabyte clearly isn't one of the reasons you'd want to shop for the highest-capacity disk available. Instead, we'll have to look at the performance and efficiency of these storage products.

We're throwing in Hitachi's Deskstar 5K3000, Deskstar 7K3000, Seagate’s Barracuda 3 TB, Barracuda XT 3 TB, and Western Digital's Caviar Green 3 TB up against the 7K4000 and 5K4000 to gives you a comprehensive picture of today's 3 and 4 TB hard drives.

3 TB: Hitachi Deskstar 5K3000 (HDS5C3030ALA630)

We first looked at the Deskstar 5K3000 when we compared various 3 TB hard drives last fall. Check out Four 3 TB Hard Drives, Tested And Reviewed for more information. Briefly, the 5K3000 is a five-platter drive (just like the new 4 TB version). It features 32 MB cache and a SATA 6Gb/s interface.

The rated internal transfer rate of 1366 Mb/s (170 MB/s) is actually faster than the new drive's specified speed of 1293 Mbt/s (162 MB/s). In the real world, however, both drives actually offer very similar throughput, which hovers around 130 MB/s at best. Hitachi's 4 TB drive even facilitates better average results at slightly improved peak and minimum transfer rates.

When it comes to I/O performance, though, the Deskstar 5K3000 is slightly better than the newer 4 TB drive, although this discipline isn't particularly relevant for 5400 RPM drives. This 3 TB disk also enjoys a slight advantage over the larger drive in PCMark 7, even if the differences are small enough to be ignored.

If you’re looking for a high-capacity drive and you can’t decide between the 3 and the 4 TB models from Hitachi, buy the model with the lowest cost per gigabyte (almost certainly the 5K3000) if that fourth gig won't be missed too much.

3 TB: Hitachi Deskstar 7K3000 (HDS723030ALA640)

Hitachi’s second 3 TB drive spins at 7200 RPM and has 64 MB of cache memory. As a result, its specified throughput is much higher: 1656 Mb/s instead of 1366 Mb/s (that's 207 versus 170 MB/s). Real-world throughput reaches up to 152 MB/s according to h2benchw and 148 MB/s in our Iometer streaming reads test. These are commendable results that are only beaten by the new Deskstar 7K4000 and Seagate’s Barracuda 3 TB, which actually gets close to 200 MB/s.

The 7K3000 family is also available in 2 TB and 1.5 TB flavors, and they’re all specified for 24/7 operation. Power consumption is higher on this 7200 RPM drive than on Hitachi's 5K3000. That's simply the price you have to pay for higher throughput. We measure 4.9 W at idle for the 5400 RPM 3 TB drive and 7.1 W at idle on this 7200 RPM model. The 4 TB versions have very similar power requirements.

Since the performance difference between the 3 and 4 TB drives isn't too significant, we again have to recommend that anyone shopping for a high-capacity disk consider living with the 3 TB model if the last gigabyte isn't absolutely imperative. You simply spend a lot less money per gigabyte when you compromise down to the smaller drive.

You can find comprehensive hard drive performance comparison in our 2012 Desktop HDD Charts.

3 TB: Seagate Barracuda (ST3000DM001)

Seagate’s Barracuda brand has been around for 15 years. The first Barracuda drive was also the first disk with a spindle speed of 7200 RPM (though it was only offered in 1.2 and 2.5 GB capacity points sporting a SCSI-2 interface).

Today’s Barracuda sheds all of the product name suffixes that Seagate has used over there years. There was a Barracuda ATA family (from I to V), Barracuda 7200.7 to .12, Barracuda LP at a power-optimized spindle speed (5900 RPM), the Barracuda Green, and finally a Barracuda XT, which was one of the first SATA 6Gb/s-capable models available. Various versions of the Barracuda XT and LP still exist, but the latest family covering all capacity points is simply referred to as Barracuda after all of these years. Our 3 TB sample sells for around $170 (140 Euros in Europe).

There are 250, 320, 500, 750 GB, 1, 1.5, 2, and 3 TB capacities available. Everything from 750 GB and up comes with 64 MB cache memory; the smaller capacities have to live with 16 MB and target low-cost applications with single-platter designs. Perhaps the Barracuda's most interesting attribute is its 1 TB per platter density. This means that the 3 TB drive only uses three platters, while Hitachi's drive requires five to hit the same capacity point. As a result, Seagate's Barracuda 3 TB is the coolest 3 TB hard drive on the market.

One item in the Barracuda's data sheet related to reliability caught our eye. Although the document lists NAS and desktop RAID applications as best-fit applications for Seagate's offering, the specified 2400-hour power-on rating indicates that this drive was not designed for environments requiring 24/7 availability. If that were the case, its specification should say 8760 hours. The 2400-hour specification, which represents only 100 days of continuous operation, assumes that drives are used eight hours per day, five days per week. This means that the drives should take full advantage of any available power-saving technologies that shut down the spindle motor during idle times. This applies equally to desktops and storage servers. Should you decide to use Barracuda drives in a NAS device, for instance, don’t forget to configure its power management appropriately.

3 TB: Seagate Barracuda XT (ST3300651AS)

Barracuda XT was Seagate’s first 7200 RPM 3 TB hard drive, and the 2 TB model was the first drive equipped with a SATA 6Gb/s interface. All Barracuda XTs come with 64 MB of cache. They also have a much lower specified annual failure rate of 0.34%, whereas the regular Barracuda is specified at a <1% AFR. If you’re looking for a drive that needs to be able to operate 24/7, go for the Barracuda XT instead of Seagate's vanilla Barracuda. Just be aware that it bears a $250 price tag (210 Euros in Europe), too.

However, it is important to point out that the Barracuda XT is older than the Barracuda discussed on the previous page, which helps explain why it's also slower. A maximum throughput of 158 MB/s almost matches the 163 MB/s achieved by Hitachi's Deskstar 7K4000, but it doesn't even come close to the aforementioned Barracuda's excellent 193 MB/s result. Barracuda XT employs five platters, which means that it gets hotter than the Barracuda. But it's still slightly cooler than the Hitachi drive.

You'll probably want to avoid the 2 TB Barracuda XT, which is quite a bit slower than the 3 TB model reviewed in this article. Please check out our 2012 HDD Charts for more details.

3 TB: Western Digital Caviar Green (WD30EZRX)

The only 3 TB hard drive from Western Digital is a Caviar Green model, which aims at energy efficient systems. The drive is different from all the others in that it pairs a 5400 RPM spindle speed with 64 MB of cache and 3 TB of capacity. The company’s strategy includes the VelociRaptor drive, which recently hit 1 TB. That's WD’s performance family, which clearly outperforms any other mechanical hard drive aside from some of the enterprise-oriented models out there. So, a line of 7200 RPM 3.5” disks simply wouldn't make as much sense for Western Digital.

The Caviar Green drives come with a two-year warranty, while the Caviar Black and VelociRaptor families are backed by five-year coverage. This is similar to Seagate’s approach to the Barracuda and Barracuda XT. The Caviar Green is the slowest hard drive in this round-up in terms of throughput (122 MB/s max), but also in many of the I/O tests. Does that matter? If you're only leaning on the drive for archiving data, not really. But it would if you planned on running applications from it dependent on high throughput.

The Caviar Green WD30EZRX starts at around $170 (160 Euros in Europe). If you don't need 24/7 availability, you might want to consider Seagate's Barracuda as an alternative. It runs cooler and delivers more performance. It isn't designed to run all day and night, just like this Western Digital disk.

4 TB: Hitachi Deskstar 5K4000 (HDS5C4040ALE360)

The new Deskstar 4K4000 offers 4 TB of capacity with a 5400 RPM spindle speed. It employs a SATA 6Gb/s interface and 32 MB cache, which is pretty much standard today. Like the prior xK3000 generation, the 4K5000 is based on a five-platter design, meaning that we shouldn't expect to see the lower power consumption. However, it does, in fact, demonstrate remarkable power attributes. The Deskstar 5K4000 requires the least power of all 3+ TB hard drives at idle and under common workloads. Unlike Seagate’s and Western Digital's 3 TB offerings, the Deskstar 5K4000 is designed for 24/7 operation.

Hitachi specifies an idle power requirement of 4.9 W, which is pretty impressive when you consider that this is a five-platter design. A media transfer rate of 161 MB/s is fairly nice as well, translating to a 131 MB/s net transfer rate, measured by h2benchw. Seagate’s 2 TB Barracuda LP is a faster option, though, thanks to its 5900 RPM spindle.

We like the fact that Hitachi is not trying to optimize its drives for I/O performance anymore. This is a discipline ruled by SSDs now, outperforming hard drives in each and every possible benchmark by up to 100x. Instead, Hitachi creates a typical high-capacity storage drive that delivers nice throughput and decent robustness.

Hitachi Deskstar 7K4000 (HDS724040ALE640)

This is Hitachi's high-performance 7200 RPM 4 TB drive. It's also based on a five-platter design, but it includes 64 MB of cache, instead of 32.

Power consumption in performance-idle mode (which means the drive is still spinning) is rated at 6.9 W. That's 40% more than the Deskstar 5K4000. In exchange, however, rated transfer rates increase from 161 MB/s to 205 MB/s, representing at least a 24% increase. The effective net performance is 163 MB/s. Although this is a solid result, it cannot match Seagate's Barracuda at 194 MB/s.

The Seagate drive isn't designed for 24/7 availability though, and the 7K4000 is. In other words, this Hitachi disk is the fastest high-end hard drive for performance-oriented applications that require lots of throughput. As far as I/O performance is concerned, let's just say we'd recommend looking for SSDs instead of disk drives.

Test Setup And Comparison Table

Comparison Table

Benchmark Results: Transfer Diagrams

Data Transfer Diagrams: 3 TB Hard Drives

Data Transfer Diagrams: 4 TB Hard Drives

Benchmark Results: Interface And Throughput

Interface Performance

The interface test reflects the maximum throughput that each drive’s interface is capable of handling. Typically, this equals the peak read speed from a disk's built-in cache, since we know that the physical media is limited to less than 200 MB/s. With this in mind, SATA 3Gb/s would still be fast enough to accommodate every drive in this round-up. SATA 6Gb/s is really an enabler for solid-state storage.

Sequential Read/Write Throughput

The fastest hard drive in terms of throughput is Seagate’s new Barracuda 3 TB (ST3000DM001) at up to 193 MB/s. Quick though it may be, it's not designed for 24/7 availability. For that, you'd need one of the Hitachi drives or Seagate's 3 TB Barracuda XT, all of which provide similar throughput.

Benchmark Results: Access Times And 4 KB Random I/O

Access Time

Hard drive access times do not improve much generationally. Typically, access times are limited by the hardware's ability to relocate read and write heads. 

You want a system drive able to access data quickly, which is why even smaller SSDs are perfect for hosting operating systems and frequently-used apps. As far as typical storage applications go, it doesn't matter much whether a given disk achieves 15 or 18 ms accesses.

4 KB Random I/O

One of the most important I/O benchmarks is a drive’s ability to handle 4 KB random I/O activity. Memory pages typically have a 4 KB size and hard drive sector sizes moved from 512 bytes to 4 KB, making this block size important.

Although we see significant differences between the various drives and their performance in 4 KB random I/O, the results remain outright pathetic compared to modern SSDs, many of which push tens of thousands of IOPS. Again, we can only conclude that these high-capacity drives are best for pure storage applications.

Benchmark Results: I/O Workloads

We're including our I/O workloads, which simulate database, file server, Web server, and workstation activity, simply for the sake of comparison and completeness. They really aren't all that relevant to the desktop-oriented hard drives in this round-up.

Benchmark Results: PCMark 7

If you simply cannot get your hands on an SSD as a system drive, at least make sure you buy a 7200 RPM disk. You clearly get better application-starting performance than anything a 5400 RPM drive can deliver. Although, once again, adding flash-based storage to the equation is the best way to go.

Benchmark Results: Temperature And Power Consumption

Drive Surface Temperature (Idle)

The most important variable in determining a disk's temperature is the speed of its spindle, followed by the number of platters it employs. That's only a general guideline, though. We get surprised by Hitachi's two five-platter models, which run slightly cooler than the three-platter Seagate Barracuda.

Power Consumption

The same general truth applies to power requirements, too. Although Seagate’s Barracuda 3 TB should show us the lowest idle power due to its three-platter design, both 5400 RPM Hitachi drives are actually better. The difference between the 7200 RPM models is rather small.

Intensive I/O workloads show the Deskstar 7Kx000 drives consuming close to 10 W, while the Seagate drives are better in this discipline.

At maximum continuous throughput, Seagate's Barracuda XT requires the most power, followed by two Hitachi drives and the new Barracuda 3 TB. All 5400 RPM models remain under 8 W.

Capacity Marches Forward, Commands A Premium

Hitachi could be the only hard drive manufacturer with internal 4 TB disks for a few months longer. Seagate has its own 4 TB model, but is only putting it into external enclosures. Western Digital's Caviar Green family could also receive a facelift soon, moving into the 4 TB space. Not surprisingly, capacities will continue marching forward, and that's a great thing for businesses and enthusiasts with burgeoning capacity needs.

In the end, everyone benefits when drive vendors engage in capacity wars, even if most of us only end up buying more mainstream disks. Seagate is a great example of this. While its 1 TB platters potentially enable seriously large drives, even smaller models benefit, coming in at lower prices, generally using less power, and often proffering better performance. Additionally, the ability to put 1 TB on a platter makes it that much easier for entry-level PCs to include at least that much storage space (the sweet spot) using a low-cost disk.

Performance? Seagate Barracuda

The fastest 3.5” high-capacity desktop hard drive in this round-up is Seagate’s new 3 TB Barracuda (ST3000DM001). It offers almost 200 MB/s sequential read performance, similar write speeds, and it runs relatively coolly thanks to its three-platter design.

However, Seagate didn't design this Barracuda for 24/7 operation. If you need higher availability, you'll want to check out Seagate’s Barracuda XT or Hitachi's Deskstar 7K3000. Both are five-platter designs, robust, and still fast enough for video editing or video surveillance servers. Three terabyte hard drives are probably the best compromise between capacity and cost, as 2 TB disks might not always be big enough and 4 TB models are still very expensive.

Capacity? Hitachi Deskstar

If you're willing to live with that sizable price premium in the name of massive capacity, there are only two options available to you right now: Hitachi's Deskstar 5K4000, a 5400 RPM drive, and the 7200 RPM Deskstar 7K4000. The latter is a solid all-around performer and it's rated for 24/7 availability. The 5K4000 understandably sacrifices performance, but it’s also one of the most efficient drives if you’re looking for low-power storage.

In that context, you could also consider Western Digital's Caviar Green, though it gives up even more performance and it's not specified for high availability.

Hard Drive Performance Evolves Slowly

This bears mention again: when we look at the performance of mechanical disks and the rate at which they improve, we're reminded that SSDs are disruptively faster and more responsive. New throughput records are being broken all of the time, and we're already butting up against the limits of today's highest-bandwidth desktop interface, SATA 6Gb/s. 

Meanwhile, even after picking up 20% more throughput, today's latest hard drives aren't even able to saturate a last-generation 3 Gb/s connection. When it comes to access times and I/O performance, neither metric speeds up at all compared to the last round of drives we reviewed. 

Simply, the mechanical arms equipped with read and write heads cannot be sped up and slowed down at will; there are physical limits. Trying to be more aggressive reduces the life span of a drive, and even a small gain won't help stave off SSDs in a performance comparison. The results of our testing clearly shows that hard drive vendors don't bother trying to optimize for access times or I/O anymore. It's futile.

This is not going to change, which is interesting because the industry won't be replacing magnetic hard drives with anything else in the predictable future. Disks will continue to provide more capacity at low cost, as today's computing experience is decreasingly limited by CPU or GPU power and more dependent on storage. Pushing I/O is the key there, meaning that any workload relying on frequent accesses should increasingly rely on flash-based devices, while hard drives safeguard our digital memories and databases.

Recommendations

Three terabyte hard drives continue to offer the best cost per gigabyte of storage space. Two terabyte drives offer even better value at similar performance levels. Four terabyte disks are a good option when you need big capacity and don't have space for multiple 3.5" devices. However, their price premium keeps us from recommending broader adoption.

Three Well-Built Machines Face Off

System Builder Marathon, March 2012: The Articles

Here are links to each of the five articles in this quarter’s System Builder Marathon (we’ll update them as each story is published). And remember, these systems are all being given away at the end of the marathon, including the Bonus Customer Choice PC, which we picked out using the highest-rated components in Newegg's feedback system.

To enter the giveaway, please fill out this SurveyGizmo form, and be sure to read the complete rules before entering!

Day 1: The $650 Gaming PC
Day 2: The $1250 Enthusiast PC
Day 3: The $2600 Performance PC
Day 4: Performance And Value, Dissected
Day 5: Bonus Newegg Customer Choice PC

Introduction

Without the benefit of an unlimited budget, improving performance in one area typically means making a sacrifice somewhere else. Recent changes to our benchmark suite slightly deemphasize the importance of beefy graphics at the high-end, since CPU bottlenecks hamper some of its games. On the other hand, Paul, the guy building our entry-level machine, usually can’t afford a fast enough graphics card to see those CPU bottlenecks. Those concepts drive this quarter's cheapest and most expensive systems in opposite directions. I dropped SLI from my $2600 machine, while Paul gave up a capable Core i5 processor from his $650 contender.

More controversial were the choices made by Don, who picked parts for our $1250 PC. He went with the same high-end graphics card as me in an attempt to match the expensive machine in at least the gaming segment of our metrics. Knowing that his affordable quad-core CPU couldn't stand up to my Sandy Bridge-E-based Core i7-3930K in the content creation apps, Don made his stand where he knew he stood a chance. Frequency (rather than core count) determines where most games run into a bottleneck, so Don pinned his hopes on topping-out Turbo Boost technology, since his Core i5-2400 is one of those "partially-unlocked" models.

So, we end up with two purpose-built gaming machines taking on a fully-loaded $2600 heavyweight in the productivity and content creation suites. That's definitely not going to be a fair fight. But you could also look at this match-up as one feature-oriented (overpriced) behemoth taking on two budget-oriented game systems in a battle for the best value. That may be equally unfair. We’ll do our best to present both sides of the performance/value debate, though.

Notice that the title for each build refers to its budget limit, not its actual cost. Two days ago, we saw Don call his configuration the $1250 build, though he did have another $50 available to him. Amounts left unspent simply contribute to each PC’s price-per-performance calculations.

Benchmark And Overclock Settings

The most controversial component in the entire round-up could be the $1300 PC’s 64 GB SSD. Although performance numbers for the system drive are intended to represent general use, our benchmark image weighs in at around 80 GB. Moving data, such as our workload files, onto the secondary drive does allow our programs to fit within a 64 GB envelope, and it'd be easy to argue that a tiered storage subsystem is designed to split information up exactly like that. Even then, though, it’s hard to imagine that anyone would intentionally operate their system drive near its full capacity all the time.

The $2600 build had drive issues of a different kind: Intel’s enterprise-oriented RST drive slowed performance so much that we reverted to Windows 7’s default AHCI driver for the overclocked test.

Benchmark Results: 3DMark And PCMark

3DMark helps quantify the performance of many different variables, including a system's CPU. The $2600 machine’s hexa-core processor establishes a significant advantage at the benchmark’s Entry preset. The $1300 PC’s big graphics card steps up its game at the Extreme preset. And the $650 machine struggles to keep up using lower-cost parts.

PCMark’s overall score is tremendously drive-dependent, slamming the $650 build for its use of a mechanical disk. The $2600 PC also looks somewhat bad at its default settings, but only because we were still using Intel's RSTe driver in that out-of-the-box configuration.

Reverting to Windows' AHCI storage driver gives the overclocked $2600 PC a second boost, though it never gets close to performing twice as fast as the $1300 machine.

A breakdown of a few PCMark tests helps demonstrate how well an SSD can perform. We also see where Intel's data integrity-oriented enterprise driver hampers performance, and how far mechanical hard drives trail behind. Because these tests approximate the loads we run on a frequent basis, they are also used in the hard drive rating of our average performance chart.

Benchmark Results: SiSoftware Sandra

Sandra's Arithmetic metric shows the $2600 PC generating scores more than four times higher than the $650 machine’s Core i3. If the rest of the system performed that well, the most expensive machine would finally break through barriers in value.

Unfortunately, we already know that the rest of the top machine's hardware doesn't yield four times the performance of Paul's nifty little build, and that’s where the $1300 machine could get its shot at a value crown.

Perhaps most surprising is the big lead Intel's Core i7-3930K establishes over its own Core i5-2400. We’d expect a 50% improvement for its 50%-higher core count, but it instead returns more than 100% on our investment.

The $1300 machine’s memory was hobbled by its low-end motherboard, which is really a shame since the $37 left over in Don's budget could have gone toward a much nicer platform.

Benchmark Results: Battlefield 3

Several sacrifices were made in order to get the $1300 build armed with a very high-end graphics card, all so it could hang with the $2600 PC in as many games as possible.

A bottom-end motherboard, a multiplier-locked processor, and a read-only DVD drive were all considered acceptable in this effort to give the big build a black eye. Did the compromises pay off?

Don's approach actually worked better than we might have expected, especially since the $1300 build’s graphics card overclocked better than mine.

The $650 build reaches its Battlefield 3 limit at 1920x1080, which is fine. Paul said from the outset that he'd be happy if his machine could provide playable performance at the highest native resolution most value-seekers are expected to use.

Benchmark Results: DiRT 3

The $1300 build continues pushing for big numbers in DiRT 3, actually beating the stock $2600 build at Ultra quality and 2560x1600. It’s too bad that neither its motherboard nor CPU are suitable for “reasonable” overclocking. Another couple hundred megahertz from Turbo Boost could have helped pushed frame rates even higher at CPU-bound settings.

Amazingly, the $650 build continues to demonstrate playable 1920x1080 results at our highest tested detail levels. Paul's mission appears accomplished!

Benchmark Results: The Elder Scrolls V: Skyrim

Overclocking is the easiest way to circumvent CPU bottlenecks in Skyrim, but the $1300 PC can’t overclock as well as the unlocked Core i7-3930K. A higher graphics card overclock helps that machine take the lead at 2560x1600, though. That's an intense-enough setting that a lower-end CPU doesn't even play a role in the benchmark result.

Meanwhile, the $650 PC is still playable through 1920x1080 at our highest test settings, making squabbles between the other machines appear petty. Paul must be mighty proud right about now.

Benchmark Results: StarCraft II

The $2600 build continues to use its big CPU to squash the $1300 build in gaming resolutions up to 1920x1080. But CPU bottlenecks again disappear once we hit the GPU-punishing 2560x1600 resolution. At that point a higher graphics card overclock allows the $1300 PC to score its strategic win.

Once again, the back and forth between pricey configurations seems silly to anyone who just wants to play this game at high graphics detail using a 1920x1080 resolution. The $650 survives its fourth and final graphics test.

Benchmark Results: Audio And Video Encoding

Apple iTunes and Lame MP3 rely on high clock rates because they're single-threaded. Fortunately for me, the Core i7-3930K is the only unlocked processor in the bunch, allowing me to hit 4.6 GHz and take the win. Granted, that's hardly a reason to celebrate, since I paid $600 bucks for one if Intel's most complex desktop processors.

The $2600 machine strikes back in multi-threaded video encoding apps, demolishing the $1300 build while chasing Paul's $650 PC out of the room.

Benchmark Results: Productivity

The $2600 PC achieves roughly 50% greater Photoshop performance compared to its $1300 rival. In light of its roughly 100% higher cost, the value proposition there doesn't look particularly inviting.

Don's $1300 build’s value also shines in comparison to the $650 machine, since it serves up nearly two times the performance at less than two times the price.

Unfortunately for Don, his $1300 configuration gets demolished in both performance and value by my $2600 build in 3ds Max, though my expensive machine fails to outperform Paul's $650 configuration by the expected four-to-one ratio that prices would dictate.

File compression is a mixed bag, since 7-Zip responds well to multi-core processors and WinZip doesn’t. WinRAR falls somewhere in the middle, and the dual-core $650 machine’s value shines overall.

After seeing so many staggering losses in office application apps, we begin to wonder if Don's broken motherboard is hammering its overall performance picture. Don't worry, though. We'll get the marginal motherboard replaced before anyone wins the $1300 configuration.

Although it appears slow, the $650 machine still manages more than one-quarter of the flagship build's performance.

Power And Efficiency

Power consumption is one place where slower dual-core processors shine, which gives Paul's effort a chance to top our corresponding chart, where lower numbers are better.

On the other hand, my Core i7-3930K consumes more than two times the idle power of Don's Core i5-2400, even though it only employs two additional cores.

Since efficiency compares work to energy, we first collect performance averages for all tests to determine how much more gets done on the faster systems. Using the slowest configuration as our 100% baseline makes this measurement easy.

Dividing average performance by average power would give us average efficiency, but we weighted our performance charts to account for the 10% maximum amount of time a user might spend waiting for various files to load. The baseline from these calculations starts at 100%, but since nothing is 100% efficient we move the baseline to 0% by subtracting one from the results.

Surprisingly, the $2600 PC actually reaches baseline efficiency when it's overclocked, even though moderate CPU voltage increases were required. The $1300 build shines brightest however, combining an efficient Sandy Bridge quad-core processor with AMD’s moderately-efficient Radeon HD 7970 graphics card.

Three Different Goals, One Value Conclusion

As I suggested yesterday, when it comes to building performance-oriented PCs, most enthusiasts start seeing some form of diminishing returns after spending between $600 and $800. That’s because less-expensive parts generally offer far less performance, while better-performing parts generally cost far more money.

This quarter, however, Paul took a big risk by pairing an overpriced performance-oriented part (his graphics card) with a platform that wouldn't overclock at all. We understand where he was going there. More important than simply winning our Marathon, he wanted to give gamers a build that'd manage playable 1920x1080 frame rates, even if it meant getting wasted on the productivity apps.

His approach could have lost him the value competition, had Don not tried the same strategy with a more generous budget. In retrospect, we have to wonder how much more value he could have extracted from a slightly more expensive motherboard (with fully-functional memory) in the mid-priced machine.

Meanwhile, my $2600 build creates value from some of its overpriced components by overclocking well. We're left wondering if sacrifices on the flashy cooler, quiet case, and Blu-ray writer could have been parlayed into even better go-fast hardware, which may have overcome the value enabled by Don's machine. Of course, then I'd have to hear it from everyone in the comments section poking fun at cheap-looking components in a high-end build.

Don certainly proved his point when it comes to graphics performance. He used the same card as me to achieve similar performance at 2560x1600. A little luck with GPU overclocking gave him the upper hand, even, bringing his $1300 machine within 2% of my higher-end box.

But few of us could recommend the $1300 system to a friend or family member. Even Don seemed pretty worried about the quality of his finished product. We recommend that anyone considering a revamp of that build at least read a few of our motherboard reviews before making a revision. Consult with our forum members on parts that haven’t been reviewed yet. And consider whether the recently-released (and less expensive) GeForce GTX 680 might yield even better performance for your money.

Giving It Up For More Gaming Performance

System Builder Marathon, March 2012: The Articles

Here are links to each of the five articles in this quarter’s System Builder Marathon (we’ll update them as each story is published). And remember, these systems are all being given away at the end of the marathon, including the Bonus Customer Choice PC, which we picked out using the highest-rated components in Newegg's feedback system.

To enter the giveaway, please fill out this SurveyGizmo form, and be sure to read the complete rules before entering!

Day 1: The $650 Gaming PC
Day 2: The $1250 Enthusiast PC
Day 3: The $2600 Performance PC
Day 4: Performance And Value, Dissected
Day 5: Bonus Newegg Customer Choice PC

Introduction

Up until now, we've used Intel’s Core i5-2500K in our mid-range enthusiast builds without ever having been disappointed. We gave AMD’s lower-priced FX-6100 a chance last time around, but that build wasn’t very successful. Despite its impressive graphics subsystem (a powerful pair of Radeon HD 6950 cards in CrossFire), it just couldn't compete.

I suppose we could crawl back to the Core i5-2500K, begging for forgiveness. It's the safe bet, after all. But we're tired of making it, and we're never going to learn anything new if we go down that road again.

No. This time we’ll try something a little different, and we’ll give up some of the processor and platform budget to make room for a shiny new Radeon HD 7970. We think the result has the potential to be a very capable gaming machine.

Prices haven't changed much between the time we ordered our parts and now, although Crucial’s m4 SSD did go up by $5. The original cost of this box ducked in just over our $1250 budget. Now we're slightly north of $1260.

CPU, Motherboard, And Cooler

CPU: Intel Core i5-2400

With our sights set on a $550 Radeon HD 7970 for our $1250 build, we have to cut some corners. And it's not like the $190 Core i5-2400 is a bad CPU. After all, we’ve been recommending it in the sub-$200 space for some time now. Moreover, in our recent sub-$200 CPU gaming comparison, this processor performed very closely to the Core i5-2500K overclocked to 4 GHz.

Read Customer Reviews of Intel's Core i5-2400

Although the Core i5-2400’s multiplier ratio is locked, we’re interested in seeing how much speed this 3.1 GHz model gains when we peg it at its highest available 38x Turbo Boost ratio.

Motherboard: ASRock P67 Pro3 SE

Read Customer Reviews of ASRock's P67 Pro3 SE

We heard good things about ASRock’s P67 Pro3, so we thought we’d give the SE version a try. At $95, it certainly didn’t hurt that this was also the lowest-priced motherboard we could find with Intel's P67 chipset.

On paper, it has everything else we need: DDR3 memory support up to 2133 MT/s and a PCIe x16 slot. Reality turned out to be a little different though, as we suffered from from a BIOS limitation that we’ll discuss in the overclocking section. Know this: if you plan on overclocking, spend a few more dollars on a platform with more complete firmware settings. If you’re not overclocking, you might appreciate the Pro3 SE’s thin footprint, which fit into our mid-tower enclosure with ease.

CPU Cooler: Cooler Master Hyper TX3

Read Customer Reviews of Cooler Master's Hyper TX3

We know that there are better coolers out there, but you’ll have a hard time finding one for $20. Given tight budget constraints, we opted for Cooler Master’s low-budget option. It just so happens to be a decent model for the price, and it's significantly better than the solution you get bundled with Intel's Core i5-2400. 

Video Cards, Power Supply, And Case

Video Cards: PowerColor Radeon HD 7970

This card is the reason that we’re scaling back on everything else in today's build. If you're a gamer, though, the sacrifices are worthwhile.

PowerColor's Radeon HD 7970 is a very fast board with modest power requirements. It's well worth $550 if you're looking for playable frame rates at high resolutions. The company doesn't skimp on bundled adapters, including mini DisplayPort-to-DVI, a mini DisplayPort-to-DisplayPort, HDMI-to-DVI, and DVI-to-VGA adapters in its bundle. The card also comes with DiRT 3, which happens to be one of the games in our updated SBM benchmark suite.

Read Customer Reviews of PowerColor's Radeon HD 7970 3 GB

Power Supply: Corsair TX650 V2

Read Customer Reviews of Corsair's TX650 V2

We keep waiting for something better than Corsair’s TX650 to come along, but, for less than $100, it hasn't happened yet. As such, we're happy to use this popular $90 power supply once again.

PC Case: Apevia X-Trooper Junior

Read Customer Reviews of Apevia's X-Trooper Junior

Operating within a tight budget, I'll always favor go-fast hardware over a fancy chassis, and I do enjoy giving inexpensive enclosures a shot.

To that end I opted for Apevia’s X-Trooper Junior, a small mid-tower case that costs a mere $40. It is surprisingly roomy inside, with enough clearance for our beefy Radeon HD 7970. Two included 120 mm fans are particularly nice touches, particularly in this price range.

Memory, Hard Drive, And Optical Drive

Memory: Mushkin Enhanced Redline 8 GB (2 x 4 GB) DDR3-1600 Memory Kit 

Frankly, I’d be happy with 4 GB of lower-latency memory. But your feedback tells us that you want to see 8 GB in these builds. At $60, this Mushkin Enhanced dual-channel kit boasts 7-9-8-24 timings at a 1600 MT/s data rate. 

Read Customer Reviews of Mushkin's Enhanced Redline 8 GB DDR3-1600 Kit

SSD: Crucial m4 CT064M4SSD2 2.5" 64 GB

Read Customer Reviews of Crucial's 64 GB m4

I’m not a fan of restrictive 32 GB boot drives, but Crucial’s M4 64 GB SSD delivers two times that capacity, along with great value and performance.

As with any SSD, remember to check for firmware updates, as it seems many manufacturers (including Crucial) end up fixing stability issues once the drives are on store shelves. We updated our m4 to version 0309 before running any tests.

Hard Drive: Hitachi Deskstar 7K1000.D 750 GB

Read Customer Reviews of Hitachi Deskstar 7K1000.D 750 GB

Hard disk prices are still much higher than they were last year, and because we’re on a strict budget, we have to make every dollar count.

Now that this system boots from an SSD, we only need a conventional hard disk for  general storage. Hitachi’s $100 Deskstar 7K1000.D offers 750 GB of capacity, it spins at 7200 RPM, and includes a 32 MB cache.

Optical Drive: Samsung TS-H353C OEM

Read Customer Reviews of Samsung TS-H353C

We also chose the cheapest optical drive we could find. This particular example is a real-only device, and not a burner. So, if you need to write your discs, you'll want something else.

Samsung's TS-H353C has been out of stock for a while, so simply look for the least-expensive drive to take its place in your own build.

Building And Tweaking Around A Radeon HD 7970

When cost is the primary concern, parts rarely seem to fit together ideally. So, we were really surprised at how easily ASRock's P67 Pro3 SE and PowerColor's Radeon HD 7970 fit inside the Apevia X-Trooper Junior case. It turns out that the small enclosure is extra wide and designed specifically to accommodate long graphics cards. A little bit of flexibility in hard drive mounting makes things even simpler. Also, the fact that ASRock's board is skinnier than more ATX platforms helped quite a bit. It’s been a while since I’ve worked with a case this small that didn’t cause some misery when it came time to pack parts inside.

The Apevia X-Trooper Junior isn’t without its flaws, though. Two of the motherboard standoff locations simply weren’t threaded properly, and I had to use brute force to jam them in. Moreover, a lack of cable management makes routing leads and then cleaning up more of a challenge. Those are the only two issues that keep me from recommending X-Trooper to friends and family.

More unfortunate is that our off-the-shelf ASRock P67 Pro3 SE doesn't work properly, refusing to run in dual-channel mode. The Mushkin Redline kit that we bought isn't on this platform's approved memory list; however, we tried modules from Corsair and OCZ as well. None of them got our board working the way it was supposed to. Increasing voltage didn’t help, and so we’re forced to run our tests in single-channel mode. This probably won’t affect the game benchmarks much, but it almost certainly will have a negative impact on memory-intensive application benchmarks.

Overclocking

The 3.1 GHz Core i5-2400 isn’t multiplier unlocked, but it does have to operate within the constraints of a 38x maximum Turbo Boost multiplier ceiling. This allows us to force all four cores up to 3.8 GHz.

That would have been a reasonable tradeoff, considering the Core i5-2400's relatively low price compared to Intel's Core i5-2500K. But because many vendors are now encouraging overclocking through Turbo Boost offsets, the technology remains on all of the time without an option to disable it. So, despite our 38x multiplier setting, our processor ended up running at 3.6 GHz in fully-threaded workloads and 3.7 GHz in less demanding tests. That's better than default, but not quite the 3.8 GHz we wanted.

The PowerColor Radeon HD 7970 was more than happy to hit the core and memory frequency ceilings imposed by AMD's Overdrive tool. Stock 925 MHz core and 1375 MHz memory clocks jumped up to 1125 and 1575 MHz, respectively. We also slid the power control setting to +20%, lowering the chances we'd see throttling under load. This specimen might have overclocked even higher, but we've been having some trouble unlocking higher clock rates in the latest version of MSI's Afterburner utility.

Test System And Benchmarks

It will be interesting to see how AMD’s FX-6100 will fare compared to Intel’s Core i5-2400, especially when overclocked since the FX chip has an unlocked multiplier and two extra execution cores. Plus, the FX-6100 system’s dual Radeon HD 6950 CrossFire setup has a theoretical advantage over the Radeon HD 7970 when it comes to raw frame rates, although the single card is a more elegant solution.

And here's the list of benchmarks:

Benchmark Results: Synthetics

We kick off our synthetic tests with 3DMark 11, which gives us an immediate first impression of how this quarter's single Radeon HD 7970 compares to last quarter's Radeon HD 6950s in CrossFire.

It’s surprising to see how close these results end up. AMD's Radeon HD 7970 does perform a bit better at lower resolutions. But, at the highest detail setting, our CrossFire setup claims a slight advantage.

As usual, PCMark’s component tests favor Intel's Core i5 over AMD’s FX. Crucial's m4 SSD demonstrates a significant advantage in application launch speed compared to OCZ's Vertex Plus.

The results in Sandra are much closer, as AMD's FX-6100 earns its share of wins in the Arithmetic and Multimedia tests. Note the poor memory bandwidth from our current build, which is easily explained by the single-channel memory configuration we are forced to use due to this platform's uncooperative motherboard. The FX-6100 system's dual-channel bandwidth is much higher than what we reported in last quarter's System Builder Marathon, as an update to the software now properly reports the Bulldozer architecture's throughput.

Benchmark Results: Media Encoding

Intel's Core i5-2400 does better than AMD's FX-6100 when it comes to media encoding, despite the fact that we're stuck with a single memory channel and a modest overclock. Although the FX is specified as a six-core processor, it's architecture isn't that cut and dry. Even at a lower clock rate, four of Intel's Sandy Bridge-based cores simply get more work done.

AMD does take a single victory in HandBrake when its FX-6100 is overclocked to 4.5 GHz.

Benchmark Results: Productivity

Productivity-oriented benchmarks deliver another mixed bag of results. The ABBYY FineReader and 7-Zip numbers are close, which is interesting because the former is heavily threaded, while the latter is much less so. Nevertheless, Intel's Core i5-2400 holds onto a significant lead in the WinZip and WinRAR compression workloads.

When it comes to graphics applications, the Core i5-2400 manages slight wins in 3ds Max and Adobe Photoshop at stock speeds, while AMD's FX-6100 takes the lead when we overclock it.

This is actually a little surprising, and it may be that the Intel system's single-channel memory handicap is hurting it. Fair's fair, though, and AMD gets an advantage as a result.

Benchmark Results: Battlefield 3 And Elder Scrolls V: Skyrim

Battlefield 3

Battlefield 3 is a very graphics-dependent game, so the results should come down to the horsepower of our Radeon-based graphics cards.

Whoa. The Radeon HD 6950s in CrossFire from last quarter's System Builder Marathon beat the Radeon HD 7970 at every combination of resolutions and settings, except 2560x1600 at Ultra details.

Skyrim

Unfortunately, Skyrim automatically gets patched through Steam. As we discovered yesterday in Paul's story, a patch that appeared between last quarter (when we ran our comparison data) and today affected performance so significantly that we can't compare our two systems in this title. As such the numbers you're about to see don't mean a whole heck of a lot.

Benchmark Results: DiRT 3 And StarCraft II

DiRT 3

DiRT 3 is often limited by host processor performance, except at its highest detail settings and resolutions, and that’s what we see here. The Core i5-2400 system and Radeon HD 7970 combo are notably faster overall. But when we apply the Ultra preset and 8x MSAA, the overclocked FX-6100 and Radeon HD 6950s in CrossFire pulls ahead at 2560x1600.

StarCraft II

We've seen that StarCraft II can also be limited by CPU performance when a map is filled with units. But the second generation of our in-house benchmark is less stressful in that regard, and therefore more dependent on graphics performance.

As such, the Core i5-2400 and Radeon HD 7970 combo retains an advantage until 2560x1600 at Ultra quality settings with 8x MSAA applied.

Power And Temperature Benchmarks

AMD's FX-series CPUs are not known for power efficiency under load (and certainly not when overclocked), so the results we're about to see shouldn't take anyone by surprise. Of course, a single Radeon HD 7970 tells a much more compelling efficiency story than two Radeon HD 6950 cards in CrossFire, so expect to see a confluence of factors impact the final consumption numbers.

The new build’s modest Hyper TX3 cooler isn’t much of a mach for Xigmatek’s Gaia, but it does a decent job considering its less expensive price. The Radeon HD 7970's load temperatures look a little higher than we’d expect, but keep in mind that the Radeon HD 6950 CrossFire setup from the previous build also benefits from non-reference coolers.

A Surprisingly Robust Gaming System

Before we wrap this up, let’s consider the aggregate performance data. Keep in mind that we removed the Skyrim results from the following chart because the game's patch altered performance too significantly, rendering a comparison completely invalid.

First up is application performance. On average, the Core i5-2400 does quite well against AMD's FX-6100, especially in light of its motherboard, which artificially hamstrung it with a single memory channel and an inability to specify a constant 3.8 GHz. We did see a couple of benchmark situations where the FX-6100 took a first-place finish when it was overclocked to 4.5 GHz, due to its high clock rate, dual-channel memory, hexa-core architecture, or a combination of the three. In general, though, the Core i5-2400 comes out on top.

As far as game performance goes, the new build's Radeon HD 7970 performs admirably up to the charted resolution of 1920x1080. If we were to focus more intently on 2560x1600, though, the benchmarks suggested to us that the two Radeon HD 6950s in CrossFire would have fared much better.

It’s a real shame that ASRock's P67 Pro3 SE isn't equipped with better BIOS overclocking options, and especially unfortunate that it was unable to run our memory kit in dual-channel mode. We’re not sure how much of an impact this had on the final results, but it's possible that we'll try a Core i5-2400 on a different motherboard next time around just to compare the two platforms.

Regardless, looking at the average performance and power consumption of this quarter's build, a Core i5-2400 and Radeon HD 7970 combo are a better bet than an overclocked FX-6100 with two Radeon HD 6950s. Despite its well-documented flaws, the new build is a powerful little system more deserving of our $1250 budget.

We're especially looking forward to Day 4 of our System Builder Marathon, where Thomas will compare the overall value of our three systems. Not to give too much away, but our top-end machine also centers on Radeon HD 7970, so it'll be interesting to see how much platform alacrity impacts gaming performance.

Hitachi has officially confirmed two new models of 4TB hard drives in the form of the the Deskstar 5K4000 Internal Hard Drive Kit and the Touro Desk External Hard Drive.

The 4TB Deskstar 5K4000 is a 3.5-inch hard drive with a 32MB cache buffer and Hitachi's CoolSpin power-management technology. Combined with Hitachi's other power-management tools, CoolSpin offers up to a 28-percent idle power savings over the company's Deskstar 7K3000 7200 RPM drives and an idle acoustic level of 2.5 bels. The 5K4000 also uses Hitachi's Advanced Format technology, which increases the sector size on the hard disk drive from 512 bytes to 4096 bytes for increased capacities and improved error correction capabilities.

The new 4TB Hitachi Deskstar 5K4000 Hard Drive Retail Kit is currently shipping in limited quantities to online and retail stores with a suggested retail price of $399.99. Hitachi will be shipping the Deskstar 5K4000 (bare drives and bulk pack) to OEMs, distributors and channel partners worldwide in Q1 2012.

The external solution, the Hitachi Touro Desk External Drive is a USB 3.0 storage solution with a Deskstar 5K4000 HDD under the hood and is both Mac and PC compatible. It comes bundled with 3GB of cloud storage for free from HitachiBackup.com and will be available in January for approximately $420.

The impact of the flood in Thailand is much more dramatic than previously estimated. Along with a 10 to 18 percent decrease to 41 to 45 million units, according to a report published by Digitimes, it appears that the HDD industry's shipments will decline by more than 35 percent in the current quarter.

According to Digitimes, WD, Seagate, Hitachi and Toshiba will ship only about 103 million units in Q4, down from 162 million in Q3. WD may be hit the hardest, with shipments estimated to be in the 22 to 26 million range, down from 58 million in Q3. Hitachi may drop from 32 million to 20 million and Toshiba from 22 million to 12 million.

According to Seagate, the global demand calls for about 180 million drives in Q4.

Supersize My Storage, Please

The race to achieve higher and higher hard disk capacities recently entered its next stage. While Western Digital just expanded the GoFlex line with its very first 3.5" 4 TB drive, physically smaller disks are setting records of their own. Now, the very largest 2.5” models (represented by Samsung’s M8 [HN-M101MBB] and Western Digital’s Scorpio Blue [WD10JPVT]) offer 1 TB of storage space.

That’s only half of the innovation, however. After all, 1 TB 2.5“ hard disks have been around since mid-2009. Back then, 1 TB drives used three 333 GB platters, which bumped their height to 12.5 mm (0.5“) and thus prevented them from being installed in most notebooks. Samsung and Western Digital resolved this issue; their 1 TB drives now only sport two 500 GB platters, resulting in a z-height of 9.5 mm (0.374“), which is suitable for laptops.

Advanced Format (AF) Leads to Higher Data Density

The breathtaking speed at which storage density progresses can be partly attributed to the Advanced Format (AF), though up until now it mostly benefited 3.5“ drives. AF sports a sector size of 4 KB, eight times the size of traditional 512-byte sectors. Thus, this format has only one-eighth of the gaps between sectors and one-eighth of the synchronization and error correcting blocks. According to disk manufacturers, AF is responsible for increasing disk capacity by 7 to 11 percent on its own.

In order to maintain compatibility with the outside world, AF-based hard disks emulate 512-byte sectors. Current Windows and Linux systems work flawlessly with AF-based drives; you simply can’t tell the difference. However, older operating systems like Windows XP, Windows Server 2003, and Windows Home Server may exhibit performance problems because they don’t align their partitions to 4 KB boundaries. You can overcome that limitation by downloading an alignment tool from each manufacturer’s website.

Comparison Test of Six 2.5“ Hard Disks with 3 Gb/s SATA Interfaces

Only two of the notebook hard disks we tested offer 1 TB of storage capacity, namely the Samsung M8 HN-M101MBB and Western Digital Scorpio Blue WD10JPVT. The main topic of this test is, however, Advanced Format, as three of the remaining four test candidates are AF-equipped: Hitachi’s Travelstar 5K750 HTS547575A9E384 (750 GB), Samsung’s Spinpoint M8 HN-M500MBB (500 GB), and Western Digital’s Scorpio Black WD7500BPKT (750 GB). In total, five of the six hard disks we’re testing employ AF technology.

The only exception is Toshiba and its MK6461GSYN (640 GB), which does not merely emulate 512-byte sectors, but still uses them internally. The Japanese manufacturer didn’t miss the AF train; it already announced its MQ01ABD100, a 1 TB drive that features 4 KB sectors and a 9.5 mm (0.374“) height.

Hitachi Travelstar 5K750 HTS547575A9E384 (750 GB)

Hitachi addresses a broad range of applications with its 2.5“ Travelstar 5K750 product line, including mobile computing and mobile storage, game consoles, blade servers, network routers, and video surveillance systems. There are three models available: 500 GB, 640 GB, and 750 GB. We’re testing the flagship model, the 750 GB HTS547575A9E384, which costs about $80.

The Travelstar 5K750 hard disk has two platters with 375 GB each and features a storage density of 472 Gb per square inch. All models in this product range employ AF and emulate 512-byte sectors. The disks are 9.5 mm (0.374“) high, have an 8 MB cache, rotate at 5400 RPM, and feature a SATA 3Gb/s interface, which is still sufficient for notebook hard disks.

All Travelstar 5K750 models can optionally be ordered with BDE (Bulk Data Encryption) support, which includes a hardware encryption unit, or as the Enhanced Availability (EA) version, certified for 24/7 operation. Unfortunately, as with base and BDE-equipped drive, Hitachi does not specify the EA SKU’s MTBF. The model numbers of the BDE variants end in 1 instead of 4; the model number of the EA variants begin with HTE instead of HTS.

Performance

According to its specification sheet, the maximum data rate of the Travelstar 5K750 is 998 Mb/s. Average access time is quoted as 12 ms, which is a tad optimistic, according to our tests. We measured access times between 18 and 20 ms, with the 750 GB model clocking in at the high end of this range. Hitachi specifies 1.5 W power consumption while active and 0.8 watts while idle. Our tests confirm the idle value, but the power consumption of an active disk varies between 1.2 W and 2.4 W, depending on the benchmark.

With a 73.7 MB/s average sequential read speed and 73.1 MB/s average sequential writes, the Hitachi Travelstar 5K750 places last in the data rate tests. Most of the other competitors with a 5400 RPM spindle speed are significantly faster. Only Samsung’s Spinpoint M8 HN-M500MBB is in the same speed category. Hitachi’s sample is outclassed by the Spinpoint M8 HN-M101MBB and, even more so, the Western Digital Scorpio Blue WD10JPVT.

Samsung Spinpoint M8 HN-M500MBB (500 GB) And HN-M101MBB (1 TB)

Samsung targets the mobile computing and notebook market with its 2.5“ Spinpoint M8 family. You can choose between four different storage sizes: a 1 TB flagship (HN-M101MBB) and 750 GB model feature two 500 GB platters, whereas the 500 and 320 GB models only employ one platter. The manufacturer does not specify its storage density, but mentions AF support and the capacity increase enabled by AF.

Samsung sent us two Spinpoint M8 disks for this round-up, the HN-M500MBB (500 GB, $55 street price) and HN-M101MBB (1 TB, $90 street price).

Apart from the different capacities and platter configurations, all Spinpoint M8 models are identical. They have an 8 MB cache, a SATA 3Gb/s interface, and a 5400 RPM spindle. Compared to the M8’s predecessor, Samsung claims to have increased performance by about seven percent through a combination of AF and a higher-clocked SoC (system on chip) controller.

Performance

The performance of both Samsung notebook hard disks is average. The 500 GB model has a slightly higher sequential data rate than the 1 TB one, achieving a read rate of 84.7 MB/s and a write rate of 84.6 MB/s, which is quite good. It’s not good enough for the top spot, however, as Western Digital’s Scorpio Blue WD10JPVT is faster.

The 1 TB Samsung Spinpoint M8 loses out to its smaller sibling, sporting read speeds up to 79.9 MB/s and writes as fast as 79.7 MB/s, topping Hitachi’s Travelstar 5K750. If your main focus in choosing a 1 TB notebook hard disk is performance, the Western Digital Scorpio Blue WD10JPVT will serve you better.

Compared to 7200 RPM drives, the Samsung disks offer lower power consumption. According to its data sheet, the Spinpoint M8 draws 2.2 W while in use and 0.7 W when it’s idle. Our tests confirm the idle power number, and demonstrate between 1.2 and 2.6 W for a disk in use.

Toshiba MK6461GSYN (640 GB)

The Toshiba series MKxx61GSYN offers up to 640 GB of storage space, primarily targeting home users. Buyers can choose among five sizes: 160 GB, 250 GB, 320 GB, 500 GB, and our test subject, the flagship MK6461GSYN with 640 GB capacity, which costs approximately $90.

Contrary to the other five hard disks we’re testing, the Toshiba disk does not use Advanced Format, though this doesn’t negatively affect height or capacity. While the 500 GB and 640 GB models use two platters, the smaller models use one. All of them, however, are only 9.5 mm (0.374“) high, which is the standard height of notebook drives. According to Toshiba, the storage density is up to 506.3 Gb per square inch.

The MK6461GSYN is equipped with a SATA 3Gb/s interface, a 16 MB cache, and a 7200 RPM spindle speed. Incidentally, Toshiba’s MK6461GSYN and Western Digital’s Scorpio Black WD7500BPKT are the only two members of the ”7200 RPM club” in this test. A version of the MK6461GSYN with a free-fall sensor is available, too. Should a notebook with this drive in it fall, the hard disk retracts its read-write head to the park position, avoiding a crash.

Performance

It is hardly surprising that 2.5” hard drives with 7200 RPM spindles achieve the highest data rates. Toshiba doesn’t specify its performance specs, so we’ll simply present the benchmark results. 

Toshiba’s MK6461GSYN achieves a sequential read rate of 89.5 MB/s and an almost identical write speed of 89.4 MB/s. It loses its match-up to the other 7200 RPM drive, Western Digital’s Scorpio Black WD7500BPKT, which barely misses three-digit data rates.

By the way, the 5400 RPM WD Scorpio Blue WD10JPVT almost catches up to Toshiba’s solution, too.

We expected this drive’s power consumption to be higher than the 5400 RPM drives. But, according to its data sheet, Toshiba should actually be on par with the slower-rotating drives. The company specifies 2.1 W while active, and a 0.8 W idle power. Our benchmarks confirm the idle figure, but show higher power consumption under load. Depending on the benchmark, it consumes between 1.5 and 3.2 W.

Western Digital Scorpio Blue WD10JPVT (1 TB) And Scorpio Black WD7500BPKT (750 GB)

Western Digital sent us the flagship models from both of its product lines, Scorpio Blue and Scorpio Black.

Due to their different spindle speeds, they don’t play in the same league. The Scorpio Black rotates at 7200 RPM, emphasizes performance, and is targeted at desktop replacement notebooks. Hard disks in the Scorpio Blue product line, on the other hand, spin at 5200 or 5400 RPM, depending on the model. They are optimized for low noise and low power consumption, while still offering good performance. The manufacturer sets its sights on many different types of portable devices with the Blue family. And the 1 TB WD10JPVT belongs to that more power-friendly line-up.

Western Digital Scorpio Blue WD10JPVT (1 TB)

WD takes no prisoners when it comes to data density, fitting 500 GB onto each platter. The flagship model (WD10JPVT, with a street price around $115) features a SATA 3Gb/s interface, a height of 9.5 mm (0.374“), and an 8 MB cache. The Scorpio Blue family contains 13 other disks.

Be forewarned, the other 1 TB drive (WD10TPVT) is 12.5 mm (0.5“) high, and thus unsuitable for most notebooks.

Western Digital Scorpio Black WD7500BPKT (750 TB)

The Scorpio Black line-up only has six models, ranging from 160 GB to 750 GB. All of them have a SATA 3Gb/s interface, a 16 MB cache, and a 7200 RPM spindle speed. Only the two largest disks are AF-equipped: the WD5000BPKT and our test candidate, WD7500BPKT, which costs $90.

Performance

Western Digital earns top marks. The sequential read and write performance of both drives winds up at the top of their respective segments. The 5400 RPM Western Digital Scorpio Blue WD10JPVT achieves a sequential read speed of 87.3 MB/s and a sequential write rate of 87.2 MB/s, beating Samsung’s Spinpoint M8 HN-M500MBB by a slim margin. The 7200 RPM Western Digital Scorpio Black WD7500BPKT wins against the competition quite decisively, posting a sequential read rate of 98.0 MB/s and a sequential writes as fast as 97.4 MB/s.

Our test shows that the Western Digital Scorpio Black WD7500BPKT has the highest power consumption amongst the samples measured, at up to 3.3 W (the Scorpio Blue hits 2.5 W). Even at idle, the ammeter needle swings to the right more than it does with the other disks, revealing that the WD7500BPKT draws 1.0 W (the Scorpio Blue is lower, at 0.7 W).

Comparison Table And Test Configuration

Test Setup

Benchmark Results: Data Rate And Interface Bandwidth

The results of the sequential read and write tests are clearly grouped by spindle speed. The 7200 RPM Toshiba MK6461GSYN and Western Digital Scorpio Black WD7500BPKT drives achieve the highest performance, with a significant gap between the fastest WD and second-place Toshiba models. The 5400 RPM WD10JPVT nips at the heels of Toshiba’s offering, making it the fastest disk in its class.

Both the 500 GB Samsung Spinpoint M8 HN-M500MBB and 1 TB Spinpoint M8 HN-M101MBB make the most out of the SATA 3Gb/s interface’s available bandwidth.

Benchmark Results: Streaming Reads/Writes

It seems obvious: performance-oriented hard drives with 7200 RPM spindles and correspondingly low access times should top the charts in Iometer’s streaming benchmark.

The reality is a little bit different, though. While Western Digital’s WD7500BPKT sits unchallenged at the top of the chart, the other 7200 RPM disk, Toshiba’s MK6461GSYN, and two 5400 RPM models (Samsung’s Spinpoint M8 HN-M500MBB and Western’s Digital Scorpio Blue WD10JPVT), are clustered together below. Samsung’s Spinpoint M8 HN-M101MBB and the slowest drive, Hitachi’s Travelstar 5K750 HTS547575A9E384, trail far behind.

Benchmark Results: Access Time And I/O Performance

The low-level benchmark h2benchw directly accesses the hard disk and ignores the OS-imposed sector alignment. In light of the fact that five out of six test candidates employ 4 KB AF sectors and emulate 512-byte sectors, we find this benchmark very interesting.

The results of h2benchw aren’t so clear-cut, though. The 7200 RPM models tend to be in front. But the Western Digital Scorpio Blue WD10JPVT encroaches into their territory and posts comparable access times. The Toshiba drive dominates the read test and is almost twice as fast as the competitors.

As notebook hard disks, the test candidates are not really optimized for database, Web server, or workstation access patterns. But the benchmark results are quite clear: both Western Digital disks top the performance charts, while Toshiba’s MK6461GSYN posts mid-range results. Both Samsung disks and the Hitachi Travelstar 5K750 do not win any laurels in these workloads.

Benchmark Results: PCMark Application Performance

The Western Digital Scorpio Black WD7500BPKT is the overall performance winner in PCMark Vantage. The Toshiba MK6461GSYN and Western Digital Scorpio Blue WD10JPVT trail at a distance, surprisingly close to each other. Both Samsung drives and the Hitachi Travelstar 5K750 seem to focus on quietness and low power consumption; they simply cannot keep up.

Power Consumption

Samsung and Hitachi keep their power consumption in check quite well, but the Western Digital Scorpio Blue WD10JPVT does even better across all of the simulated use cases to which we subjected it.

As expected, the Scorpio Blue WD10JPVT can’t be beat when it comes to performance per watt. Surprisingly, the power hog in this round-up, Western Digital’s Scorpio Black WD7500BPKT, winds up in second place. While it draws more power than its competitors, it overcomes that with consistently higher performance.

Who Sells The Fastest High-Capacity 2.5" Hard Drive?

When we assess the overall performance of 2.5“ hard drives, we find both Western Digital drives at the top of the stack. Overall, the Scorpio Black WD7500BPKT achieves the best benchmark results, impressing us with its high data rates and low access times. While this 7200 RPM drive certainly won’t win any awards for power consumption, you have to expect that from a performance-oriented storage device. On the other hand, it’s not an energy hog either, as it draws a mere 0.9 watts more power than its competitors.

If you value an energy-conscious (and yet, still fast) hard drive, consider the second-place finisher, WD’s Scorpio Blue WD10JPVT. It is not as nimble as the Scorpio Black, but it sports more reasonable power consumption. In spite of its huge 1 TB capacity, the Scorpio Blue is, like the rest of the field, only 9.5 mm (0.374“) high, making it compatible with all notebooks.

The Toshiba MK6461GSYN shows that hard drives don’t necessarily have to switch to Advanced Format to offer high capacity and decent performance. In many cases, the Toshiba drive barely trails the second-place Scorpio Blue WD10JPVT. Overall, there are only minor speed differences.

If your focus isn’t performance-specific, but you instead value energy-saving features instead, Samsung’s Spinpoint M8 models (HN-M500MBB and HN-M101MBB) get a nod, as does Hitachi’s Travelstar 5K750 HTS547575A9E384. None of those drives achieve the performance of the WD offerings or Toshiba’s MK6461GSYN, though.

Who Sells The Best 3 TB Hard Drive?

The latest 3.5” hard drives offer up to 3 TB of storage capacity. Western Digital was first with its Caviar Green WD30EZRS, and now Hitachi and Seagate have added competing products to their portfolios.

In this comparison, we put the Western Digital drive up against its 3 TB competitors: two drive from Hitachi (the Deskstar 5K3000 HDS5C3030ALA630 and Deskstar 7K3000 HDS723030ALA640), as well as Seagate's Barracuda XT ST33000651AS.

Sure, the 3 TB labels on the hard drives look impressive, but you have to exercise caution on many computers, which are incapable of recognizing and utilizing the drives properly. Making them run as intended often requires particular platform choices and a suitable operating system. And you can forget about using them as boot drives if you're not using a UEFI-equipped motherboard.

Despite those configuration-related hurdles (which we're going to have to collectively overcome at some point anyway if we want to see even larger disks), 3 TB drives still represent a very important pillar in storage environments, either at home or in enterprises. SSDs are superior in many ways, but they're definitely not in the position to displace conventional hard drives when it comes down to basic data storage. As a result of high capacity, low cost per gigabyte, and low power consumption per gigabyte, magnetic storage remains the only reasonable choice for backup, archiving, and user data.

Performance and Energy Saving

Our test candidates can be divided into two different camps. One is made up of the Western Digital Caviar Green WD30EZRS and the Hitachi Deskstar 5K3000 HDS5C3030ALA630. Their moderate spindle speeds translate to low power consumption, moderate temperatures, and minimal noise.

The second group is made up of the Hitachi Deskstar 7K3000 and Seagate Barracuda ST33000651AS XT. Their 7200 RPM speeds promise more performance. Power saving is not the main focus for these products, which is immediately reflected in their somewhat-higher power consumption.

Hitachi Deskstar 5K3000 (HDS5C3030ALA630) And Deskstar 7K3000 (HDS723030ALA640)

Hard drives that try to emphasize low power consumption purposely avoid drawing your attention to their spindle speed. This applies to Hitachi's Deskstar 5K3000 as well. Instead, the manufacturer touts the benefits of its CoolSpin feature that employs a rotational speed said to maximize performance, power consumption, and operating noise. Because the drive is purportedly optimized for power, it most likely employs a single speed somewhere between 5400 and 6000 RPM.

Compared to the Hitachi Deskstar 7K3000 (a much more performance-oriented hard drive with its 7200 RPM spindle speed), the Deskstar 5K3000 requires up to 29% less power as a result of its CoolSpin technology. You can tell the difference just by looking at each disk's data sheet. The 5K3000's active idle is rated at 5.1 W, while the 7K3000 is rated at 7.3 W. Both of these top-of-the-line models use five platters, and they have a areal density of 411 Gb per square inch.

Other than the 3 TB version (HDS5C3030ALA630), the Hitachi Deskstar 5K3000 is also available in capacities of 2 TB (HDS5C3020ALA632) and 1.5 TB (HDS5C3015ALA632). The same goes for the Deskstar 7K3000 series: the 3 TB flagship (HDS723030ALA640) is followed by a 2 TB version (HDS723020BLA642) and a 1.5 TB model (HDS723015BLA642).

The Deskstar 5K3000 drives use a 32 MB cache and a 6 Gb/s Serial ATA interface, while the Deskstar 7K3000 comes with a larger 64 MB cache. Otherwise, the technical details separating the two only differ slightly. Specs like platters used, weight, and temperature (around 5°C) are where you'll notice them. 

Performance

The different foci of the two Hitachi hard drives is reflected in their performance. The Deskstar 5K3000 has an average data read speed of 97 MB/s, while the Deskstar 7K3000 is clearly faster at 118 MB/s average throughput. The same goes for the access times: 6.1 ms for the Deskstar 7K3000 compared to 7.7 ms for the Deskstar 5K3000.

As with the other hard drives in this round-up, the two Hitachi drives are ideal solutions for capacity-heavy applications. The Deskstar 7K3000 is a bit faster due to its higher rotational speed and larger cache, but none of the hard drives in the test were really designed for environments with high I/O loads. For users building large storage arrays, our recommendation is (perhaps surprisingly) the Deskstar 5K3000, as it easily beats the faster 7K3000 when it comes to performance per watt.

Seagate Barracuda XT (ST33000651AS)

In addition to its impressive 3 TB capacity, the Seagate Barracuda XT holds its own against the rest of the competition when it comes to performance as well. Just like the Hitachi Deskstar 7K3000, Seagate's drive is equipped with a 64 MB cache, a 6 Gb/s SATA interface, and a spindle speed of 7200 RPM. Besides the 3 TB flagship model (ST33000651AS), Seagate also offers the Barracuda XT with four platters instead of five, resulting in a reduced capacity of 2 TB (ST32000641AS).

With its high speed, Seagate suggests that the Barracuda XT is particularly well-suited for high-performance PC gaming systems, HD video editing, desktop RAID arrays, and home/small business servers. Use that word server carefully, though. This product was primarily designed for desktop use, and servers have different requirements, often starting with the interface all four of these disks employ.

Performance

This drive has the highest sequential read rate, with an average of nearly 120 MB/s. However, the Hitachi Deskstar 7K3000 is closely behind at 118 MB/s. Thanks to its impressive speed, the Seagate drive also does well in tasks that require quick access times, such as application launching. In typical server workloads, the Seagate Barracuda does not perform up to the level suggested by its manufacturer. In the Iometer benchmark, which assesses typical Web server, workstation, and database performance, even the supposedly slower drives from Hitachi and Western Digital perform better. That's a clear sign that the Barracuda XT is not optimized for such tasks. Then again, we already knew that, right?

When it comes to saving power, Seagate overtakes Hitachi and its Deskstar 7K3000. While Hitachi's device uses up to 9.2 W, depending on the benchmark, the Barracuda XT uses a maximum of 7.7 W.

Western Digital Caviar Green (WD30EZRS)

Western Digital is something of a pioneer in this story because its 3 TB was the first one available. Along with the previously-tested WD30EZRS, the manufacturer now provides the WD30EZRX, another drive with the same capacity. The only difference is the interface. While our test candidate uses a SATA 3 Gb/s interface, the WD30EZRX has twice the potential interface bandwidth at 6 Gb/s. Of course, as we all know, neither drive can even come close to saturating those links.

Both drives are part of the Caviar Green product line, which along with the 3 TB capacity comes in 2.5, 2, 1.5, 1, and .75 TB capacity points.

Yet again, we see another drive vendor playing the same game. Western Digital reveals nothing about the exact speed of the Caviar Green drives, just like Hitachi with its Deskstar 5K3000. Hitachi calls the feature CoolSpin, while Western Digital opts for IntelliPower. With these mechanisms, the hard drives spins at a rate that achieves the best balance of performance, noise level, and power consumption. Because the manufacturer promises that performance is in no way compromised by the IntelliPower function, you can assume that the speed is somewhere between 5400 and 6000 RPM, just like the Hitachi drive.

Performance

While the ED30EZRS makes no claims of stellar performance potential, it lives up to its stated purpose of quiet data storage. The device is by no means slow, though. Its average read throughput of 94 MB/s is nearly as fast as its most applicable competition, Hitachi Deskstar 5K3000. However, the high average access time of more than 20 ms and the comparatively low interface performance of 211.6 MB/s speak for themselves. The 7200 RPM drives (like Seagate's Barracuda XT) deliver results almost twice as fast. The Western Digital WD30EZRS also must yield to the competition when it comes to the application benchmarks.

If you're looking for a large drive for data storage or several disks for a NAS server, the Western Digital WD30EZRS is ideal. However, the competing Hitachi drive is worth its slightly higher price if you're planning to use it in large storage arrays. With an idle power consumption of 6.2 W, it uses almost 1 W less, putting it in the lead for performance per watt.

Comparison Table And Test Setup

Comparison Table

Test Setup