Date: March 15, 2018
Author(s): Jamie Fletcher
With games getting bigger and movies switching to 4K resolutions, the amount we need to store for our digital lives is forever increasing. While cloud services help, sometimes you need something local to keep everything on. We take a look at one of Seagate’s highest density drives on the market, the Barracuda Pro 12TB.
In our media-rich world, storage is in constant demand. Cloud services and video streaming have helped curb the need for huge amounts of local storage, but not everyone has the bandwidth or desire to keep everything on someone else’s computer. Hard drives are still the cheapest option for bulk and infrequent access storage, something Seagate has provided for many years.
A couple of months ago, Seagate updated its range of storage solutions with 12TB drives from its NAS ready IronWolf range, and under review today, Barracuda Pro desktop drives. While IronWolf is meant for RAID-enabled disk controllers due to error correction limits and vibration dampening mechanisms, the Barracuda Pro is about speed for local storage pools on workstations and desktops.
It’s certainly been a while since we’ve taken a look at any HDDs, since a lot of the market is switching over to solid-state storage. With this in mind, we have a new test system in place to manage this, since the old system no longer exists. During this time, drive densities have exploded. It doesn’t feel so long ago when 2TB drives were causing issues due to the MBR partition tables unable to cope with such large densities, and here we are now, looking at a 12TB drive.
The Barracuda Pro 12TB under review today, is a 7200RPM SATAIII 8-platter PMR, helium filled drive, built for high performance (for a mechanical drive) and high density for local storage on desktops. While its speed won’t match that of an SSD, its density and price will more than make up for it. Backed by a 5 year warranty and a 2-year data recovery service included, this also has some durability too. For a desktop drive, this is one of the highest capacity drives currently on the market, with nearly all others being for NAS or Enterprise storage solutions.
Helium in hard drives isn’t new, and it’s something we’ve covered before, but here’s a quick summary of why. Disk drives are typically nitrogen/air-filled, as they require some form of gas or air for the read/write heads to stay floating above the spinning platter. Nitrogen is quite a dense gas compared to helium, so there is a certain amount of drag involved on the platters and motor. Switching to helium means it lowers the drag, allowing the motor to spin faster, or in the case of high-density drives, attach more platters to the motor. This is why we’ve gone from 4-6 platters on older drives, to 8 and even 9 on modern ones. More platters not only increases the density, but also the overall performance in sequential reads/writes (providing the controller can keep up).
The test system for this drive is not ‘new’ but is more than enough for hard drive testing. The drive is hooked up to a SATA port connected to the PCH, using an Intel disk controller. Windows 10 Pro is installed on its own SSD, and any real-world file transfers are pulled from a separate SSD to prevent any interference from the OS drive.
The suite we’re using isn’t as extensive as what we would use for SSDs, simply because IOPS and queue depths are not a functional metric any more for HDDs, since SSDs have superseded them by an insurmountable margin. However, sequential read/write and a certain amount of random access is still important. As such we’re using synthetic tests from CrystalDiskMark, HD Tune, and Futuremark’s PCMark 8 storage benchmark. Real-world testing is performed with Window’s built-in tool called Robocopy, which transfers our long-standing mixed-media 50GB backup, consisting of files of various sizes.
All tests are performed a minimum of three times, or more if there are any irregularities, which will be mentioned if necessary.
Full system specifications are available in the table below:
|SmartKevin Hard Drive Test System|
|Processor||Intel Core i7-4770K @ Stock|
|Memory||Corsair Vengeance 16GB DDR3-2133|
Corsair Force LS 120GB SATA 6Gbit/s SSD
HyperX Fury 120GB SATA 6Gbit/s SSD
Seagate Barracuda Pro 12TB (ST12000DM0007, 256MB Cache, 7200 RPM)
|Power Supply||Corsair HX850i 850 Watts|
|Cooling||Cooler Master Hyper 212 EVO|
|Et cetera||Windows 10 Pro 64-bit, Open bench system|
When preparing our HDD testbed for benchmarking, we follow these guidelines:
Drives are initialized but left unpartitioned when tested with HD Tune, and are formatted with 4KB cluster sizes for PCMark 8, CDM and real-world testing.
Due to the new test system and a lack of comparative drives at this time, results will be short and should be taken as they are for the time being. However, as more drives are tested, graphs and tables will be populated in future articles.
We’ll start off our testing with PCMark 8’s Storage benchmark. PCMark 8 is normally used for a full system score, such as CPU or memory focused tests, but it can break things down into focused testing, and as such we’re using the storage test in this scenario. This typically takes over an hour to run as it simulates real-world file transfers and disk access behavior from a variety of popular games and applications.
We should note a couple of things first. At this time, PCMark 7 is no longer used due to its age, and PCMark 10’s storage benchmark is currently not out yet. Also, the drive needs to be formatted first, in this case, it’s NTFS with default block sizes. Due to a bug with PCMark that causes the system to stall, the entire 12TB disk partition cannot be used as the target, and instead a smaller partition of 200GB is used.
|Seagate Barracuda Pro 12TB PCMark 8|
Since these results are taken without context, it’s hard to say how this performs relative to other drives, but things will be made clearer as we move on to other tests where we have a better understanding of the results.
HD Tune is a favorite test tool for us, as it has a variety of disk metrics it can generate to check the overall health of a drive, including error rates, SMART information diagnostics, as well as performance metrics such as throughput, IOPS and random access. The drive splits up results into read only and write only, and is based on raw performance without a filesystem.
Due again to lack of comparative data, we’ll be showing captures of the test results and highlighting the important metrics below. You can see the performance of the drive decline as it becomes full over time; something that is often overlooked when looking at averages and maximums.
With the full-disk tests, you can see how performance tapers off as the hard drive becomes full, peaking at over 250MB/s at the start, and dropping down to 110MB/s at the end of the drive. Even without comparative results to highlight, we can say that this is indeed a fast drive. At 255MB/s peak and 195MB/s average, this thing is faster in sequential than the old VelociRaptors from WD in our old tests. At 110MB/s on the low-end, even when this drive gets full, it’ll still be pretty fast.
With write speeds, things are not too different, only knocking off about 5MB/s across the entire drive compared to the read speeds. Access times are a little odd though, with read, we see 14.4ms, which is respectable compared to old drives, but in writes, it drops down to half that at about 7ms. This appears to be a function of the very large 256MB of cache on the drive, and can be shown with the IOPS tests, as write IOPS at different transfer sizes are doubled that of read across the full range of transfer sizes.
Another long time favorite test for storage, as it’s a simple test that anyone can run and compare results with. Hard drives need to be formatted first, so there is some filesystem overhead, but ultimately negligible in difference, plus more realistic for actual real-world use.
CDM can be configured to run different volume sizes, but the end result is pretty much the same, with only a few percentile differences between the largest and smallest volumes. In this case we used nine passes of 4GB and 32GB.
The results tally up with what we were expecting, very high sequential performance, in excess of 250MB/s, and very typical 4K performance that we see with HDDs, due to the high access times.
Robocopy is a simple command-line tool built into Windows that can be scripted to automate large and complex file transfers and backups, complete with logging information. Its logs are a bit different from usual, as they are reported in bytes per second, or megabytes per minute.
For our results, we simply scale the bytes per second to megabytes by dividing the value twice by 1024, to convert bytes to kilobytes, and then again to megabytes. The transfer consists of a mixed media 50GB folder, containing nearly 4000 files in 370 directories. We also transfer the archive directly, as a single file. This transfer is done on an empty drive, with the transfer being deleted each time we start a new pass. As stated before, the transfer was done from a separate SSD than the main OS drive.
|Seagate Barracuda Pro 12TB: Robocopy|
|Single 50GB File||247.3 MB/s|
|50GB Folder||225.0 MB/s|
Without much in the way of comparative drives at this time, it’s hard to make any final judgments, however, there are a number of things we can point out. At $450, Seagate’s Barracuda Pro 12TB drive is not the cheapest drive on the market ($37.50/TB), but it is one of the densest. It’s meant to be used inside desktops and direct-attached storage systems (think USB/TB hard drives), so it doesn’t have the motor vibration dampening like the IronWolf Pro drives, but it is cheaper as a result.
Having a single dense drive is good for low power and low-profile systems, or where you want to minimize complexity without using raid controllers. This Barracuda Pro is also surprisingly fast, so archiving will take less time, if that’s your plan. That speed also makes it good for those ever-growing Steam libraries that we’re accumulating. Games such as Final Fantasy XV take up a massive 150GB, which certainly puts the pressure on lower capacity SSDs.
Because this is an 8-platter drive, subjectively speaking, the drive is loud. Compared the various 4 platter drives I have kicking around, this one is certainly noticeable. It’s not distracting, but it is more akin to the classic hard drive sounds of the previous decade.
For a better cost to density ratio, Seagate’s non-Pro drives are better value, at about $200 for 8TB ($25/TB), but you trade in the 5 year warranty and 2 years recovery service for just a 3 year warranty, and no recovery service.
If you are looking for high density drives you intend to use in a desktop, and are looking for reliability, then Seagate’s Barracuda Pro range is worth considering, especially with the extended warranty and recovery service. The fact they can be up to 12TB using PMR, rather than the more troublesome SMR method makes them a lot more useful as a daily work drive, too.
Copyright © 2017 SmartKevin