Chad takes a couple of Seagate’s 160GB SATA Barracuda 7200.7 drives to see what NCQ is all about. Check out the article for single drive NCQ performance and RAID 0 performance. It’s an interesting read!
Seagate was the first drive manufacturer to put serial ATA, or SATA, hard drives on the streets. In 2003, Seagate delivered its Barracuda ATA V in a SATA version for all those early adopters starving for new technology. That Barracuda was more like an average trout when it came to performance, though. However, it did offer the low noise-level that Seagate has a reputation for delivering. A lot has changed in the nearly two years since 2003, though. Now, the top drives from major manufacturers feature the SATA interface and twice the capacity. The competition in this space has really been heating up since early 2003. Western Digital released its Raptor line of SATA hard drives that features 10,000 RPM spindle speed and near-SCSI performance levels. And Maxtor recently released its new DiamondMax 10 family of drives, which boasts features such as a 16 megabyte buffer and dual-processor technology.
Not to be outdone, this year Seagate became the first company to introduce SATA drives featuring Native Command Queuing, or NCQ, a feature that is included in the next generation SATA specification. Command Queuing has long been a feature in SCSI technology, but it is just now making its way into the ATA storage world. Native Command Queuing is a technology designed to improve the performance of SATA hard drives by optimizing the execution order of read and write requests sent to them. A standard SATA drive without NCQ technology will execute commands as they are received, but a SATA NCQ drive’s internal command queue will be reordered for optimal performance on the fly. To optimize performance, NCQ technology reorders the commands in such a way that reduces the amount of seek time the drive needs to access data. In other words, the NCQ reordering algorithms utilize such information as the physical location of the data on the disk. A common analogy for NCQ is the way an elevator works – it stops at near floors first rather than the order in which the buttons were pressed, thus reducing the total amount of time to meet all requests (drop everyone off at the floor they selected). Native Command Queuing has the most profound effect on performance in multi-tasking and applications with heavy asynchonous I/O loads. An added benefit of NCQ is that it will probably increase the lifespan of drives that use it since the amount mechanical movement for seeking is reduced.
Today, Seagate offers a wide selection of Barracuda SATA NCQ hard drives. The SATA Barracuda 7200.7 has been around for more than a year, and it offers capacities from 80 to 160 gigabytes and an eight megabyte cache. The company’s newest model and succesor to the 7200.7 is the Barracuda 7200.8, and it features capacities ranging from 200 to 400 gigabytes and either an eight megabyte or sixteen megabyte cache. In order to get the word out about NCQ and Seagate’s implementation of it, Seagate sent Bjorn3D two 160 GB Barracuda 7200.7 SATA NCQ drives. In an effort to see if NCQ could make a noticeable difference in performance, I ran some tests on different drive configurations using my Intel 915P Express testbed. As you will see on the next page, the results show that NCQ is more than just hype.
One important thing to note about NCQ technology is that it requires not only a drive that supports it but also host-side support, meaning the PC that the drive is connected to must have a controller capable of using NCQ. Today, only Intel’s 915 and 925 platforms with the Intel 82801FR I/O controller hub (ICH6R) have this feature, but AMD fans will have it when motherboards based on NVIDIA’s nForce4 hit the streets.
In order to see how much of a difference NCQ can make, I could only use one drive, because if SATA RAID is enabled on the Intel 915P chipset, NCQ comes along for the ride. The way the system is set up is determined by one option in the Award BIOS on my Gigabyte GA-8GPNXP. It is found under the Integrated Peripherals section and labeled SATA RAID/AHCI. The options for this feature are Disabled, RAID, and AHCI. If Windows XP is installed with this parameter set to Disabled, standard Windows drivers will be used to control SATA devices as if they are standard IDE devices. AHCI, which stands for Advanced Host Controller Interface, is the option that allows NCQ to work on a single drive and will allow installation of the Intel SATA AHCI controller driver. Actually, RAID can be selected with only one drive, too. This makes the system “RAID-ready” as Intel calls it and will allow users to migrate from a single drive to a RAID array if they desire. I chose the AHCI option for my NCQ tests. Just for fun, I also decided to test RAID 0 performance with the pair of drives that Seagate provided. The main thing that I will focus on in the results is the comparison between the two different single drive configurations’ performances.
Test System Configuration
- Intel Pentium 4 550 (article)
- Gigabyte GA-8GPNXP Duo (review)
- OCZ EB DDR PC3200 1 GB kit – 2 x 512MB (review)
- Seagate Barracuda 7200.7 SATA NCQ hard drive(s)
- Albatron PCX5750 – DirectX 9.0c, ForceWare 61.77
- Microsoft Windows XP with Service Pack 2
- Hyperthreading – enabled
FutureMark PCMark04 (1.2.0)
The HDD test suite in PCMark04 attempts to simulate the usage of hard disks in real applications. This includes activities such as loading Windows XP, opening and using applications and copying files. The results of these tests are combined to calculate the HDD Score.
|Drive Configuration||HDD Score|
|One 7200.7 – NCQ Not Enabled||4564|
|One 7200.7 – NCQ Enabled||5114|
|Two 7200.7s – RAID 0||6723|
NCQ delivers a performance boost of about 12% in PCMark04. That’s a nice boost for a feature you get essentially for “free” as long as you have everything set up correctly and the appropriate hardware. Let’s take a look at how these configurations perform on the individual PCMark04 HDD tests that comprise the HDD test suite.
The XP Startup test does not actually boot the system. It simply initiates operating system calls that would occur during a boot process. NCQ definitely provides a nice performance jump for this test. During the Application Loading test, various commonly used applications, such as Acrobat Reader and Windows Media Player, are opened and closed. Once again, NCQ comes through with slightly better performance. NCQ seems to make the smallest difference with the File Copying, which simulates copying 400 MB of files. However, it comes back with another decent performance gain in General HDD Usage. The General HDD Usage simulates accessing the hard drive in popular applications like Microsoft Word, WinZip, and Internet Explorer.
SiSoftware Sandra (2004.10.9.133) File System
The SiSoftware Sandra File System benchmark is fairly straightforward. It tests access time and buffered, random, and sequential reading and writing performance of a hard drive. The Drive Index value is an overall performance rating based on the average of the read, write, and seek tests, and file and cache size.
Overall, NCQ does not seem to add to the performance this time. I’m not sure why the Buffered Read performance is so much higher with NCQ enabled, but that is the result I observed numerous times.
PassMark PerformanceTest V5.0
PassMark PerformanceTest V5.0 allowed me to do something a little bit different than the previous two benchmarking tools. It allowed me to create a custom benchmark that simulated a web server and a database running simultaneously in two separate threads and stressing the disk subsystem. This is the type of test that should be able to show the benefit of NCQ, and it definitely did. I configured each of the two threads that were running to perform asynchronous, random access and almost exclusively perform reads. Each time the benchmark ran for five minutes.
|Drive Configuration||Data Written
|Average Transfer Rate
|One 7200.7 – NCQ Not Enabled||7.73||487.74||1.66|
|One 7200.7 – NCQ Enabled||11.80||594.55||2.03|
|Two 7200.7s – RAID 0||22.46||1091.67||3.71|
The 7200.7 with NCQ enabled really flexes its muscles here. In the same amount of time, the NCQ-enabled configuration was able to write over 50% more data and read about 22% more data than when NCQ was not enabled.
Real World Booting and Loading Times
I also wanted to get a feel for how some real world booting and loading times might be affected by NCQ, so I broke out the stopwatch and timed Windows XP bootup (from the OS selection screen until the Desktop could be seen) and Half-Life 2 loading times.
|Drive Configuration||Windows XP Boot
|Half-Life 2 Load
|Half-Life 2 Start New Game
|One 7200.7 – NCQ Not Enabled||39.0||65.0||29.7|
|One 7200.7 – NCQ Enabled||30.0||59.0||26.3|
|Two 7200.7s – RAID 0||N/A*||41.7||24.0|
* – Initialization of the RAID hardware added significant time to the boot process making it not a good comparison in this case.
It looks like NCQ can make a substantial difference in Windows XP boot times, as we saw above with the PCMark04 results. Half-Life 2 load times are also reduced by a few seconds thanks to NCQ.
I am pleased to see that Native Command Queuing seems to make a difference in various computing disciplines that average users might actually be able to notice. For people with a motherboard based on Intel’s 915 or 925 Express chipsets (with the appropriate I/O controller hub that supports NCQ) who want to maximize their system’s performance, I would definitely recommend taking advantage of the native NCQ capability of their system by pairing it with an NCQ-capable drive or, better yet, multiple drives configured in a RAID array.
The Seagate Barracuda 7200.7 SATA NCQ hard drives performed well in these tests, and I was never able to hear them over the other noises coming from the test system. I also feel it is worth noting that these (as well as all other new drives from Seagate) include a five year warranty. I have to give Seagate props for that, especially considering it was just a year ago that most manufacturers reduced their three year warranties to one year. If you are interested in these drives, keep in mind that the 7200.7 line includes parallel ATA and SATA non-NCQ drives, so obviously care must be taken when shopping for them.
Available drives supporting SATA NCQ:
- All Barracuda 7200.8 SATA units have NCQ.
- Barracuda 7200.7 SATA with these model numbers:
- 160GB – ST3160827AS
- 120GB – ST3120827AS
- 80GB – ST380817AS
- Maxtor DiamondMax 10 drives with the SATA interface