Storage Benchmarks: Intel vs. Everyone
If you've been around a few years, you can probably remember when hard drive manufacturers began introducing features like native command queuing (NCQ) and Intel created the advanced host controller interface (AHCI) standard. The purpose of AHCI was to expose the more advanced capabilities of SATA as compared to the old parallel ATA (PATA) standard, while maintaining backwards compatibility with the older IDE standard. Even now, years (and several OS releases) after AHCI first debuted, it's still a feature most motherboards disable by default. Enabling it isn't as easy as flipping a BIOS switch; the OS almost always has to be reinstalled to enable or disable AHCI support. Even once that's done, there's still the issue of which driver one should use, Microsoft provides native AHCI drivers for Vista and Windows 7, but third party controller manufacturers still provide their own custom solutions.
The reason anyone bothers in the first place is because without AHCI, it's impossible to hot-swap SATA hard drives and NCQ doesn't function (some utility programs will report NCQ support is available, but this is based on whether or not the hard drive supports it, not whether or not it's actually running). Since the X58A-UD3R is such a storage-centric board, we opted to focus on comparative AHCI performance between the three separate controllers Gigabyte included. We tested PCMark Vantage in the following scenarios, all tests were conducted using a WD300 VelociRaptor hard drive. PCMark Vantage was looped 3x in each instance.
We've broken eSATA, IDE, and AHCI performance out into three separate charts and included the aggregate PCMark Vantage score as well as the results of any suite subtests that showed potentially meaningful variation.
- Gigabyte GSATA2: This is a rebadged JMicron JMB362. Tested using both the Gigabyte-provided drivers and a newer JMicron driver from the company website.
- eSATA Ports: These are also controlled by a JMicron JMB362. Tested using the same set of drivers as the GSATA chip.
- Marvell 9128: The only 6G controller chip around at the moment. Tested using the latest company driver.
- Intel ICH10: Tested using Intel's own driver.
Performance scarcely budges when all of the controllers in question are run in IDE mode. The Intel controller is about 7 percent faster than the GSATA2/JBM362, with the Marvell halfway between the other two. The HDD test was the only suite that demonstrated any variation; all three controllers were within 2-3 percent of each other in all other tests.
AHCI performance is much more varied. Intel's ICH10 leads the aggregate score by a whisker, but decisively wins out over the JMicron controllers in the HDD test suite. Given Vantage's noted tendency to throw wild scores, the six percent gap between ICH10 and Marvell's 9128 is too close to comfortably call, but the Intel chip is a full 30 percent faster than the JMicron 362 using Gigabyte's driver and ~25 percent faster than the reference driver from JMicron. JMicron's reference driver is an interesting case. Flip back a page to our eSATA comparisons using the Seagate BlackArmor P110, and you'll note that PCMark Vantage again reported a much higher result in the "Music" suite when using this particular driver than when using anything else. With the exception of that particular series of tests, however, the driver's performance is consistent. It's typically a few percent faster than the Gigabyte-provided driver, but falls behind the Intel and Marvell controllers.
If you check the HDD Test suite scores, you'll note that while testing in AHCI mode increased aggregate PCMark Vantage results across the board, HDD Test Suite results moved in the opposite direction. Intel and Marvell performance rose by six percent and four percent, but the JMicron controller's performance fell 9-13 percent depending on which driver we tested. One of the good things about the X58A-UD3R is that it allows you to manually select which controllers you want to operate using AHCI and which you want to use as IDE. Given the results we've seen, it may be best to keep the JMicron ports on IDE, particularly if you plan to hook a CD drive up to them. We've not had problems personally, but some CD-ROM drives don't play nice with AHCI drivers and third-party controllers.
Before we wrap this up, let's check eSATA performance again, with the faster VelociRaptor instead of our mobile Seagate.
eSATA performance follows the same trend we saw when testing the board's three controllers, although the baseline scores aren't as high. Again, JMicron's own driver outperforms Gigabyte's included version.