SATA 6Gb/s on Your New Motherboard?

There has been a lot of talk about the new "Serial ATA Revision 3.0 specification." New motherboards are now on the market with "SATA 6Gb/s" controllers, but what does all this mean to the enthusiast? Heck, what does "SATA 3" mean? We take a quick look at what "SATA Revision 3.0" is supposed to do for the world and what it will actually do for us on the desktop currently with RAID 0 and single disk configurations.

continued...

Performance Testing

Our performance testing here today is kept very narrow, using only HDTach. You can download it for free at the previous link. While HDTach is heavily dated, it still provides us with the specifics we need to gauge basic drive or RAID array performance. If you are wanting to get some of your own numbers at home HD Tune is another good program but will cost you some green beyond the free trial. We passed on using our standard IOMeter software this time due to some questions we had on whether or not it was working correctly with the new SATA 6Gb/s controllers. Overall IOMeter is probably the best tool, but its results can be a bit convoluted for everyday desktop use. Also, you should know that to use IOMeter, you overwrite the entire drive, so be careful about firing it up on your rig at home.

Testing Setup and Particulars

For our testing, we have used the same Western Digital hard drive, model number WD1002FAEX. This is the WD Caviar Black Desktop Hard Drive in the 1TB size. It is native 6Gb/s, 64MB of Cache, and 7200 RPM. HardOCP has standardized on this hard drive for current testing.

Article Image

While keeping our drive model constant (although we did add and Intel X-25M 80GB SATA 3Gb/s SSD for comparison), we did change motherboards. We used the Gigabyte GA-X58A-UD7 and Asus P7H57D-V EVO (our review will publish next week). Both motherboards have USB 3.0 as well. I suggest checking out the HardOCP USB 3.0 testing, as it looks very promising. Keep in mind these are very different motherboards. We used 4GB in each for testing at 1066MHz 8-8-8-24-1T. We also matched processor speeds at 2.8GHz. The Gigabyte motherboard uses the Marvell 88SE9128 6Gb/s controller, while the Asus board uses the Marvell 88SE9123 6Gb/s controller. (Interestingly enough, the 9123 controller found on the Asus board is not listed in the Marvell PDF spec sheet or the Asus documentation, but we assume the functionality to be between the 9125 and 9120 model controllers.) The 9128 supplies RAID functionality where the 9123 does not. This is not meant to be a comparison between the two motherboards, but rather give you a look at what is available for purchase and what you will get performance-wise. Latest BIOS and controller firmware was used.

Controller Performance

Article Image

Article Image Article Image Article Image Article Image Article Image Article Image Article Image

Directly above, we have added the individual HDTach graphing screens which we have labeled as a supplement to the chart we have created with the data gleaned from HDTach.

Single Drive Performance

From the data it looks like the Marvell 9123 bests both the 9128 and ICH10R in terms of Burst Speed, by a fairly large margin. The Marvell 9128 also pulls ahead of the ICH10R in Burst Speed as well and here is a quick explanation from SATA-IO as to why.

By increasing the transfer rate up to 6Gb/s or 600 MB/s (megabytes per second), SATA technology enables faster transfer of short bursts of data to and from the drive cache.

This makes sense as we can actually utilize the massive 6Gb/s bandwidth in a Burst Speed situation where the hard drives' caches are emptied quickly. Keep in mind the huge 64MB cache on the Western Digital drive is faster than flash memory on the Intel SSD and that data caching is limited to the SSD controller to address overall performance issues. Comparing the Marvell 9128 and 9123 to the Intel ICH10R we see the Marvell solutions bury the Intel controllers in Burst Speed by differences of around 75MB/s.

Looking beyond Burst Speed, we also see that the Average Read Speed between the 9128, 9123, and ICH10R in a single disk configuration is flat with less than 1MB/s difference between boards with read speeds on a particular board being identical between the Marvell and Intel controllers. This is simply the limitation of the hard drive being seen.

Random access times are flat across our drives as well except obviously in the case of the SSD. Here we are seeing limitations of rotating media.

RAID 0 Performance

The Burst Speed on our Marvell 9128 in a RAID configuration is actually moving backwards compared to single drive Burst Speed numbers. RAID 0 is slower than a single disk. Looking at the ICH10R we see it scale somewhat appropriately with a RAID 0 configuration besting its own single disk transfer by 138MB/s. I did not expect the RAID 0 array on the 9128 and 9123 Marvell controllers to be slower than the single disk Burst Speeds.

Average Read Speed on our RAID 0 2 disk configurations is very close, but the Intel ICH10R does edge out the Marvell 9128 by 10MB/s. The ICH10R looks to be better tuned than the new Marvell 9128. Remember that the ICH10R is a SATA 3Gb/s controller, yet still posting better Average Read numbers. Keep in mind that we are no where even near the 600MB/s theoretical transfer speed of SATA 6Gb/s using 2 mechanical hard disks in RAID 0. Still, the Intel ICH10R is better tuned or simply more efficient in this example than the Marvell 9128. The Marvell 9123 does not have RAID capabilities.

Again in RAID 0 we see our access times remain flat as expected.

SSD

We put the Intel X25-M 80GB SATA 3Gb/s SSD into the mix for a couple of reason. First we know that many of you are interested in purchasing these drives as they move down in price. From our graph you can see that a single SSD bested our single and RAID 0 mechanical drive arrays. We would have loved to get a SATA 6Gb/s drive into the mix, but those are simply not available to us yet. While SATA-IO tries to dress up SATA 6Gb/s technology, we all know that it is "waiting" on 6Gb/s Solid State Disk drives outside of the enterprise arena.

One of the major market needs SATA 6Gb/s addresses is the throughput bottleneck facing solid state drives (SSDs). SSDs provide faster data access and are more robust and reliable than standard HDDs because they do not incur the latency associated with rotating media.

With SATA 3Gb/s, SSDs are already approaching the performance wall with sustained throughput rates of 250-260 MB/s. Next-generation SSDs are expected to require 6Gb/s connectivity to allow networks to take full advantage of the higher levels of throughput these devices can achieve.

Here is a look at a 6Gb/s SSD prototype being tested at Marvell. Although it is not tremendously impressive, at least we see the technology in the developmental state. In the IOMeter testing towards the end of the video, we are seeing data rate speeds that simply would not be possible with SATA 3Gb/s technology we use currently.

Looking forward we are surely going to see SSD hard drives need the SATA 6Gb/s bus to perform best.

The Bottom Line

In our single disk testing, we saw the hard drive operating on the SATA 6Gb/s bus turning in faster Burst Speed data rates of around 20%. Sadly though, given how fast Burst Speeds are with today's drives even these large jumps in speed really do not equate to much more than a graph best used by marketing folks. You would not see much impact from this increased Burst Speed on real world computing. SATA-IO does feel as though Burst Speed is a bit more valuable (Your Mileage May Vary):

By increasing the transfer rate up to 6Gb/s or 600 MB/s (megabytes per second), SATA technology enables faster transfer of short bursts of data to and from the drive cache. For example, high-end PCs used for bandwidth-hungry applications like video editing would be good candidates for SATA 6Gb/s because they require high performance access to memory. High-end solid state drives (SSDs) are already pushing the limits of available interfaces, impacting read speeds. Moving to SATA 6Gb/s will remove the bottleneck for SSDs to enable faster read and write speeds.

In terms of everything else, we saw basically no difference from the same single drive being plugged into a SATA 6Gb/s or SATA 3Gb/s port. All of this is obviously due to the hard drive not being able to actually take advantage of the 6Gb/s bus. We will see SATA 6Gb/s start to shine with SATA 6Gb/s SSD drives, but for now it looks as if you would have to configure a fairly big RAID 0 array to push the technology and we do not currently have a motherboard with more than a single Marvell 9128 controller supporting only two drives.

In a RAID 0 2 disk array situation we see that Intel's SATA 3Gb/s ICH10R controller bested the Marvell's SATA 6Gb/s 9128 controller. We triple checked this data and then went and then asked motherboard builders to confirm our results and all had seen the same thing.

When buying your new motherboard there is no performance reason we can point to that suggests you should select a motherboard with SATA 6Gb/s. In fact, if you are running a RAID 0 setup, you are actually better off going with Intel's ICH10R solution. I see a lot of tech editors complaining about Intel's lack of currently having a SATA 6Gb/s controller on the market right now, and I have to ask why that is. The bottom line is that Intel knows there is no advantage to currently bringing the product to market until it has its SATA 6Gb/s Solid State Disk drives for sale. From all the data, that makes perfect sense to us here at HardOCP.

The "only" reason to buy a motherboard with SATA 6Gb/s currently is if you want to "future proof" your computer. Seeing how Intel's ICH10R bests Marvell's SATA 6Gb/s controller in a RAID configuration, you have to wonder about how much tuning is going to be done on Marvell's future controllers when the "new stuff" is not as efficient as the "old stuff." I am going to make a guess here and say when Intel releases its SATA 6Gb/s technology, we are going to see true advantages. There is another reason to buy a SATA 6Gb/s motherboard, and that is because most of those products seem to be bundled with Super-Speed USB 3.0. That is something you will likely get a lot of use out of before SATA 6Gb/s technology as you can currently purchase USB 3.0 devices. We are excited to see Super-Speed USB 3.0 connected SATA 6GB/s external Solid State Disk drives, hopefully 1H11.

Editor's note: Upcoming 256GB SSD hard drives that use SATA 6Gb/s technology are getting read speeds of around 350MB/s and write speeds of around 250MB/s and are projected to cost about $3.15 per GB. 256GB x3.13 =$800

Discussion