So what about the controller?
So if the Barracuda XT is the first hard drive with a SATA 6Gbps interface, what about an equivalent disk controller? The only one on the market that supports the new 6Gbps standard is Marvell's 88SE9123, a chip that got off to a bit of a rocky start earlier in the year. Problems with an initial version of the chip forced motherboard makers to drop it from their first wave of P55-based offerings. The issue was apparently traced to the ATA component of the chip rather than its SATA logic, and it's since been resolved with a new silicon revision.
Like the Barracuda XT, the 9123 is perhaps not the most appropriate poster boy for SATA 6Gbpsnot because of past indiscretions, but due to purely technical limitations. You see, the 9123's link to its host system is a one-lane PCI Express 2.0 interface that, at best, offers 500MB/s of bidirectional bandwidtha full 100MB/s short of 600MB/s. The Marvell chip has two SATA ports, too, so dual-drive or RAID configs will have to share the bandwidthnot that it's going to matter much to the Barracuda XT.
The 9123 has already found its way into a batch of refreshed P55 motherboards from Asus and Gigabyte. The latter has updated its entire P55 line with new revisions that feature the Marvell chip. Asus has been somewhat more conservative, rolling out the 9123 in a couple of models, including the P7P55D Premium.
In the Premium, the 9123 rides alongside the 3Gbps SATA controller embedded in Intel's new P55 PCH, making the board a perfect candidate for our new storage test platform. However, the P55's half-rate PCI Express 2.0 lanes only offer 250MB/s of bandwidth, forcing Asus to employ a PLX 8613 PCIe bridge chip to keep the Marvell controller properly fed. The bridge chip sits between the Marvell controller and the P55 PCH, transforming two of the P55's 250MB/s PCIe lanes into a single 500MB/s link. This arrangement puts the 6Gbps SATA ports one more hop away from the CPU. Data must first pass through the DMI link down to the P55, over a pair of PCIe lanes to the bridge chip, and then through another PCIe link to reach the Marvell controller.
Another curiosity associated with the Marvell controller is that its latest 126.96.36.1997 drivers section off a portion of main system memory to use as a transfer cache. That seems like an odd sort of thing to do, given how aggressively Windows preemptively populates system memory on its own, but Marvell claims Intel and JMicron use the same technique in their drivers. Seagate also recommends disabling write-cache buffer flushing with the Marvell driver to make it "behave more like" the Intel one. Write-cache buffer flushing forces applications to wait for data written to a drive to make its way through the cache and onto the platter, where it will persist in the event of power loss. The switch doesn't appear to have much impact on the Marvell driver's performance.
We'd already gathered a full set of performance data for the XT before Seagate dropped that particular configuration recommendation on us, along with a new firmware revision for the drive that promised better performance. We re-tested the XT using the latest firmware and the recommended driver config, and our results barely budged.
Of course, we didn't limit ourselves to testing with the Marvell controller. The Barracuda XT also worked its way through our storage test suite connected to the 3Gbps SATA controller in Intel's P55 chipset. We left the Intel driver in its default config, which has write-cache buffer flushing enabled, contrary to Seagate's claims.
Since Western Digital's Caviar Black 2TB is the XT's most natural competitor in the desktop world, we also threw it into the mix on both the Intel and Marvell controllers. That's as many configurations as we had time to test, but you can extrapolate how the XT might fare against a small collection of other drives by checking out the results of our Caviar Black 2TB review, which will put that drive's performance in context.
We used the following system configuration for testing:
|Processor||Intel Core i7-870 2.93GHz|
|CPU/chipset link||DMI (2GB/s)|
|Motherboard||Asus P7P55D Premium|
|Chipset||Intel P55 Express|
|Memory size||4GB (2 DIMMs)|
|Memory type||Kingston HyperX DDR3-1600 at 1333MHz|
|CAS latency (CL)||9|
|RAS to CAS delay (tRCD)||9|
|RAS precharge (tRP)||9|
|Cycle time (tRAS)||24|
|Audio||Via VT2020 with 6.1.7600.16385 drivers|
|Graphics||Gigabyte GeForce 8600 GT 256MB with ForceWare 190.62 drivers|
Seagate Barracuda XT 2TB with CC12 firmware
Western Digital Caviar Black 2TB
|OS||Windows 7 Ultimate x64|
Our test system was powered by an OCZ GameXStream power supply unit.
With the exception of our power consumption and noise levels, all tests were run at least twice, with the results averaged. We used the following versions of our test applications:
The test systems' Windows desktop was set at 1280x1024 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
|Intel Computex keynote confirms Kaby Lake and Optane for 2016||18|
|Asus shows off Avalon modular case and GX800 liquid-cooled laptop||5|
|Samsung designs minuscule single-package NVMe SSD||20|
|Thermaltake shows off The Tower and more at Computex||9|
|Adata shows NVMe and TLC SSDs at Computex||1|
|Corsair@Computex 2016: fans that levitate, fans that illuminate||8|
|Patriot adds 2TB model to Ignite SSD lineup||12|
|Intel boosts the high-end desktop with its Broadwell-E CPUs||80|
|EVGA@Computex 2016: Custom Pascal cards, new PSUs, and more||8|
|Everyone from Asus to Zotac has announced a non-reference GTX 1080. I see what you did there!||+46|