Re-Lighting Patriot’s Inferno

This drive was very pleasing to review the first go-round. As a practical performer, it completely blew me away. Unfortunately the numbers were slightly marred by an oversight of monumental proportions – the benches were run in IDE mode. Thanks to astute observer jonsrud, who pointed out the error of my ways,  we’re back with some new numbers with a touch of analysis thrown in!

AHCI – Advanced Host Controller Interface – What is it?

Most modern SATA controllers offer AHCI along with the default Legacy IDE mode for compatibility with older drives. The two big advantages AHCI has are the ability to hot-swap SATA devices and support for native command queuing (NCQ). There are others (improved power management, port multiplication), but the largest performance gain comes from NCQ.

When it was first introduced, NCQ was for mechanical hard drives. Thanks to the ever-astute Wikipedia, we learn that NCQ “is a technology designed to increase performance of SATA hard disks under certain conditions by allowing the individual hard disk to internally optimize the order in which received read and write commands are executed.”

Basically, the drive can pay more attention to where the data is on the platter and less on which track it’s on. The image below explains it better than I could.

With NCQ, the read/write heads can jump from track to track and read the data as it comes around. Without NCQ, the drive read/write heads go from inner track to outer track in order, which causes more drive revolutions to read the same amount of data. Obviously, that is not what NCQ does to help with SSDs.

While it’s used to compensate for rotational delays in standard HDDs, in SSDs it’s used to compensate for the system lagging behind the drive. Yep, they’re so fast, you need NCQ to help the rest of your system to keep up.

NCQ allows your system to parallelize (spell check says that’s a word) commands to your drive – up to 32 at a time. Not only does that speed things up quite a bit, it also helps reduce the number of writes to the flash on a SSD, assisting in increasing the longevity of flash in your drive.

Not all SSD controllers support NCQ but most in the current generation do. Even so, not all will benefit from it. From this PC Perspective article,

• “Indilinx controllers, while very good, take almost no advantage of NCQ at all. In the rare case that it does, it’s only up to a queue depth of 4…”
• “The new JMicron controller is leaps and bounds above their older offerings, but it does not see any performance advantage with NCQ.”
• “Sandforce scales nicely when hit with parallel IO’s and will easily scale past 10,000 IOPS in such a case.”
• Intel is “the single best example of an SSD to demonstrate…” the difference between AHCI and IDE.

As you can see, there are definitely advantages over running plain old IDE mode. It’s easy to measure the advantages thankfully…give the drive a secure erase (via hdparm – here’s how) and re-bench in AHCI mode. I thought that was going to be somewhat painless; it’s easy enough to switch to AHCI mode in Windows 7 even after the OS is installed (Vista is similar and XP is a bit more complicated), so that’s what I did.

Unfortunately, that wasn’t the best way to go about it in this case. Upgrading from IDE to AHCI mode on an existing Windows install was not netting the expected performance gains on my system. After conferring with several forum members, we decided a fresh install with native AHCI from the start was the only way to properly evaluate the drive. The full benchmark suite wasn’t run on both chipsets prior to the fresh install, but like me, you may see some gain by reinstalling from scratch. “How much?” is questionable in every-day usage.

So since I had the drive all cleaned and happy and since I had two systems practically right next to each other, I reinstalled fresh OS copies on both of them. That way this is a worthy read, not only to compare IDE to AHCI, but also to compare chipsets! Who doesn’t like a little healthy competition?

Systems Compared

Both fresh installs used the most recent AHCI drivers from Intel and AMD. To remove as many variables as possible, for the purposes of these tests both CPUs were run at their stock speed (coincidentally, that’s 3.2 GHz for both) and RAM was run on both systems at DDR3-1600 / 9-9-9-24.

AS SSD

AS SSD Benchmark (Read Tests)

No question how much AHCI helps with reads. The SB850 comes out on top here, with P55 just edging it out in sequential read tests.

AS SSD Benchmark (Write Tests)

SB850 / AHCI combination are on top when writing too. IDE actually didn’t have a bad showing for write tests, coming very close to the AHCI results.

AS SSD Benchmark Score

The total score reflects the healthy gains from AHCI and shows the SB850 just beating the P55.

CrystalDiskMark

Crystal DiskMark x64 (Random Data – Read Tests)

Crystal DiskMark x64 (Random Data – Write Tests)

Again, read tests are squarely in AHCI’s favor. Write tests are quite the surprise. Not only are sequential and 512 K writes close, IDE beats AHCI both 4K and 4K QD32 write tests.

The SB850 is on top overall here as well.

Crystal DiskMark x64 (0Fill – Read Tests)

Crystal DiskMark x64 (0Fill – Write Tests)

Crystal DiskMark x64 (1Fill – Read Tests)

Crystal DiskMark x64 (1Fill – Write Tests)

0 and 1 fill tests tell a similar story, just with larger numbers; it’s easier to write ones or zeros instead of random data, so that’s to be expected.

As far as chipsets, with each winning two tests, reads are a draw. Write tests show the P55 overtaking the SB850 in several categories. Even so, I’d trust the random numbers over these for everyday use, and the SB850 rules the day in both benchmarks so far.

ATTO Disk Benchmark

This one is more difficult to chart, so I’ll just put the three screenshots out there for you to peruse.

Hello P55, welcome to the party! Not only did AHCI show clear gains over IDE, but the P55 walked away with this bench, besting the SB850 in both reads and writes.

IOMeter

Regretably, IOMeter wasn’t run in IDE mode. I’m introducing it to flesh out future SSD results. It’s a strong metric of performance and deserves a place alongside its faster, more user friendly brothers. The build used is 2008-06-28.

To run it I used configurations from the OCZ forums that appear to no longer be there. In case you wish to run Iometer on your own, you can get the four configurations here.

Iometer 4K Random Read

Iometer 4K Random Write

Well then, it sure looks like they are very similar, with the SB850 getting the 4K crown due to its stronger read performance. Writes were a dead heat and merely a margin of error away from each other.

CPU utilization was somewhat interesting, trading places on write vs. read, with AMD requiring more on the write tests and Intel requiring more on the read tests.

Iometer 2MB Sequential Read

Iometer 2MB Sequential Write

Intel takes over in the 2MB Sequential read category, besting AMD in all categories except CPU utilization. Considering that’s less than two percent, we’ll call that a non-issue.

Final Thoughts & Conclusion

Before we draw any conclusions, here are the chipsets side-by-side in AHCI mode.

Read Tests

Write Tests

Hmm. Well, you were promised a comparison, not a clear-cut winner. It seems the SB850 is better at reading data and the P55 is better at writing it. Even so, the advantages between the two are benchmark-detectable only and no one would ever be able to tell the difference in daily use.

To the real purpose of this article (chipset comparison was just incidental), AHCI is definitely worth the little effort it takes to enable in Windows Vista and Windows 7. If you’re on XP, it might be more convenient to wait until your next reinstall, which is just a matter of time on XP. For sure, in any situation, if you’re installing from a fresh, formatted drive it would behoove you to switch to AHCI from the get-go. There is very little effort involved and as you can see quite a bit to gain on a SandForce drive.

–Jeremy Vaughan (hokiealumnus)