|Introduction and Specifications|
An interesting thing about NAND Flash SSD (Solid State Drive) technologies, beyond the fact that the market is flush with competitive product offerings, is that the technology itself is very flexible and adaptable to a number of different design approaches, other just the straight-forward SATA-based SSDs. Take for example the RAID 4-pack configuration we setup here with Intel's X25-M SSD or perhaps the omnipotent Fusion-io ioDrive. Granted, these are rather high-end, pricey setups, but you get the gist that solid state storage arena is just getting warmed up.
In a sort of hybrid version of Fusion-io's product and our little RAID 4-pack array we setup for testing, memory solutions manufacturer OCZ Technology has introduced an almost fully integrated solution of their SATA-based SSD technology, along with a third party RAID controller, all wrapped up clean and tidy on a standard plug and play (no longer pray) PCI Express X8 adapter card. Dubbed the Z-Drive, we first got a look at this wild-eyed beast back in May. It wasn't quite ready for prime time back then and it re-emerged again in early September with specs that admittedly caused a pavlovian response of our salivary glands.
Today, we get to satiate ourselves with a deep dive look at the new OCZ Z-Drive, in a tasty 256GB variant that drops in at an almost reasonable $899 price point that is on near cost parity with a standard mid-range SSD. This is a very different approach to SSD technology, one that occupies a PCI Express slot instead of a SATA port. First we'll dig into what makes it tick and then we'll see how it ticks through our benchmark time trials.
There are a few notables to take away from the spec list here that tip us off to the fact that we're on to something unique. First, the Z-Drive has a "hardware" RAID controller on board, namely the LSI SAS1068E 8-port SATA/SAS controller. In fact, the underlying main card is a Supermicro AOC-USAS-L4I controller with 256MB of on-board cache. As you'll see on the forthcoming page, the card has been heavily modified, however, and the SAS port on the front plate of the card has been removed. The card requires a PCI Express X8 slot, which means it has more than enough bandwidth to handle the task at hand. Finally, the 256GB model we tested was factory set to a RAID 0 mode configuration in the LSI BIOS with a 4x64GB OCZ Vertex style SSD setup. More details on these specific components, next...
|Under The Hood Of The Z-Drive|
This new version of the OCZ Z-Drive, versus the first iteration of the product we've seen, is significantly more elegant in terms of its design approach. The product assumes a short full-height X8 PCI Express card form factor. However, the width of the card does make it a little tight if there is another card in an adjacent slot. If the adjacent card is something that runs hot, like a graphics card, we'd suggest ensuring there is ample airflow within the case the card is being installed into. There is also a motherboard clearance issue with the card that OCZ has detailed for users here in this document. Essentially, if there are any components, directly in back of the PCIe X8 or X16 slot you're installing it into, that are taller than 8mm (.320-inches) high, the card won't seat in the slot properly. This situation represents a small percentage of motherboards on the market but it's definitely something to be wary of, if you're considering the Z-Drive.
The Z-Drive m84 256GB PCI Express SSD we tested is what OCZ calls their "mainstream" product. It's based on a 4x64GB SSD RAID array and offers a bit less performance versus the p84 (Performance) or e84 (Enterprise) versions of the product. The Enterprise banded model is built with SLC NAND flash technology while the Performance and Mainstream products are made of MLC NAND flash. As you can see, disassembled, the Z-Drive is comprised of four primary components; the base RAID controller board and SSD shell housing, a backplane and power connector plate that provides data and power interfaces to the SSDs and the two SSD cards themselves. The on-board LSI RAID controller also talks to the SSDs over a single connector on the backplate as well.
The flash memory boards are rather interesting designs with a pair of SSDs on a single PCB with a single SATA interface, but each driven by their own Indilinx Barefoot SSD controller. Essentially, the design employs on-board RAID on a single PCB and across board RAID between the two SSD boards, hosted by the LSI RAID controller. Again, since the BIOS is locked down at the factory to a RAID 0 config, there is no significant utility within the LSI RAID BIOS itself, though we'd encourage OCZ to perhaps offer user configurable options (other RAID modes?) in future iterations of the product.
Regardless, as you can see, the solution is nicely packaged and again rather elegant, requiring a single 4-pin Molex power connector for the SSD array while the controller board derives its power from the PCIe slot. Speaking of power; let's power it up.
|The Setup, Methodology and SANDRA|
Our Test Methodologies: Under each test condition, the Solid State Drives tested here were installed as secondary volumes in our testbed, with a standard spinning hard disk for the OS and benchmark installations. The SSDs were left blank without partitions wherever possible, unless a test required them to be partitioned and formatted, as was the case with our ATTO benchmark tests. Windows firewall, automatic updates and screen savers were all disabled before testing. In all test runs, we rebooted the system and waited several minutes for drive activity to settle before invoking a test.
Also, you'll note that we performed all of our individual and RAID SSD testing on an Intel X58 chipset-based motherboard via its ICH10R Southbridge SATA controller. The OCZ Z-Drive was tested in a PCIe X16 slot on the same test bed. For a few of our file transfer tests specifically, we utilized Fusion-io's ioDrive as a source drive to read files from, for our write performance testing, or write files to, for our read performance testing of OCZ's Z-Drive or Intel SATA SSDs.
In our SiSoft SANDRA testing, we used the Physical Disk test suite. We ran the tests without formatting the drives and both read and write performance metrics are detailed below. Please forgive the use of these screen captures and thumbnails, which will require a few more clicks on your part. However, we felt it was important to show you the graph lines in each of the SANDRA test runs, so you are able to see how the drives perform over time and memory location and not just an average rated result.
Looking at these preliminary, high level numbers from SANDRA, the OCZ Z-Drive puts up solid numbers in excess of 400MB/sec for reads and near 700MB/sec for writes, though there are definitely saw toothed variations across different areas of the drive volume. Write performance is obviously a strong suite for the Z-Drive and compared to a 160GB Intel X25-M (Gen 1 80GB SSDs) RAID 0 setup, the Z-Drive shows a 4X performance advantage. Next we'll dig into the numbers a bit deeper with ATTO and CrystalDiskMark.
|ATTO and CrystalDiskMark Benchmarks|
ATTO is a more straight-forward type of disk benchmark that measures transfers across a specific volume length. It measures raw transfer rates for both reads and writes and graphs them out in an easily interpreted chart. We chose .5kb through 8192kb transfer sizes over a total max volume length of 256MB. This test was performed on blank, formatted drives with default NTFS partitions in Windows Vista. We've also provided test results from a couple of different Intel single and RAID 0 SSD setups for reference.
Here the Z-Drive shows its Achilles Heel, that being small file transfers below 64K. In fact, in terms of read performance, the Z-Drive m84 doesn't really stretch its legs until you hit 256K file sizes. Read performance is just plain lackluster in fact, down at 32K and below, where the Z-drive doesn't even keep up with the average single SSD. What's odd is the product's write performance which begins to scale nicely at 16K file size transfers on up. Regardless, you can see, with respect to large sequential transfers, the Z-Drive has significant potential. This test hints to us that there is definitely available optimization to be done with the product, at least with respect to small block reads. If write performance can scale so well, there's really no reason read performance should suffer so much with small transfer sizes. Historically small file writes have been much more of a bottleneck for the average SSD, while reads have been a strength.
CrystalDiskMark is a new synthetic test we've started looking at that tests both sequential as well as random small and large file transfers. It does a nice job of providing a quick look at best and worst case scenarios with SSD performance, best case being large sequential transfers and worse case being small 4K transfers. Once again we've provided some tests on various Intel X25-M SSD configurations just for reference as a baseline.
Intel X-25M G2 - 160GB Single SSD
Intel X25-M G1 X4 RAID 0 - 320GB
OCZ Z-Drive m84 256GB
This test paints a clearer picture of the OCZ Z-Drive under various workloads. If we look at large sequential transfers, the Z-Drive competes handily with even a 4 x Intel X25M RAID 0 array, in fact blowing it out of the water in write performance, but falling a bit short on read throughput, though still offering impressive numbers. At 512K reads and writes the Z-Drive looks similar with significantly better write performance and almost performance parity with the Intel quad RAID 0 array. However, when we drop down to tiny 4K file sizes the Z-Drive takes a knee in a big way and it's not even able to keep pace with a single Intel X-25M SSD. Obviously, looking at all of the numbers at the 4K level, even the Intel RAID array took a massive hit and the Z-Drive actually more than doubled its performance in read throughput, while falling way behind on small block writes.
|HD Tach Testing|
Simpli Software's HD Tach is described on the company's web site as such:
HD Tach shows a very different picture of the Z-Drive. Here the new PCI Express-based SSD tops out with performance that is about on par with the average standard SSD, clocking in at around 200MB/sec for read and write throughput. Comparatively, a dual drive Intel SSD array sports over 500MB/sec for reads and 150MB/sec or so for writes. One thing to note is the significantly lower CPU utilization of the Z-Drive versus the Intel SSD RAID array. The obvious advantage of the Z-Drive's LSI hardware RAID controller is that even under a RAID 0 setup, it requires significantly less host CPU resources; 1% versus the ICH10R software RAID solution's 7% CPU utilization result.
|PCMark Vantage HDD Test Module|
Next we ran the Intel RAID 0 arrays and our 256GB OCZ Z-Drive through a battery of tests in PCMark Vantage from Futuremark Corp. We specifically used only the HDD Test module of this benchmark suite to evaluate all of the drives we tested. Feel free to consult Futuremark's white paper on PCMark Vantage for an understanding of what each test component entails and how it calculates its measurements. For specific information on how the HDD Test module arrives at its performance measurements, we'd encourage you to read pages 35 and 36 of the white paper.
We like PCMark Vantage's HDD Performance for its real-world application measurement approach to testing. From simple Windows Vista start-up performance to data streaming from a disk drive in a game engine and video editing with Windows Movie Maker, we feel that these tests illustrate more of a real-world performance profile of an SSDs in an end user/consumer PC or Workstation usage model.
This specific set of PCMark Vatange HDD tests are generally read-intensive measurements, whether reading images in Windows Photo Gallery or scanning the hard drive for threats in Windows Defender. The Vista Startup and gaming tests are indicative of application loading performance, which are also a read-intensive operations. Unfortunately, under these conditions, the Z-Drive just couldn't keep up with the Intel SSD/ICH10 RAID arrays and with the Windows Defender and Gaming tests it didn't keep pace with even a single Intel X25-M SSD.
|PCMark Vantage HDD (cont.)|
Our next series of Vantage tests will stress the current weakness of most NAND Flash, that being write performance. Applications like video editing, streaming and recording are not what we would call a strong suit for the average SSD, due to their high mix of random write transactions.
Interestingly here, the performance of our Intel X25-M RAID 0 arrays completely skews the graph in the Windows Media Center test. In their white paper, Futuremark claims this specific test "measures concurrent disk drive performance of Media Center tasks, including SDTV video playback, SDTV video streaming to a Windows Media Center Extender and SDTV video recording," with about 50% read and 50% write operations. The Intel RAID arrays completely out-classes the Z-Drive here and in fact the only victory it was able to chalk up was in the application loading test versus a single Intel X25-M SSD.
|Iometer Workstation and Database Results|
The IOMeter Question:
As we noted in a previous SSD round-up article, though IOMeter is clearly thought of as a well respected industry standard drive benchmark, we're not completely comfortable with it for testing SSDs, as well as comparing their performance to standard hard drives. The fact of the matter is, though our actual results with IOMeter appear to be accurate, it is debatable whether or not certain access patterns, as they are presented to and measured on an SSD, actually provide a valid example of real world performance, at least for the average end user. That said, we do think Iometer is a solid gauge for relative available bandwidth with a given storage solution. Regardless, here's a sampling of our test runs with Iometer version 2006.07.27 with the OCZ Z-Drive and our complement of Intel's SSDs.
Here we dropped in a single Intel SSD as well, for a reference baseline metric. In our database or server access pattern, which is comprised of completely random access with 33% dedicated to write transactions, you can see the Intel X25-M RAID array scales dramatically as you add more drives to the equation and turn up the number of IO requests per target. Unfortunately this was not the case with the OCZ Z-Drive where the single card RAID solution topped out at roughly 11K IOPs and didn't scale beyond that as we increased IO queue depth.
In our Workstation access pattern, which consists of only 20% write operations and a bit more sequential access work, there are some rather interesting observations. Again, as you add drives to the Intel RAID 0 array, performance scales relatively well. In this test, the OCZ Z-Drive performed better as compared to its performance in the Database test but it still couldn't keep up with a single Intel 80GB gen 1 SSD.
Again, one thing to note is the transfer size of our standard Iometer tests. At small 8K transfer sizes, the Z-Drive takes a beating. In this next test, we'll simplify things a bit and show you how the Z-Drive fairs versus a single 160GB gen 2 Intel X25-M SSD but with larger file transfers
Here you're looking at an Iometer test run that is similar to our Workstation test pattern, except that we've taken the transfer request size up to 64K, which actually isn't all that large. Though overall IO throughput it lower since obviously it takes longer to transfer the larger file size, you can see the OCZ Z-Drive outperforms the Intel SSD by a large margin, on the order of about 40%.
So let's recap. The Z-Drive's strong suit is larger, sequential file transfers. Where it falls down is with respect to a heavy workload of small random read/write operations. We've observed this in a number of scenarios thus far in our testing. Let's look at a usage model that should play to the Z-Drive's strength, next.
|File Transfer Tests|
Our final series of tests are what you might call more "crude measurements" in that we simply fired up our trusty stop-watch and measured the time it took to complete a copy and paste command of a single large file or a bunch of large files from one storage volume in our test system to another.
** Please note that we utilized a Fusion-io PCI Express SSD card as our source drive in some of the following tests, to read files from or copy files to, for our read and write measurements of both the OCZ Z-Drive and Intel SSD. This affords us the luxury of much higher available bandwidth from the source or target drive (depending on whether we were reading from or writing to it), such that it would not be the limiting factor in a given test condition. Remember the 160GB Fusion-io drive we're using here retails for ~ $7,000, so it's in a class by itself and performs as such. It does make for an excellent test vehicle, though we've also included some measurements with a standard WD Raptor 150GB hard drive as our source or target drive, to offer a more practical usage model and benchmark measurement.
In our single, large file transfer test, we're moving a 3.5GB zip file comprised of several HD video clips, across from a source drive to either the Z-Drive or Intel SSD as our targets for the write test. For the read test, we're moving that file from the Z-Drive or Intel SSD to another target drive in the system. As you can see, we're also showing the Fusion-io drive as a source/target drive, as well as the standard WD Raptor hard drive in this role. Obviously, under the virtually limitless bandwidth setup with the Fusion-io drive as a companion target or source, the OCZ Z-Drive m84 proves itself to be 2 - 5 times faster than a single Intel X25-M gen 2 SSD with this very large file transfer workload. Under the limitation of the WD Raptor drive, we only see a 10 - 25% edge for the Z-Drive but regardless, the potential available bandwidth is there.
Our multi-file transfer test, which is comprised of a dozen or so large 100 - 300MB+ sized files shows more of the same advantage for the OCZ Z-Drive, again when working with a source or target that has large amounts of available throughput itself.
|Game Level Load Tests|
Our final test case couldn't get any more straightforward. Here we'll show you how the OCZ Z-Drive performs under a real-world application loading condition like a game level load-up. Here we simply installed Left 4 Dead on each of the drives we've tested and then proceeded to launch single player games on either the No Mercy or Dead Air levels. Also, as you'll note, we've included the Fusion-io card in this test, for a relative measure of performance.
Again, here in this random, mostly small file read scenario, the OCZ Z-Drive shows itself to be just a tad slower than an Intel 160GB X25-M gen 2 SSD. As an aside, we should note that the Fusion-io drive that we included here for reference, is currently not a "bootable" solution but the Z-Drive is. We quickly loaded up Windows Vista on the Z-Drive and Intel SSD and compared stop-watch times to that of our WD Raptor 150GB hard drive which was the OS volume. As it turned out the Z-Drive booted Vista in about 2 seconds slower than the Intel SSD but 15 seconds faster than the standard WD 10K RPM hard drive.
And actually, if you consider the Z-Drive's cost per GB value proposition, the picture looks a little better. Currently, the average 160GB Intel SSD weighs in at around $3.12 per GB, while the Z-Drive's $899 price tag for a 256GB version weighs in at $3.50 per GB. More on the subject of value proposition in our wrap-up, next...
|The Bottom Line, Sort of...|
To say our results with the OCZ Z-Drive were a mixed bag would be an understatement. Early on in our testing, we literally got the notion that we needed to "look" for ways to demonstrate the product's strength. Too many times under synthetic benchmark testing, the Z-Drive fell flat, but there were also test conditions that showed big gains for the product, as was seen in our ATTO larger file transfer tests and CrystalDiskMark tests. One thing is for certain, small file random operations are not a strength for the product, but clearly large file transactions, especially write transaction are its strength. These results had us digging deeper with respect to demonstrating real-world test conditions for the Z-Drive where it would excel and also where it wouldn't. As we garnered from our custom large file transfer tests, the Z-Drive was handily up to the task. And as we showed you in our Left 4 Dead game level load tests, the Z-Drive shows mostly competitive performance on par with only a single standard SATA SSD.
To be completely frank (or Ralph or Harry for that matter), when we consider the sum of its parts, we expected more from the Z-Drive. What we're looking at here is an SSD RAID card with a full-fledged RAID processor on board along with dedicated cache memory, all talking over a PCI Express X8 link with a ton of bandwidth at its disposal. We expected the product to put up a fight with the likes of at least an SSD RAID solution driven by a software RAID controller over a motherboard's integrated Southbridge. That, however, simply wasn't the case under many everyday test conditions we looked at, save for large file transfers. In reality, to us, the Z-Drive feels a little less than fully baked. We have seen similar oddities with hardware RAID controllers on the market, in conjunction with SSDs in RAID 0, so perhaps that points to the primary issue.
We haven't explored other RAID modes much, but it appears that at least some of the RAID controllers on the market today aren't optimized all that well for SSDs and are apparently still tuned to hide the latencies of standard spinning hard disks, when they really need to be tuned to take advantage of the exponentially lower latency response and access times of the average SSD. Perhaps with some time and a few BIOS revisions, OCZ can get together with LSI and improve the product's performance for more mainstream tasks.
That said, we feel that currently, the Z-Drive is best suited for a few key niche' applications and usage models only. If you're the type of end user that needs to drive large file write and read operations, as would be the case for a digital video production for example, then the OCZ Z-Drive m84 256GB model we tested would serve you well. In addition, if you consider the cost per GB of the product, versus competitive SSDs on the market, the Z-Drive drops in with a moderate price premium over a single 256GB SATA SSD, on the order of about 20 - 25%. However, you'll still need to consider the reliability aspects of a SSD RAID solution even with OCZ's 1 million hr MTBF specification and standard 3 year warranty. When all is said and done, we think the Z-Drive is an interesting product that partially exploits current technologies available in the market, but perhaps needs a bit more maturation time in the oven in order to appeal to the enthusiast, power user or mainstream user.