OCZ’s Vector 180 solid-state drive reviewed

Stop me if you’ve heard this one before. An SSD maker walks into a fab and orders a round of NAND built on the latest process. The chips are swapped into an existing drive based on a familiar controller, the firmware is tweaked as necessary, and the model number is incremented by an arbitrary margin. With a fresh sticker affixed, a new solid-state drive is born.

Lousy punchline, I know. But the joke is on me, because I have to figure out how to write about OCZ’s new Vector 180 without putting you all to sleep.

This is OCZ we’re talking about, so maybe that task won’t be too difficult. The company’s SSDs went through quite a rough patch a few years ago, in part due to their early adoption of SandForce controller technology that was still rough around the edges. OCZ has been rehabilitating its reputation ever since, and it seems to be on the right track since being acquired by Toshiba in late 2013.

So the Vector 180 isn’t just another Barefoot-based SSD with upgraded NAND. It’s the latest chapter in an increasingly convincing comeback story. And it’s no joke.

OCZ’s new hotness replaces the high-end Vector 150 SSD introduced in 2013. It uses the same proprietary Barefoot 3 M00 controller, but it’s based on a newer, “A19” revision of Toshiba’s 19-nm MLC NAND. This is the third OCZ SSD to adopt A19 flash. The budget-oriented ARC 100 and AMD-branded Radeon R7 SSDs use the same NAND—and similar Barefoot 3 controller technology.

In some respects, very little has changed in the SSD arena since 2013. Serial ATA is still the predominant interface, and eight-channel controllers remain the norm. The Barefoot 3 M00 fits right in despite its age.

Higher-capacity SSDs have become more common over the past couple years, and the Vector 180 is OCZ’s first to reach toward terabyte territory. The drive is available in capacities up to 960GB, while the rest of OCZ’s consumer-grade SSDs top out in the 480-512GB range. Here are the performance specs for each capacity:

If those speed ratings look familiar, that’s because the Vector 150 has the exact same specs. Upgrading the NAND doesn’t appear to impart any performance benefits, likely because the 6Gbps SATA interface is the primary bottleneck. We’ve heard numerous SSD makers lament the limitations of Serial ATA, and we can’t wait for the PCIe revolution to begin.

As expected, the 120GB variant is the runt of the litter. Eight-channel controllers typically require at least 32 flash dies to achieve peak performance, and at 8GB per die, the 120GB drive simply doesn’t have enough NAND. The 240GB and 480GB versions also use 8GB dies, while the 960GB taps larger 16GB ones. Interestingly, the top capacity has a lower sustained write speed rating than the 480GB, despite the fact that both employ 64-die arrays.

Like other Barefoot-based SSDs, the Vector 180 can encrypt bits in hardware using a 256-bit AES algorithm. However, it lacks support for the associated IEEE and TCG Opal standards—and, by extension, for Microsoft’s eDrive spec. Plenty of other SSDs support those standards, including drives that cost a lot less than the Vector.

In a new twist, OCZ has added a measure of power-loss protection dubbed Power Failure Management Plus, or PFM+. If the Vector detects “power anomalies,” it uses onboard capacitors to fuel “system-critical routines” before shutting down automatically. As far as I can tell, those routines are largely limited to saving a copy of the mapping table to the flash. This information is normally housed in the drive’s volatile DRAM cache, making it especially sensitive to power loss. The Vector also takes periodic snapshots of its mapping table during normal operation.

Although PFM+ is a nice perk, it’s not as robust as the power-loss protection common on server-grade SSDs. The scheme only protects “at-rest” data that’s already written to the flash—it doesn’t preserve “in-flight” data that has yet to be committed to the NAND.

The Vector 180 is rated to withstand 50GB of writes per day for the length of its five-year warranty. That’s more than enough endurance for most workloads, and it probably under-sells the drive’s true write tolerance. As we’ve seen in our ongoing SSD Endurance Experiment, SSDs can survive far more writes than their official specifications claim.

Unlike typical SSD warranties, the Vector’s Shield Plus coverage is bolstered by an advanced replacement policy. If the drive fails, OCZ will ship a replacement along with a pre-paid shipping label for the faulty unit. There’s no need to dig up the original receipt, either. Only the drive’s serial number is required to initiate a return.

Obviously, it’s better to avoid failures in the first place. The problematic SSDs of OCZ’s past still cast a dark shadow, but the company’s recent products have fared much better. According to internal data OCZ shared with us last summer, the return rates for Barefoot-based drives with Toshiba flash are under 0.1%.

That figure seems a little too good to be true, but the anecdotal evidence suggests failure rates have plummeted since the bad old days. The Amazon and Newegg user reviews for OCZ’s latest SSDs are largely positive, with loads of glowing praise and relatively few complaints. Most of the negative reviews apply to the Vertex 450, which was released before Toshiba bought OCZ, and to the 480GB versions of the ARC 100 and Radeon R7. Customers seem pretty happy with the rest of the lineup, including the Vector 150, the Vertex 460 and 460A, and the other ARC 100 and Radeon R7 capacities.

User reviews are far from gospel, but they’re the best gauge we have for assessing real-world reliability. All the data we’ve seen indicates that OCZ’s newer SSDs are a lot more reliable than their predecessors.

Speaking of data, let’s move on to our performance results…

Betwixt benchmarking rigs
We’re in the midst of replacing our old storage test rigs with new machines—and an updated suite of benchmarks. That transition, coupled with timing for this review, limited the testing we were able to do. We’ll have a more comprehensive look at the Vector 180’s performance in a future article. In the meantime, we’ve run the 240GB and 960GB flavors through arguably the most important test from our old suite, DriveBench 2.0, plus a handful of the new tests we’ll be using moving forward.

TR DriveBench 2.0 — Disk-intensive multitasking
DriveBench 2.0 is a trace-based test comprised of nearly two weeks of typical desktop activity peppered with intense multitasking loads. More details on are available on this page of our last major SSD round-up.

We measure DriveBench performance by analyzing service times—the amount of time it takes drives to complete I/O requests. Those results are split into reads and writes.

To no one’s surprise, the Vector 180 behaves much like OCZ’s other Barefoot-based SSDs. The 240GB and 960GB versions perform similarly—and very well overall. They have the lowest mean write service times we’ve ever measured, and they’re not exactly slouches with reads, either.

All the SSDs execute the vast majority of DriveBench requests in one millisecond or less—too little time for end users to perceive. We can also sort out the number of service times longer than 100 milliseconds, which is far more interesting data. These extremely long service times make up only a fraction of the overall total, but they’re much more likely to be noticeable.

Through nearly two weeks of simulated disk activity, the Vector 180s log just 11 writes over 100 milliseconds. That’s an astoundingly low total considering the sheer volume of I/O involved.

Depending on how one looks at the results, the picture is either better or worse with reads. Although the Vector 180s both post over 900 extremely long service times, that’s fewer than for any other SSD apart from the Vector 150, which is in the same league. All of OCZ’s Barefoot drives crunch through DriveBench 2.0 with relative ease.

IOMeter
The next batch of benchmarks is part of our new suite. We tested the Vector 180s against a couple of other Barefoot drives to get a sense of how the new model compares to its predecessors. Expect our collection of comparative results to grow as benching ramps up on our new rigs.

IOMeter fuels a large component of the new suite, including our sequential and random tests. Those are run across the full extent of the drive at two queue depths. The QD1 tests simulate a single thread, while the QD4 results emulate a more demanding desktop workload. (87% of DriveBench 2.0 requests have a queue depth of four or less.) Click the buttons below the graphs to switch between queue depths.

The Vector 180 240GB runs away with the sequential write speed tests, especially at the higher queue depth. It leads with reads, too, though it’s matched by the Vertex 150 in the QD1 test. Interestingly, the 960GB variant is a step slower with reads—and a full stride behind with writes.

The older OCZ drives aren’t running the latest firmware, so that may explain some of the performance differences between them and the Vector 180. We need to update the firmware on a stack of SSDs before building a full set of results on our new test rigs.

Our first random I/O tests show slim differences between the Vertex 180 and 150. The deltas are just fractions of a millisecond. That said, the 960GB drive has a more notable edge with writes, especially at the higher queue depth.

We’ve reported the median of three consecutive three-minute runs for the preceding tests. Sequential and random read performance is typically consistent over that period, but random write performance worsens as the SSD’s overprovisioned area is consumed by incoming writes. We run a separate, 30-minute random write speed test to capture that decline. This test uses a higher queue depth of 32 to push SSDs into a steady state more quickly.

Click the buttons below the graph to switch between the drives. And note that we’re graphing IOps rather than response times.

The Vector 180 960GB hits a much higher peak random write rate than its 240GB counterpart, and it sustains that burst for longer. Its write rate is more consistent as performance flattens out, too, but it also drops to zero at regular intervals. The 240GB’s slowdowns are somewhat more irregular and less severe. As the test wears on, the smaller drive also benefits from higher spikes than its larger sibling.

Next, we’ll look at scaling across multiple queue depths using a database access pattern made up of 66% reads and 33% writes, all of which are random. We’ve extended this test up to 128 simultaneous requests to accommodate next-gen PCIe drives with deeper queues. Don’t expect SATA drives to scale effectively beyond 32 requests, which is the depth of that interface’s command queue. This time, the clicky-button magic toggles between total, read, and write IOps.

Barefoot drives typically scale well in this test, and the Vector 180 is no exception. Note how the 960GB starts strong but starts lagging behind as the queue depth climbs past 16, though. That dynamic is evident with both reads and writes.