Sign in with
Sign up | Sign in
Adata Premier Pro SP920 SSD: From 128 To 1024 GB, Reviewed
By ,
1. Adata's SP920: Quite Literally, A Familiar Face

Adata is one of several solid-state storage purveyors in somewhat of a bind. These companies, SandForce partners all, are currently facing a problem: the next-generation controller is late. Like, super-late. If you're Adata, relying on SandForce's turnkey controller and firmware combination, no new architecture means no new products.

The last time we checked, SF-3000-series silicon was scheduled to ship in the fall. Now it's looking like we might have to wait until 2015. The hardware is real. And it's fast, from what we've already seen. But it's not yet ready to sell. 

So, what do you do if you're Adata?

You start looking for something to fill the gap, particularly in the retail space (OEM sales aren't as sensitive to the typical consumer mentality). Us enthusiasts are all about higher-end specs and lower price tags. But we also tend to track the perception of brands. And Adata certainly doesn't want to fall off the face of the planet. Maintaining a presence is important, and that's hard to do without a compelling product receiving attention.

That's why Adata needs this new Premier Pro SP920. Without a new LSI SandForce-based product to wedge into its line-up, the company was looking for something fast and fierce to introduce, keeping the brand relevant in an always-evolved storage market.

But First, A Little History

In August of last year, I had the opportunity to walk the exhibit floor of Flash Memory Summit 2013. There was one constant amongst the big SandForce partners: a lack of upcoming products rocking new silicon. At LSI's Accelerating Innovation Summit, some of those same vendors had prototype SF-3000 based drives on display. Even then, we knew the next-gen controller still wasn't ready. And you know what? That's fine. If the company needs more time to iron out firmware kinks and optimize some corner cases, we're not going to complain.

But a partner like Adata might. The same for Mushkin or PNY. SandForce became the processor of choice for much of the SSD industry, and if you're only selling LSI-based drives, you have to take the lean times with the fat. Now is one of those lean times, which is why Adata is introducing the SP920, a product not based on SandForce technology.

You won't see the SP920 in mSATA or M.2 form, nor will it be an OEM product. It's retail-only, making it a product built for you, the enthusiast reading Tom's Hardware.

All of this would be business as usual, except for a curious turn of events that started when Adata delayed the SP920 for a few weeks. Then, we published The Crucial M550 SSD Review: Striking Back With More Performance. And as soon as the virtual ink dried, we received Adata's drives. Coincidence? Not exactly...

Adata SP920

128 GB
256 GB
512 GB
1024 GB
Controller
Marvell 88SS9189-BLD2
Form Factor/ Interface
7 mm, 2.5" SATA 6 Gb/s
NAND128 Gb density, Micron 20 nm L85A ONFi, 200 MT/s
NAND Part
MT29F128G08CBCABH6MT29F256G08CECABH6MT29F512G08CKCABH7
DRAM
Micron LPDDR3 128 MB
Micron LPDDR3 256 MBMicron LPDDR3 512 MB
Micron LPDDR3 1024 MB
Max Seq Read/Write (MB/s)
560 / 180
560 / 360
560 / 470
Max 4 KB Read/Write (IOPS)
80,000 / 45,000
96,000 / 80,000
98,000 / 88,000
Price
$90
$169
$335
$530
Warranty3 Years

So, here's the line-up. As with Crucial's just-reviewed M550, the Marvell 9189 controller make another appearance. Notice also that all four drives use IMFT L85-series NAND. In fact, if you read my M550 review, you're probably seeing a lot of similarities.

One notable difference is that Crucial's M550 uses 64 Gb die to bolster performance via interleaving in its 128 and 256 GB capacities. Adata's comparable models use 128 Gb NAND, though, halving the number of dies and conveying specifications that look a lot more like the M500 as a result.

Adata includes a 7 mm-to-9 mm spacer and a 3.5\Adata includes a 7 mm-to-9 mm spacer and a 3.5" sled for desktop installs

Inside The Adata SP920

Here's where the story gets more complicated. Normally we'd show you some interior shots of the drive's internals and proceed on to our performance-oriented evaluation. However, we have to pull up short and point out the striking similarities between the SP920 and Crucial's M550.

We begin with the least-functional part of the SSD, its chassis.

Surprise! That's actually the enclosure shot from my M550 piece. Why is it here? First, because I took a fairly nice photo, if I may say so. Also, the SP920 and M550 have this component in common, right down to the production line QR code and pink gap pads. And that's not all they share.

Inside, we find a PCB that also looks familiar. This 512 GB SP920 hosts eight packages of Micron MT29F256G08CECABH6-10:A. That's a mouthful, but it means we're looking at 20 nm NAND sporting two dies per package, along with two chip enables per. Add that up, and we get 32 GiB per package, totaling 256 GiB on each side, and 512 GiB total. For those of you keeping score at home, Crucial's 512 GB M550 also uses this flash, though Micron's abbreviated part code on the package itself is slightly different. This flash can operate synchronously or asynchronously, and is purported to operate at just 200 MT/s. 

Flip the drive over and things get even spookier. There are the other eight packages, plus Marvell's 9189 controller and Micron's low-voltage LPDDR3 (512 MB, in this capacity). The new processor offers improved DevSlp and idle state power draw, but also bumps up bandwidth.

Also present are capacitors to help enable power-loss protection. Again, where have you seen this before?

It's time to stop being coy and just come out with it: Adata's Premier Pro SP920 is not the company's own design. Of course, it isn't 2009 anymore, and we tend to expect more in the way of innovation from new products. But a lot of companies have been down this road before, and in Adata's case, it's probably a smart decision.

Why? Well, although Adata has experimented with non-SandForce-driven SSDs before, this time it needed something fast, and building a new storage product takes time. Adata has extensive production and engineering capabilities. But taking a processor like Marvell's 9189, building up a PCB, and optimizing a firmware doesn't happen overnight. The company had to find a stopgap quickly, so it took drastic measures and adopted the sexy-hot Crucial M550 platform. How can I be so sure? Great question...

2. A Primer: The Art Of The Platform, SMART, And You

There are a couple of ways we can tell that Adata's SP920 shares more than its controller with Crucial's M550 if you don't want to dismantle your own SSDs.

Before I popped the top on our own samples, I thought to myself, "Oh, it looks like Adata leased Micron's firmware IP for the SP920." On the next page, you'll see that both drives use the same firmware revision, MU01. The M500 is up to MU03 MU05 now, but Crucial's M550 launched on MU01. That the SP920 employs similar nomenclature was the first real indicator of something amiss. And there's more to it than that.

You see, Marvell doesn't sell firmware. It simply sells its controllers. Whether Plextor, Micron, or SanDisk chooses to implement a feature is up to each company. Generally, that also means firmware has to be developed, creating a pretty big barrier to entry before you're able to take a Marvell-powered SSD and start selling it. For Adata to come out with something new from Marvell as it waits on LSI, and to do that quickly, required getting its hands on software that was fully-baked already.

And if the matching firmware versions hadn't tipped me off, there is another surefire way to know that two Marvell-based solutions are identical: SMART data.

Self-Monitoring, Analysis and Reporting Technology data is defined by the firmware architecture. If you want to track host LBAs written or power-on time, the firmware author has to enable that functionality. Since each Marvell-based implementation is unique, we see SanDisk using certain SMART attributes, while Plextor uses others. Sometimes they overlap; sometimes they don't. Micron's own implementation tracks RAIN activity. That is, if some flash fails, RAIN uses parity to recover the information that would have been otherwise unrecoverable. Marvell's newest controllers can use parity to protect against data loss as well, but RAIN is a beast of Micron's own creation.

So, when we see this, we know what's up:

SMART Attributes (Raw Values, Decimal)
Crucial M500 480 GB
Crucial M550 512 GB
Adata SP920 512 GB
01 Raw Read Error Rate
1
0
0
05 Reallocated Sector Count
1
0
0
09 Power On Hours
181
247
83
0C Power Cycle Count
50
23
16
AB Program Fail Count
0
0
0
AC Erase Fail Count
0
0
0
AD Average Block Erase Count
56
56
40
AE Unexpected Power Loss
36
17
12
B4 Unused Reserve NAND Blocks
8218
4403
4403
B7 SATA Interface Downshift
0
0
0
B8 Error Correction Count
0
0
0
BB Reported Uncorrectable Errors
0
0
0
C2 Temperature
171800002583
197570068506
180390133785
C4 Reallocation Event Count
17
16
16
C5 Current Pending Sector Count
0
0
0
C6 Smart Offline Uncorrectable Error Count
0
0
0
C7 Ultra DMA CRC Error Rate
0
3
1
CA Percent Lifetime Used
1
1
1
CE Write Error Rate
0
0
0
D2 Successful RAIN Recovery Count
346
0
0
F6 Total Host Sector Writes
26638204322
20741567941
15230511355
F7 Host Program Page Count
803332972
652193216
480081235
F8 FTL Program Page Count
1181515239
1240751859
724621376

I don't have a complete list of Micron's SMART data, but most attributes are carried over from previous models. Adata's SP920 isn't detected as exactly the same drive, so attribute names may vary based on the utility you use to view them. For example, Adata's own toolbox doesn't know the true names of each attribute. CrystalDiskMark has it mostly correct for the M500, and thus the M550 and Adata SP920. The raw values are don't change either way.

Some of these attributes are found in other drives as well. Some aren't. Intel's dalliance with Marvell's 9175 controller in the SSD 510 (a short-lived product that tided the company over until its SandForce-based solution was ready, oddly enough) featured Intel's own SMART attributes from the X25 days. And when the transition to SandForce silicon happened, those attributes followed.

So there it is. Had Adata used its own PCB and enclosure, we might not have stumbled upon this mystery. But we did, dug deeper, and the SMART data doesn't lie.

While we're here, though, let's make some observations. First, we're displaying raw decimal values. D2 is RAIN recovery count. The M550 and Adata SP920 both report values of zero. However, our 480 GB M500 didn't make it through the last year unscathed. I don't know exactly what that field represents. It could be a number of blocks, or actual data in KB or MB. Perhaps part of the flash went belly-up, prompting the drive to recalculate affected area values from parity.

And that's why it's good to have RAID support on Crucial's M550 and Adata's SP920. Micron was aggressive in transitioning to 128 Gb, 20 nm flash. While the process is much more mature today, we still like having a safety mechanism to protect our valuable information. The M500's parity ratio is 1:15, which is where the 480 GB capacity comes from. The M550 platform employs 1:127 parity to storage blocks, tying up a lot less of the drive's capacity. One gigabyte out of every 16 is reserved for RAIN on the M500. That's only one out of every 128 on the M550 and SP920.

I also love that Micron includes thorough write counter metrics. F6 is total host sector writes (think of this as the writes requested by the operating system in 512-byte sectors). Do the math for our 480 GB M500 and you get 12,702.09 GiB, or 26,638,204,322 sectors * 512 bytes each. Most drives expose this counter, and it's useful as a tool for calculating the beating a drive has endured.

Write amplification wears a drive down over time, reducing its performance. It's often a product of internal overhead, for example, shuffling a partially-filled block around during a program/erase cycle. This is normal. But it can greatly reduce the life of a client drive used for high-intensity write workloads, which is why enterprise-oriented SSDs are often substantially over-provisioned. The additional free space helps keep P/E cycles to a minimum.

We can see the result of that additional overhead through the F8 attribute, FTL program page count, which measures the number of 16 KB pages the controller has had to program. Our M500 went through difficult testing; I punished it over and over with long runs of full-span high-queue depth writes. Over its life, I've hit it with 18,028.50 GiB worth of page program operations. Compare that number to the host writes (12,702.09 GiB), and you can calculate that I've subjected the M500 to an extra 5300 GiB of shuffling and churning.

3. Test Setup And Benchmarks

Our consumer storage test bench is based on Intel's Z77 Platform Controller Hub paired with an Intel Core i5-2400 CPU. Intel's 6- and 7-series chipsets are virtually identical from a storage perspective. We're standardizing on older RST 10.6.1002 drivers for the foreseeable future.

Updates to the RST driver package occasionally result in subtle performance changes. They can also lead to some truly profound variance in scores and results as well, depending on the revision. Some versions flush writes more or less frequently. Others work better in RAID situations. Builds 11.2 and newer support TRIM in RAID as well. Regardless, results obtained with one revision may or may not be comparable to results obtained with another, so sticking with one version across all testing is mandatory.

Test Hardware
ProcessorIntel Core i5-2400 (Sandy Bridge), 32 nm, 3.1 GHz, LGA 1155, 6 MB Shared L3, Turbo Boost Enabled
MotherboardGigabyte G1.Sniper M3
MemoryG.Skill Ripjaws 8 GB (2 x 4 GB) DDR3-1866 @ DDR3-1333, 1.5 V
System Drive Intel S3500 480 GB SATA 6 Gb/s, Firmware: 0306
Drive(s) Under TestAdata SP920 1024 GB SATA 6 Gb/s, Firmware: MU01

Adata SP920 512GB SATA 6 Gb/s, Firmware: MU01

Adata SP920 256 GB SATA 6 Gb/s, Firmware: MU01

Adata SP920 128 GB SATA 6 Gb/s, Firmware: MU01
Comparison Drives
Crucial M550 1024 GB SATA 6 Gb/s, Firmware: MU01

Crucial M550 512 GB SATA 6 Gb/s, Firmware: MU01

Intel SSD 730 480 GB SATA 6 Gb/s, Firmware: L2010400

Samsung 840 EVO mSATA 120 GB, Firmware: EXT41B6Q

Samsung 840 EVO mSATA 250 GB, Firmware: EXT41B6Q

Samsung 840 EVO mSATA 500 GB, Firmware: EXT41B6Q

Samsung 840 EVO mSATA 1000 GB, Firmware: EXT41B6Q

SanDisk X210 256 GB, Firmware X210400

SanDisk X210 512 GB, Firmware X210400

Intel SSD 530 180 GB SATA 6Gb/s, Firmware: DC12

Intel SSD 520 180 GB SATA 6Gb/s, Firmware: 400i

Intel SSD 525 180 GB mSATA, Firmware: LLKi

SanDisk A110 256 GB M.2 PCIe x2, Firmware: A200100

Silicon Motion SM226EN 128 GB SATA 6Gb/s, Firmware: M0709A

Crucial M500 120 GB SATA 6Gb/s, Firmware: MU02

Crucial M500 240 GB SATA 6Gb/s, Firmware: MU02

Crucial M500 480 GB SATA 6Gb/s, Firmware: MU02

Crucial M500 960 GB SATA 6Gb/s, Firmware: MU02

Samsung 840 EVO 120 GB SATA 6Gb/s, Firmware: EXT0AB0Q

Samsung 840 EVO 240 GB SATA 6Gb/s, Firmware: EXT0AB0Q

Samsung 840 EVO 480 GB SATA 6Gb/s, Firmware: EXT0AB0Q

Samsung 840 EVO 1 TB SATA 6Gb/s, Firmware: EXT0AB0Q

SanDisk Ultra Plus 64 GB SATA 6Gb/s, Firmware: X211200

SanDisk Ultra Plus 128 GB SATA 6Gb/s, Firmware X211200

SanDisk Ultra Plus 256 GB SATA 6Gb/s, Firmware X211200

Samsung 840 Pro 256 GB SATA 6Gb/s, Firmware DXM04B0Q

Samsung 840 Pro 128 GB SATA 6Gb/s, Firmware DXM04B0Q

SanDisk Extreme II 120 GB, Firmware: R1311

SanDisk Extreme II 240 GB, Firmware: R1311

SanDisk Extreme II 480 GB, Firmware: R1311

Seagate 600 SSD 240 GB SATA 6Gb/s, Firmware: B660

Intel SSD 525 30 GB mSATA 6Gb/s, Firmware LLKi

Intel SSD 525 60 GB mSATA 6Gb/s, Firmware LLKi

Intel SSD 525 120 GB mSATA 6Gb/s, Firmware LLKi

Intel SSD 525 180 GB mSATA 6Gb/s, Firmware LLKi

Intel SSD 525 240 GB mSATA 6Gb/s, Firmware LLKi

Intel SSD 335 240 GB SATA 6Gb/s, Firmware: 335s

Intel SSD 510 250 GB SATA 6Gb/s, Firmware: PWG2

OCZ Vertex 3.20 240 GB SATA 6Gb/s, Firmware: 2.25

OCZ Vector 256 GB SATA 6Gb/s, Firmware: 2.0

Samsung 830 512 GB SATA 6Gb/s, Firmware: CXMO3B1Q

Crucial m4 256 GB SATA 6Gb/s Firmware: 000F

Plextor M5 Pro 256 GB SATA 6Gb/s Firmware: 1.02

 Corsair Neutron GTX 240 GB SATA 6Gb/s, Firmware: M206
Graphics
MSI Cyclone GTX 460 1 GB
Power Supply
Seasonic X-650, 650 W 80 PLUS Gold
Chassis
Lian Li Pitstop
RAID
LSI 9266-8i PCIe x8, FastPath and CacheCade AFK
System Software and Drivers
Operating
System
Windows 7 x64 Ultimate
DirectX
DirectX 11
Drivers
Graphics: Nvidia 314.07
RST: 10.6.1002
IMEI: 7.1.21.1124
Generic AHCI: MSAHCI.SYS
Benchmarks
ULINK DriveMaster 2012
JEDEC 218A-based TRIM Test
Tom's Hardware Storage Bench v1.0
Trace-Based 
Iometer 1.1.0# Workers = 1, 4 KB Random: LBA=16 GB, varying QDs, 128 KB Sequential, 8 GB LBA Precondition, Exponential QD Scaling
PCMark 7
Secondary Storage Suite
PCM Vantage
Storage Suite
4. Results: Sequential Performance

Fantastic sequential read and write performance is a trademark of modern SSDs. To measure it, we use incompressible data over a 16 GB LBA space, and then test at queue depths from one to 16. We're reporting these numbers in binary (where 1 KB equals 1024) instead of decimal numbers (where 1 KB is 1000 bytes). When necessary, we also limit the scale of the chart to enhance readability.

128 KB Sequential Read

As we've seen from some of our recent SSD reviews, performance at low queue depths starts slow, and scales up as outstanding commands pile up. I speculate that our 128 KB sequential workload isn't intense enough to fully saturate Adata's SP920s. Perhaps its an artifact of larger page sizes (16 KB in this case).

All four drives are more or less equal in this test, peaking at almost 540 MB/s at a queue depth of four. I'm using 128 KB chunks of data, whereas a test utility like AS SSD is going to use somewhere in the range of 1 to 4 MB.

128 KB Sequential Write

128/256 GB Adata SP920

I'll start by write testing the two smaller SP920s, and comparing them to the older Crucial M500 equivalents. Why am I dipping back to the 120 and 240 GB M500s? Unfortunately, Crucial didn't send along its 128 and 256 GB models for review, which benefit from twice as many dies as Adata's SSDs. So instead of making this face-off about interleaving, it's instead a measure of firmware and controller performance. Even with the same NAND, we realize a notable speed-up from the two Adata offerings.

The 128 GB Adata SP920 achieves 177 MB/s, which is quite a bit faster than the 120 GB M500's result in the mid-130 MB/s range. The 256 GB model's jump is even more impressive; it registers an extra 100 MB/s. That's awesome, and very close to Adata's official specification. However, the 256 GB M550 is supposed to achieve 500 MB/s. Those extra dies prove useful in a test like this one.

There is a tradeoff here.

512/1024 GB Adata SP920

The larger models fare far more predictably, and we see total parity between the bigger drives with the same number of dies. Performance that falls just shy of 500 MB/s is a result of Micron's flash and firmware. It's about 100 MB/s more than the M500 delivers, though the real fruits of this platform are found elsewhere.

Here's a breakdown of the maximum observed 128 KB sequential read and write performance with Iometer:

The SP920s fall a few megabytes per second short in our benchmark, which are enough to put Adata's largest two models behind the Crucial M550s at the top. In any case, that small amount of difference is imperceptible.

5. Results: Random Performance

We turn to Iometer as our synthetic metric of choice for testing 4 KB random performance. Technically, "random" translates to a consecutive access that occurs more than one sector away. On a mechanical hard disk, this can lead to significant latencies that hammer performance. Spinning media simply handles sequential accesses much better than random ones, since the heads don't have to be physically repositioned. With SSDs, the random/sequential access distinction is much less relevant. Data are put wherever the controller wants it, so the idea that the operating system sees one piece of information next to another is mostly just an illusion.

4 KB Random Read

Testing the performance of SSDs often emphasizes 4 KB random reads, and for good reason. Most system accesses are both small and random. Moreover, read performance is arguably more important than writes when you're talking about typical client workloads.

128/256 GB Adata SP920

Against the smaller M500s, Adata's SSDs establish a slim lead at a queue depth of one and lead from there. Low-queue depth random read performance is a very relevant measurement, and the SP920s manage to improve, even using (more or less) the same flash. Likely, the gains come from the updated controller and firmware instead. 

Of course, none of these drives can come close to Samsung's 840 EVO, which is up to 25% quicker at a queue depth of one.

With 32 outstanding commands, Adata's 128 GB SP920 is on par with the 240 GB M500.

512/1024 GB Adata SP920

It comes as no surprise that the 512 and 1024 GB SP920s are functionally identical to Crucial's M550s at the same capacity points, even though were able to coax a few hundred more IOPS out of Crucial's drives. That's pretty standard though, and we might see the same variance from the same SSD on different runs.

4 KB Random Writes

Random write performance is also important. Early SSDs didn't do well in this discipline, seizing up even in light workloads. Newer SSDs wield more than 100x the performance of drives from 2007, though we also recognize that there's a point of diminishing returns in desktop environments.

128/256 GB Adata SP920

Sporting fewer dies, these lower-capacity drives tap out earlier. Our workload is sufficiently intense to saturate the higher-density configurations at lower queue depths.

Still, we get 16,000 more IOPS from the 256 GB SP920 compared to the 240 GB M500, and 10,000 more IOPS from Adata's 128 GB model versus the 120 GB M500. The drives feature the same number of addressable NAND elements, so most of the speed-up, again, likely comes from the controller.

512/1024 GB Adata SP920

We hit more than 80,000 IOPS with just four outstanding commands. The M550 platform wrings out over 91,000 tested similarly. These drives plateau at a queue depth of eight, so latency is best there. Beyond that, every model offers the same performance, albeit at higher latency.

Random Performance Over Time

My saturation test consists of writing to each drive for a specific duration with a defined workload. Technically, it's an enterprise-class benchmark, where the entire LBA space of the SSD is utilized by a random write at high queue depths.

Here's 12 hours of a 4 KB write with 32 outstanding commands. First, we secure erase each drive. Then we apply the 4 KB write load, showing the average IOPS for each minute (except for the last 20 minutes, where we zoom in and show you one-second average increments).

After the first drive fill, performance drops off fast, since the SSD no longer has free blocks to write to. Instead, they have to be erased prior to subsequent writes.

Today we're showing a breakout of both drive families at steady state for a 4 KB write at a queue depth of 32.

I wanted to use this workload to show that the Crucial M550 and Adata SP920 drive families behave similarly. More so than just the raw performance numbers, we want to characterize these SSDs. And as I might have predicted, this test shows that two identically-configured devices (the 512 GB models, specifically), are more or less the same.

Here's a breakdown of the maximum observed 4 KB sequential read and write performance with Iometer. The order the drives appear in our chart is determined by maximum combined read and write performance.

Drives like the 1024 GB SP920 are about as fast as the SATA 6Gb/s interface will let them be.

The overhead associated with small transfers brings down the throughput ceiling. That is, it's easy to saturate SATA with large sequential operations. But these drives demonstrate you can do somewhere around 400 MB/s with aligned 4 KB blocks.

6. Results: Tom's Hardware Storage Bench v1.0

Storage Bench v1.0 (Background Info)

Our Storage Bench incorporates all of the I/O from a trace recorded over two weeks. The process of replaying this sequence to capture performance gives us a bunch of numbers that aren't really intuitive at first glance. Most idle time gets expunged, leaving only the time that each benchmarked drive is actually busy working on host commands. So, by taking the ratio of that busy time and the amount of data exchanged during the trace, we arrive at an average data rate (in MB/s) metric we can use to compare drives.

It's not quite a perfect system. The original trace captures the TRIM command in transit, but since the trace is played on a drive without a file system, TRIM wouldn't work even if it were sent during the trace replay (which, sadly, it isn't). Still, trace testing is a great way to capture periods of actual storage activity, a great companion to synthetic testing like Iometer.

Incompressible Data and Storage Bench v1.0

Also worth noting is the fact that our trace testing pushes incompressible data through the system's buffers to the drive getting benchmarked. So, when the trace replay plays back write activity, it's writing largely incompressible data. If we run our storage bench on a SandForce-based SSD, we can monitor the SMART attributes for a bit more insight.

Mushkin Chronos Deluxe 120 GB
SMART Attributes
RAW Value Increase
#242 Host Reads (in GB)
84 GB
#241 Host Writes (in GB)
142 GB
#233 Compressed NAND Writes (in GB)
149 GB

Host reads are greatly outstripped by host writes to be sure. That's all baked into the trace. But with SandForce's inline deduplication/compression, you'd expect that the amount of information written to flash would be less than the host writes (unless the data is mostly incompressible, of course). For every 1 GB the host asked to be written, Mushkin's drive is forced to write 1.05 GB.

If our trace replay was just writing easy-to-compress zeros out of the buffer, we'd see writes to NAND as a fraction of host writes. This puts the tested drives on a more equal footing, regardless of the controller's ability to compress data on the fly.

Average Data Rate

The Storage Bench trace generates more than 140 GB worth of writes during testing. This tends to penalize drives smaller than 180 GB and favor those with more than 256 GB of capacity. So, it's not wise to take a trace from a 240 GB SSD and wrap it around something, say, 40 GB-large. Using a small trace on a larger SSD is no problem, but going the other way results in different trace timing, affecting the results.

If you've followed us so far, you won't be surprised to see Crucial's M550s and the 1024 GB SP920 tripping over one another. Nor should the 128 and 256 GB model's outcome require much explanation. But the 512 GB SP920? That doesn't end up where we might have guessed.

First, we tried running our trace a second time. Then we tried other systems, all of which spat back the same result. Really though, losing 15 MB/s doesn't mean much at the end of the day. But we do come away knowing that the 512 GB SP920 isn't exactly the same as the 1024 GB model. Nor is it identical to the M550s. Something is slightly different on what should be an identical drive. Perhaps a look at mean service times on the next page will shed some light...

7. Results: Tom's Hardware Storage Bench v1.0, Continued

Service Times

Beyond the average data rate reported on the previous page, there's even more information we can collect from Tom's Hardware's Storage Bench. For instance, mean (average) service times show what responsiveness is like on an average I/O during the trace.

It would be difficult to graph the 10+ million I/Os that make up our test, so looking at the average time to service an I/O makes more sense. For a more nuanced idea of what's transpiring during the trace, we plot mean service times for reads against writes. That way, drives with better latency show up closer to the origin; lower numbers are better.

Write latency is simply the total time it takes an input or output operation to be issued by the host operating system, travel to the storage subsystem, commit to the storage device, and have the drive acknowledge the operation. Read latency is similar. The operating system asks the storage device for data stored in a certain location, the SSD reads that information, and then it's sent to the host. Modern computers are fast and SSDs are zippy, but there's still a significant amount of latency involved in a storage transaction.

When we get four models from the same product family split between a range of capacities, it's not uncommon for there to be a big delta between the smallest and largest drives.

For the most part, our tests show that read performance is similar on each drive, even though the 128 GB SP920 is just one-eighth the size of the 1024 GB model. Writes take a massive hit, though, since the smaller SSDs employ fewer dies to spread workloads across.

Mean Read Service Time

Adata's SP920s demonstrate similar read performance as the other SSDs equipped with IMFT's flash.

Finally, Adata's SP920s are vindicated. The 1024 GB model beats out the largest M550 for first place. No desktop-oriented SSD we've tested boasts such an excellent result. Yes, the margin of victory is tiny, but you can see just how compelling the competition is, as well.

And we're again presented by a mystery in the 512 GB SP920. It isn't at the top of the chart alongside Crucial's M550, which is itself identical to the 1024 GB model. The difference is notable enough for us to think twice about some of the assumptions we've made about the data generated thus far. We're not talking about a crazy deficit or anything, but the drive is marginally slower in our trace-based testing. Just bear in mind that you'd have a hard time telling a difference in a real-world workload. 

The 128 GB SP920 beats Samsung's 120 GB 840 EVO and Crucial's M500 at the same capacity point, which is a testament to Marvell's updated controller. Both competing SSDs use 128 Gb die as well.

8. Results: TRIM Testing With ULINK's DriveMaster 2012

We've been utilizing ULINK's DriveMaster 2012 software and hardware suite to introduce a new test for client drives. Using JEDEC's standardized 218A Master Trace, DriveMaster can turn a sequence of I/O (similar to our Tom's Hardware Storage Bench) into a TRIM test. JEDEC's trace is months and months of drive activity, day-to-day activities, and background operating system tasks.

ULINK strips out the read commands for this benchmark, leaving us with the write, flush, and TRIM commands to work with. Execute the same workload with TRIM support and without, and you end up with a killer metric for further characterizing drive behavior.

DriveMaster is used by most SSD manufacturers to create and perform specific measurements. It's currently the only commercial product that can create the scenarios needed to validate TCG Opal 2.0 security, though it's almost unlimited in potential applications. Much of the benefit tied to a solution like DriveMaster is its ability to diagnose bugs, ensure compatibility, and issue low-level commands. In short, it's very handy for the companies actually building SSDs. And if off-the-shelf scripts don't do it for you, make your own. There's a steep learning curve, but the C-like environment and command documentation gives you a fighting chance.

This product also gives us some new ways to explore performance. Testing the TRIM command is just the first example of how we'll be using ULINK's contribution to the Tom's Hardware benchmark suite.

On a 256 GB drive, each iteration writes close to 800 GB of data, so running the JEDEC TRIM test suite once on a 256 GB SSD generates almost 3.2 TB of mostly random writes (it's 75% random and 25% sequential). By the end of each run, over 37 million write commands are issued.

The first two tests employ DMA to access the storage, while the last two use Native Command Queuing. Since most folks don't use DMA with SSDs (aside from some legacy or industrial applications) we don't concern ourselves with those. It can take up to 96 hours to run one drive through all four runs, though faster drives can roughly cut the time in half. Because so much information is being written to an already-full SSD (the drive is filled before each test, and then close to 800 GB are written per iteration), SSDs that perform better under heavy load fare best. Without TRIM, on-the-fly garbage collection becomes a big contributor to high IOPS. With TRIM, 13% of space gets TRIM'ed, leaving more room for the controller to use for maintenance operations.

TRIM Testing

Rolling Average

Here's the chart derived from our DriveMaster JEDEC TRIM test data. We have the new Adata SSDs, Crucial's M550s, Samsung's venerable 840 Pro at 256 GB, Crucial's more mainstream M500 (240 GB), Plextor's M5P, and the 250 GB 840 EVO. Each device's NCQ-based test is plotted. The solid line represents average IOPS every 100,000 commands, but without TRIM. The hashed line represents performance every 100,000 commands with TRIM. In each case, the workload is mixed in with tons of small, random writes.

Since performance is measured over each 100,000-command segment, time is factored out of the above chart. This rolling average also hides the trace's peaky nature.

There's a lot going on in the chart above, but pay particular attention to the 512 GB Crucial M550 in green and SP920 in teal. The M550 again outpaces Adata in a small but quantifiable way through the test with TRIM enabled and without the command. We keep getting the sense that these drives are not as identical as the hardware suggests. 

Instantaneous

But I also want to know the instantaneous average of our TRIM testing. How does the drive fare servicing writes with and without TRIM during each 100,000-command window? The purple line represents IOPS across the entire trace, without TRIM. The teal line is with TRIM. Each data point represents write IOPS per 100,000-command test reporting period.

As we get more experience with this test, it's easier to identify the drives intended for enterprise-oriented environments and the ones destined to do desktop duty. The SP920 is readily identifiable as the latter. During periods of high I/O demand, the teal line shows the extra space created by TRIM allowing much higher performance. That is to say, when the system needs more write I/O, the SP920 delivers through its support for TRIM.

Throughput

We collect and report the total throughput of each drive in the NCQ with TRIM test. It's one number that helps capture overall performance in the test.

Going back to our rolling average chart, where the 512 GB M550 beat the SP920 by a slim margin, these are the results in MB/s. And now we can see that the difference is just 2 MB/s across the test. Samsung's 840 Pro 256 GB, the Crucial M550, and the theoretically-identical Adata SP920 are all very similar.

9. Results: Power Consumption

Active Idle Power Consumption

Idle consumption is the most important power metric for consumer and client SSDs. After all, solid-state drives complete host commands quickly and then drop back down to idle. Aside from the occasional background garbage collection, a modern SSD spends most of its life doing very little. Enterprise-oriented drives are more frequently used at full tilt, making their idle power numbers less relevant. But this just isn't the case on the desktop, where the demands of client and consumer computing leave most SSDs sitting on their hands for long stretches of time.

Active idle power numbers are critical, especially when it comes to their impact on mobile platforms. Idle means different things on different systems, though. Pretty much every drive we're testing is capable of one or more low-power states, up to and including DevSlp. That last feature is a part of the SATA 3.2 host specification. And while it requires a capable SSD and a compatible platform, enabling DevSlp takes power consumption down to a very small number.

As with performance, active idle is another discipline we'd expect Adata and Crucial to measure identically in. But the M550s manage to slip in under the SP920s. Again, it's a small, yet quantifiable difference.

There is one possible explanation. Take a gander at the 1 TB drives, highlighted above. Shouldn't the 1024 GB M550 and SP920 demonstrate the exact same behavior? We know active idle is largely a function of the controller itself, since the amount of NAND on-board doesn't move the dial much. As an example, when we spent time with Intel's SSD 730, its overclocked processor affected our active idle measurement substantially.

And we see the same thing here. Compare Adata's SP920 to the M550s and we see the Crucial drives demonstrating lower power use. Could the controller be to blame? We know these are identical ASICs on the surface, but I think it's possible that Micron is reserving higher-binned 9189s for its own purposes, hitting comparable clock rates using less voltage. We reached out to Marvell for comment on this and received no answer.

PCMark 7 Average Power Consumption

If we log power consumption through a workload, even a relatively heavy one, we see that average use is still pretty close to the idle numbers. Maximum power may spike fiercely, but the draw during a PCMark 7 run is light. You can see the drives fall back down to the idle "floor" between peaks of varying intensity.

The story reported previously repeats itself here. Physically, the Adata and Crucial drives appear identical. But there are subtle differences between them not apparent based on looks alone.

Charting power over time makes it easy to see the SP920s dropping back to the 1 W+ idle floor. Peak consumption is commensurate with drive capacity.

Maximum Observed Power Consumption

Maximum power use ends up close to where we'd expect it.

10. Adata SP920: Adding Value With A Nice Bundle

There are a couple of different directions I could go with this wrap-up. On one hand, it'd be easy to take a swipe at Adata for rebadging a Micron storage platform. But the fact of the matter is that I'd do the same thing in Adata's position. There are several companies needing a drive like the SP920 right now

Lots of companies need a drive like the SP920 that can tide them over until LSI's next-gen SandForce controller is ready. Adata had to look at its options and decide that it'd take too long to develop a home-grown Marvell-based platform. Going with a fast, modern, and, most important, available SSD like the M550 simply made the most sense. Adata gets a solid product able to soften the wait, and Micron (Crucial's parent company) gets to move more volume. 

Neither Micron nor Adata wants to volunteer details, and legally, they're probably prohibited from doing so. But we can fill in the gaps with educated guesses. Adata had to know that we'd figure all of this out relatively quickly, so the company gets my respect for making this decision, sticking with it, and doing what's best until the next wave of SSDs is ready. If imitation is the sincerest form of flattery, then Adata is making quite a statement about the M550. And hopefully, a variation on that theme turns into better prices on the same great performance, benefiting enthusiasts.

We know that these drives come off of Micron's production lines. See that "Product of Singapore" label attached to the SP920? Adata's production happens in Taiwan. Micron's SSDs, on the other hand, come from Singapore. This suggests that, despite its considerable manufacturing capacity, Adata isn't cranking out Micron-designed drives in its own factories, but rather doing a bit of branding. And thus, my first indication that the SP920 isn't Adata's own (the chassis, from page one), is all but confirmed by a silly little sticker. 

Now, this isn't to say that the SP920 is exactly a Crucial M550. Despite the similar boards, controllers, and firmware, there are some slight behavioral differences on the test bench. It's possible that Micron did not want to outfit the lower-capacity SP920s with the M550's equivalent flash. We know that Crucial's 128 and 256 GB models rock twice as many dies as Adata's versions with 128 Gb densities. Once you get up to the 1 TB size, though, both competing repositories match each other almost exactly.

There's one thing Adata doesn't get for the SP920 that Micron keeps for itself, or that Adata didn't want: TCG Opal 2.0 and Microsoft eDrive support, two awesome encryption features.

The important question to ask, then, is why would you want a 1024 GB SP920 rather than a 1024 GB M550? Should you buy one company's 512 GB model over the other's? Pricing should be equal. Three-year warranties cover all models. And neither brand truly outperforms the other. Adata does ship its Premier Pro SP920s with a 3.5" adapter sled and a 7-to-9 mm spacer.  It's handy, but not (pardon the pun) crucial. The SP920s also come with Adata's own toolbox utility. Firmware updates, diagnostics, TRIM, and secure erasures are all enabled through it. I like the software, and think all SSD vendors should offer something like it. Crucial does not bundle a suite of tools with its own drives. But the best value-add included with the SP920 is an OEM version of Acronis True Image HD 2013, which lets you clone an existing operating system. Unlike proprietary versions of imaging software, this one works with any drive, and there are other features built in to True Image that just cloning.

In the end, picking Adata's SP920 over the M550 makes sense when bundled extras make a difference to the value equation, providing prices do end up relatively similar. My recommendations change based on the specific model, though. The 128 and 256 GB SP920s should sell for about $10 less than the M550's suggested prices. We don't have a 128 or 256 GB M550s, so we can't recommend them based on our own experiences. But you'll probably find that they're faster than the SP920 at those capacities, since we know they're using twice the die. The 512 and 1024 GB SP920s are easier for me to stand behind. With functionally-identical performance, value is the biggest variable. If Adata's extras are enough to tip the scale for you, both capacities are solid options.

Of course, given low-ball prices on Crucial's M500, buying one of those and pocketing the difference isn't a bad way to go. Then again, the same holds true for almost every other SSD out there with a history of aggressive price cuts and almost-as-good performance.

Adata isn't the only vendor suffering in the absence of third-gen SandForce hardware. In truth, I'm a little surprised that more companies haven't diversified their storage portfolios to be less dependent on one supplier. Silicon Motion's SM2246EN controller is a solution I would have expected to start showing up in lower-capacity drives by now. Unlike Marvell, Silicon Motion will sell you a more complete SSD package. Or, if you want its controller for a completely original design, you can do that too.

Whatever circumstances led Adata to today's launch, the company made the right move to keep its name on our tongues through 2014. I'm naturally waiting with bated breath for LSI's response, but that doesn't make products based on Marvell's silicon any less compelling. We already know that the SP920 is an evolution of hardware wringing every last bit of performance from the SATA 6Gb/s interface. And there's nothing wrong with that.