by Ilya Gavrichenkov
03/14/2012 | 12:12 PM
The SSD market doesn’t live only off SandForce based drives. Although there are at least a few dozen modifications of consumer grade SSDs on SF-2281 controller, there are also completely different offerings already available in stores, even though they are certainly not as diverse. SandForce implemented a very convenient procedure for adoption of their platform: SSD makers receive not only their controllers, but also a reference PCB layout and a basic firmware. Other SSD controller makers do not offer support like this, which definitely affects their presence in the today’s market.
Nevertheless, second-generation SandForce solutions have a pretty serious competitor, which managed to win its place under the sun first of all due to its exceptional characteristics. It is Marvell 88SS9174 controller, which is undoubtedly the second most popular controller these days, even though it is less widely spread as SF-2281.
Historically, Marvell 88SS9174 came out significantly earlier than second-generation SandForce and for a while it remained the only SATA 6 Gbps controller for consumer SSDs. This controller, just like all its existing competitors, has eight flash-memory channels, but at the same time is based on a dual-core processor with ARM architecture, which allows to flexibly vary its practical characteristics by adding some changes to the firmware. Together with constantly released new hardware modifications Marvell 88SS9174 has remained up-to-date for the last 2 years. However, it is much more difficult to manufacture SSDs on these chips from the engineering prospective: the developers should come up with their own firmware and hardware layout. Nevertheless, many distinguished SSD makers, including Corsair, Intel, Micron and Plextor, prefer Marvell controllers.
In our today’s roundup we are going to discuss a few contemporary SSD models built on Marvell 88SS9174 controller. We will also try to determine whether they should be considered as a serious alternative to SandForce based SSDs. For our today’s test session we invited three solid state drives with 256 GB storage capacity: Corsair Performance Pro, Crucial m4 and Plextor M3S.
Corsair has long had a full lineup of SSDs based on SandForce controllers. However, not everyone is happy with the reputation of these controllers and therefore the drives that is why in the end of last year Corsair decided to offer an alternative to those, who weren’t certain about going with a SandForce based product – a new series of high-performance products based on a Marvell controller. The new Performance Pro series became a logical continuation of the already out-dated Performance 3 family, which is no longer a true “performance” series even in formal specifications, although it uses almost the same hardware components.
The new Performance Pro lineup looks very promising even in terms of formal specifications offering speeds comparable with those of the fastest second-generation SandForce based products. Our tests will show how true that is, and in the meanwhile let’s take a closer look at the actual 256 GB Corsair SSD based on Marvell controller.
First of all I would like to stress that the looks of Corsair Performance Pro make it stand out among other products from this manufacturer. While SandForce based drives are designed in black or red cases, the alternative SSD uses an unpainted aluminum case with rough machine marks.
Inside the case we find a PCB. There is a Marvell 88SS9174-BKK2 controller on one side – a second revision of an original Marvell chip with improved performance. I would like to point out that inside an assembled SSD the controller is pressed against the case using a layer of thermal interface, which means that Corsair requires improved heat dissipation for their drive.
The reverse side of the PCB is allocated completely for the memory chips. There are eight Toshiba flash chips marked as TH58TVG8D2FBA89. They are referred to as “Supercharged Synchronous NAND” on the official Corsair web-site, but it is a purely marketing term. In reality we are dealing with MLC chips with synchronous Toggle-Mode 1.0 interface and 32 GB capacity, each containing eight 32 Gbit NAND devices. Although this memory is manufactured with 32 nm process, it is one of the fastest modifications of MLC flash available today used in mainstream SSDs. They provide super high speed due to the fact that the eight-channel Marvell 88SS9174 controller accesses memory using eight-way interleaving instead of four-way one like in other 256 GB SSD models.
In addition to flash memory, Corsair Performance Pro also has a DDR3-1333 SDRAM based cache. In this case it is implemented using two NANYA NT5CB128M16BP-CG chips with a total capacity of 512 MB.
Although they use Toggle Mode NAND, which we normally see in more expensive SandForce based SSDs (such as OCZ Vertex 3 Max IOPs), Corsair Performance Pro are sold at a relatively affordable price. Their price per gigabyte is anywhere from $1.4 to $1.6 depending on the total size of a specific model: a 128 GB or a 256 GB one.
Crucial m4 solid state drive can be regarded as a veteran of this market. It was launched almost a year ago and in reality is an evolutionary upgrade of another popular product – Crucial RealSSD C300. Moreover, the primary message of the history behind these SSDs indicates clearly that SSD firmware for Marvell 88SS9174 based products is of great importance. Regular firmware updates allow maintaining the performance of Crucial m4 at a very competitive level. During its long-term market success, Crucial m4 firmware has received two very important updates. One of them eliminated critical errors, while another one significantly increased the read speed by anywhere between 12-20%.
However, the hardware configuration of this solid state drive also contributes substantially to its long-lasting career. It has been originally designed with intention to remain up-to-date for a long period of time.
Crucial m4 is built with a PCB carrying two types of memory chips and a Marvell 88SS9174-BLD2 controller – the third revision of this chip (from the day it was born).
The first type of memory is MLC NAND flash. For our tests we used a 256 GB drive equipped with sixteen 16 GB Micron flash chips, which is not surprising at all, since this manufacturer owns the Crucial brand. Flash memory chips are marked as 29F128G08CFAAB, which indicates that we are dealing with synchronous ONFI 2.2 interface based on two 64 Gbit semiconductor dies manufactured with 25 nm process.
The second memory type is represented by a single Micron D9LGQ chip, which is none other but a regular DDR3-1333 SDRAM chip with 256 MB capacity. This memory serves as a cache, which is an inalienable part of all solid state drives on Marvell controllers. Although Crucial’s cache is a little smaller in size, than the same cache in other products participating in our today’s roundup.
In conclusion I would like to add that Crucial m4 SSDs are currently priced very affordably: at about $1.3-$1.4 per gigabyte of storage. The stores currently carry 64, 128, 256 and 512 GB models.
Just like Crucial, Plextor uses exclusively Marvell controllers. As you may have already guessed, Plextor M3S is already the third SSD generation based on 88SS9174 controller. The previous M2S model is in fact very similar to M3S in terms of the employed hardware components, but has significantly weaker parameters because of the different firmware version.
Just like Corsair, Plextor uses Toshiba memory for their contemporary SSDs, and the new M3S family is built with the newest NAND flash manufactured with 24 nm process, which production started only in the end of last summer.
So, even though all three SSDs reviewed today are based on the same Marvell controller, they are all very unique. I have to stress that uniqueness is a typical trait of all Marvell based solid state drives, which stand out in a very positive way against the background of SandForce based clones. However, the unique hardware inside the Plextor M3S doesn’t really affect its exterior. In fact, it looks like the external casing of the Plextor M3S comes from the same exact factory as the casing for Corsair Performance Pro. The only difference is the anodized coating and a painted Plextor logo. The mechanics of the Corsair and Plextor SSD cases is completely identical.
However, what we find inside the Plextor drive is even more shocking. Corsair Performance Pro and Plextor M3S do not just look the same, but also use a PCB with the same exact layout. Moreover, once we compared with PCB from these two drives with one another we discovered that the PCB inside Corsair SSD is marked as M3S. Therefore, even though Plextor is not a legal maker of the Performance Pro, it did contribute a lot to its development.
However, Plextor M3S uses completely different type of flash memory than Corsair Performance Pro. M3S drives are built with synchronous Toshiba memory with Toggle Mode interface, which chips are designed using 64 Gbit NAND manufactured with the latest 24 nm process. Thr 256 GB Plextor M3S has eight TH58TEG7D2HBA4C chips with four semiconductor flash-memory dies in each. So, Plextor M3S uses 4-way NAND interleaving in each channel.
Other than that, everything in Plextor M3S is pretty predictable. It uses Marvell 88SS9174-BKK2 controller and has 512 MB of cache based on two NANYA NT5CB128M16BP-CG DDR3-1333 SDRAM chips. Note that only 256 and 512 GB models have a 512 MB cache. M3S models with lower storage capacity have a cache of proportionally reduced size.
By using flash memory manufactured with fine technological process Plextor managed to lower the price of the drives. As a result, the M3S lineup including 64, 128, 256 and 512 GB models comes at $1.2-$1.3 per gigabyte of space, which is lower than other SSD models using Marvell controllers. At the same time Plextor SSDs come with extended warranty, while most of the other makers offer only 3 years.
Three solid state drives participating in our today’s review - Corsair Performance Pro, Crucial m4 and Plextor M3S – are all based on the same Marvell 88SS9174 controller, but the controller revisions are different, and so are the flash memory and the firmware versions. Of course, their specifications are also quite different because of that. The table below compares the formal characteristics of the three 256 GB models, which took part in our today’s test session. For the sake of comparison we also included OCZ Vertex 3 – one of the most popular SSDs based on the second-generation SandForce controller.
Overall, the formal specifications of SandForce based SSD look better than those of Marvell based ones. However, do not forget that the major peculiarity of SandForce is its ability to compress data “on the fly”, which makes SSDs work much faster with well-compressible data. As a rule, the manufacturers list their official SSD specs with an ideal case of maximum data compression in mind, which means that the actual real-life comparison between SandForce and Marvell based SSDs may be completely different from what the table shows.
However, let’s wrap up our theoretical speculations and proceed to the actual tests.
All SSDs were performed in a testbed built around an Intel H67 based mainboard. This chipset provides support for two SATA 6 Gbit/s ports, which we use to connect the tested SSDs.
We are going to compare the performance of Corsair Performance Pro, Crucial m4 and Plextor M3S against four alternative SSDs, which represent four existing types of SandForce-based products: OCZ Vertex 3 Max IOPS 240 GB (32 nm Toggle NAND), Corsair Force Series GT 240 GB (25 nm synchronous memory), Corsair Force Series 3 240 GB (asynchronous memory), and the new Intel SSD 520 240 GB (25 nm synchronous memory and exclusive firmware).
Overall our testbed was configured as follows:
We use CrystalDiskMark 3.0.1 benchmark to test the random- and sequential read and write speed. This benchmark is convenient to work with as it can measure the speed of an SSD with both incompressible random and fully compressible recurring data. This feature is important for testing SSDs based on SF-2281 controller, which tries to compress the data before writing it into memory. So, there are two numbers in the diagrams below that reflect the maximum and minimum SSD speed. The real-life performance of an SSD is going to be in-between those two numbers depending on how effective the controller data compression is.
Note that the performance tests in this section refer to SSDs in their “Fresh Out-of-Box” state (FOB). No degradation could have taken place yet.
Marvell 88SS9174 controller doesn’t compress data during writing that is why the SSDs based on it demonstrate predictably similar results with compressible and non-compressible data. As a result, it is very difficult to compete against SandForce based SSDs in writing compressible data.
Nevertheless, SSDs with Marvell controllers inside also have their advantages. Namely, these SSDs do really well during random operations on small data blocks in case of a deep requests queue. It serves as a reminder that Marvell 88SS9174 comes from a server background, which is partially true, because Marvell is still the leading developer of server SSD controllers.
At the same time, Corsair Performance Pro and Plextor M3S also demonstrate very good results during sequential writing. Of course, if the data is well-compressible, then they can’t really compete against SandForce-based drives, but in case of “SF-2281 unfriendly” information structure, Marvell controller is capable of delivering higher results.
As for the relative performance of Corsair Performance Pro, Crucial m4 and Plextor M3S based on the same Marvell 88SS9174 controller, we can clearly see that it differs dramatically. Of course, different controller revisions, different firmware versions and different memory have a lot to do with that. And while Crucial m4 is the slowest Marvell-based SSD, which can only compete successfully against the low-cost SandForce drives using asynchronous memory , then Corsair Performance Pro and Plextor M3S built with Toggle Mode NAND look much more attractive. Moreover, Plextor M3S works much better with random operations, while Corsair Performance Pro employing much more aggressive interleaving technology leads in sequential writing.
Unfortunately, SSDs are not always as fast as in their “fresh” state. In most cases their performance goes down after some time and in real life we deal with completely different write speeds than what we see on the diagrams in the previous chapter of our review. The reason for this phenomenon is the following: as the SSD runs out of free pages in the flash memory, its controller has to clear memory page blocks before saving data into them, which causes substantial delays. Although, modern SSD controllers can alleviate the performance drop by erasing unused flash memory pages ahead of time, when idle. They use two techniques for that: idle-time garbage collection and TRIM.
Of course, users are more interested in the consistent performance of their SSDs over a long period of time rather than the peak speed they are going to see only during the initial short-term usage period, while the drive is still “fresh”. The SSD makers, however, declare the speed characteristics of “fresh” SSDs for marketing reasons. That’s why we decided to test the performance hit that occurs when a “fresh” SSD becomes a “steady” one.
To get a complete picture of SSD performance degradation we ran special tests based on the SNIA SSSI TWG PTS (Solid State Storage Performance Test Specification) methodology. The main idea of this approach is to measure write speed consecutively in four different cases. First we measure the “fresh” SSD speed. Then we measure the speed after the SSD has been fully filled with data twice. The third test occurs after a 30-minute break during which the controller can partially restore performance by running the idle-time garbage collection. And finally, we measure the speed after issuing a TRIM command.
We ran the tests in synthetic IOMeter 1.1.0 RC1 benchmark, where we measured random write speed when working with 4 KB data blocks aligned to flash memory pages at 32 requests queue depth. The test data were pseudo-random.
This is where Marvell 88SS9174 based SSDs become the stars of this show. They not only ace the TRIM execution by fully restoring their performance to the original level, but can also boast excellent efficiency of the background garbage collection algorithms, which prove capable of restoring the data-packed SSD practically to its new state. SandForce controllers, as we have already seen many times in our previous tests, don’t get anywhere close to that.
As for the three devices that are the main characters of our today’s review, Corsair Performance Pro stands out the most. Its background garbage collection is practically as efficient as TRIM execution. Plextor M3S is also good in this respect, although it doesn’t restore its performance fully to the original level. And as for Crucial m4, its garbage collection works relatively slowly compared to the other two SSDs.
However, background garbage collection kicking in when TRIM is not working cuts both ways. On the one hand, drives that are capable of restoring their parameters without TRIM are great for those applications where TRIM simply doesn’t work at all. Namely, they are strongly recommended for RAID arrays or operating systems that don’t support TRIM command, such as Windows XP, FreeBSD with ZFS or Mac OS X (although TRIM may be activated in the latest Lion versions by modifying the core extension). But on the other hand it is important to understand that background garbage collection working without TRIM causes continuous data shuffling in the flash memory, which inevitably eats up its resource.
Unfortunately, Marvell 88SS9174 controller offers very scarce S.M.A.R.T. parameters, which prevents us from estimating the Write Amplification typical of the SSDs based on it. But without doubt this Write Amplification is substantially higher than by SandForce based SSDs. In other words, Marvell 88SS9174 based solid state drives may exhaust the flash memory resource much faster than the alternative products. Therefore, it is really good that Corsair Performance Pro and Plextor M3S SSDs use relatively long-lasting memory with 5000 rewrites reserve. The third SSD, Crucial m4, uses more complicated flash-memory with only 3000 rewrites, and the garbage collection isn’t too aggressive over there, too.
Going back to the performance numbers we have to point out that in steady state the SSDs with Marvell and SandForce controllers won’t rank in the same manner as they did in the previous section, even if the operating system supports TRIM. One SandForce based drives have been filled with data completely at least once, their performance irreparably degrades, while SSDs with Marvell controllers inside manage to avoid this fate. As a result, the solid state drives that have already been in use and have reached their steady state demonstrate different write speeds (the diagrams show the speeds in CrystalDiskMark 3.0.1 with random incompressible data).
Although SSDs on Marvell controllers are in a winning situation here due to the far from ideal implementation of the TRIM execution in SandForce based drives, the overall picture doesn’t really change. Marvell controller once again puts the products based on it in leading positions during sequential writing and random writing with deep requests queue. However, SandForce based drives remain unbeatable during regular normal writing of small random data blocks.
PCMark 7 contains an individual disk subsystem benchmark. It is not a synthetic test, but is based on real-life applications. This benchmark reproduces typical disk usage scenarios and measures how fast they are completed in popular applications. Moreover, the disk access commands are not executed as a steady uninterrupted flow, but in a more realistic manner – with certain pauses caused by the need to process the data. The benchmark generates an overall disk subsystem performance rating as well as speed readings in MB/s in individual usage scenarios. Note that the absolute speed in these scenarios is not too high because of the above mentioned pauses between individual input/output operations. In other words, PCMark 7 shows you the speed of the disk subsystem from the application’s point of view. Numbers like that show us not only the pure performance of an SSD, but mostly how big of a performance gain a certain SSD can guarantee in real life.
We ran PCMark 7 on “steady” SSDs, which is what they are going to be in actual computer systems most of the time. Their performance in this case is affected not only by their controller or flash memory speed but also by the efficiency of their internal algorithms that fight performance degradation.
The PCMark 7 score is a good guide for users who want to see how SSDs perform in typical applications without delving into technical details. According to this benchmark the fastest SSD on Marvell 88SS9174 controller is Corsair Performance Pro. It manages to almost fully catch up with the best SandForce SF-2281 based SSDs built with synchronous memory. A slower drive, Plextor M3S appears to fall somewhere in the middle between SandForce based products with synchronous and asynchronous flash memory. And Crucial m4 is the last but one of the testing participants being just a little bit faster than Corsair Force 3 built with slow and inexpensive asynchronous memory.
Let’s check out the individual tests to get a more detailed picture of what our SSDs are capable of under various types of operational load:
We see almost the same overall ranking practically in all test scenarios as we have just seen in the total score diagram. The only exception is the “Starting applications” test where SSD with Marvell 88SS9174 controllers do a little better than everyone else.
Intel NASPT is another disk sub-system test that uses real-life usage scenarios. Like PCMark 7, Intel NASPT reproduces predefined disk activity traces and measures how fast they are executed. However, the default traces are designed for network attached storage devices rather than for SSDs. Therefore during our test session we replace them with the specially developed SSD Benchmarking Suite which offers more relevant usage scenarios such as compressing and decompressing files, compiling large projects, copying files and folders, loading 3D game levels, installing software, batch-processing photos, searching a digital library for data, mass-launching applications, and transcoding video.
Like PCMark 7, this benchmark gives us a true-to-life illustration of disk subsystem performance. Here the SSDs are again tested in their “steady” state.
Although INASPT and PCMark 7 are of similar nature and basically measure the time it takes to process a number of preset scenarios, they produce very different results. In our opinion, INASPT results are of greater significance, because this test and the usage scenarios in it are adjusted specifically for the solid state drives.
I have to say that INASPT considers Marvell 88SS9174 controller a higher performing solution than the Futuremark benchmark. Corsair Performance Pro wins the first place here, while Plextor M3S runs as fast as OCZ Vertex 3 Max IOPS. Crucial m4 also looks very good in this test demonstrating slightly higher results than Corsair Force GT. The secret behind this test running better on Marvell 88SS9174 based SSDs is in the fact that INASPT emulates the least favorable situation for SandForce controllers by working with files containing random incompressible data.
The detailed INASPT results help us see what usage scenarios are the most suitable for our today’s testing participants. Take note that the data-transfer rate is higher than the SATA III interface bandwidth in some subtests. That’s because INASPT is a high-level test that uses standard Windows functions to access the disk subsystem. As a result, the OS caching mechanisms also affect the results.
It is pretty interesting that the total score winner, Corsair Performance Pro, is the fastest only in two tests out of twelve. In other words, its leadership is most likely determined by the fact that in any type of usage scenario its performance is at a decent level without any serious drops. So, it turns out to be an excellent option for general-purpose use, which will fit perfectly into any desktop system.
Overall, SSDs with Marvell 88SS9174 controller are particularly fast in applications using a lot of streaming writes: saving large files, installing programs, processing video.
It is really hard to find any problems with Marvell 88SS9174 controller, which is used in the solid state drives reviewed today. Its characteristics and internal architecture allow designing SSDs with up-to-date performance, and such SSD makers as Corsair, Crucial and Plextor can’t help taking advantage of it. However, it is not as easy to compare Marvell based SSDs against competing products on SandForce controller as it might seem. I think it would make more sense to say that these products are coexisting in parallel planes.
SandForce’s strength is the ability to compress data before it gets into the flash memory. This peculiarity helps solve several problems at once: first of all, increase the performance and prolong the drive’s resource. However, it only works well until the data stored on the drive are compressible. Otherwise, the advantages of SandForce controllers become not so obvious anymore.
However, SSDs based on Marvell 88SS9174 do not depend on the type of data saved onto the drive. They demonstrate constant stable performance with compressible files as well as with incompressible photographs, videos and audio content. Moreover, Marvell 88SS9174 is capable of fighting the SSD performance degradation very effectively, so that it improves almost to its “fresh” level not only after the TRIM command, but even without it by simply performing idle garbage collection.
Therefore, solid state drives with a Marvell controller inside are a much better fit for RAID arrays or systems that do not support TRIM. As for all other cases, the choice between Marvell or SandForce based SSD should depend on the preferred usage model. SandForce devices will work faster in everyday tasks dealing with well-compressible data, while Marvell devices will deliver higher performance with incompressible data and in case of dominating random requests at a deep queue, which is more typical of servers.
If we were to choose between the three Marvell 88SS9174 based SSDs reviewed today, our favorite would be Corsair Performance Pro. It is the fastest product in most situations and its price is not unreasonable. Therefore, we decided to award this product with our Recommended Buy title. Even if you prefer SandForce platform, we strongly recommend considering Corsair Performance Pro as a possible option.