by Aleksey Meyev
01/04/2011 | 11:12 AM
As opposed to the slowly developing HDD market (our recent roundup covered the entire year), solid state drives are evolving at a much more rapid rate and bringing us more news in a shorter period. There are SSD products with new controllers, with PCI Express interface, with new firmware, etc. We, at X-bit labs, have to be running fast just to stay where we are and keep you in touch with the latest market developments.
So in this review we are going to take a look at an SSD designed as a PCIe controller and check out if it is better than a RAID array built out of two ordinary 2.5-inch SSDs. We will also benchmark low-capacity SSDs based on the SandForce controller.
Based on the popular SandForce controller, this product is already known to us. We include it into this review to check out how good today’s SSDs are in high-speed RAID0 arrays. As you will see shortly, there will be worthy opponents for such a RAID. So, we take two samples of this SSD with firmware version 30CA13F0 and unite them into a RAID0 array using the mainboard’s chipset. The test results of one such SSD will also be published for the sake of comparison.
Crucial has managed to release a product that unnerves the established leaders of this market. The RealSSD C300 is based on an exclusive controller. Inside it, there is an ARM processor with Micron’s firmware which supports SATA 3 mode, i.e. a peak data-transfer rate of 600 MBps. Yes, this is the first serially produced SSD to utilize the new interface. It is specified to have a peak read speed of 355 MBps, though. The ARM processor is accompanied by a 256MB cache memory chip. As becomes a modern SSD, this product supports TRIM.
Our sample’s firmware is version 0002. We will test it in two modes: connected to the chipset (ICH7) in SATA 300 mode and connected to a Marvell 9123 controller in SATA 3 mode. We don’t like the Marvell controller’s drivers, but we don’t have any alternative to it until SATA 3 is implemented in mainboards’ chipsets. We are looking forward to Intel’s introducing the new standard in its platform.
This is one more representative of the large number of SSDs based on the SandForce controller that have appeared over the last half-year. Why are we interested in it? We just want to check out whether low-capacity products based on that highly popular controller are slower than their large-capacity cousins.
Besides the storage capacity, this model is no different from others with the same controller. Our sample reports to diagnostic utilities that its firmware is version 2.1.
This SSD is well known to us, but we want to include it into this test session as a same-category opponent to the above-described low-capacity G.Skill Phoenix Pro. Its firmware version is 2CV102HD.
OCZ is known for its extensive range of SSD products. The company seems to have made a point of releasing a new model on a monthly basis. Following the sensational Vertex 2, we now welcome the Agility 2 which is based on the same SandForce controller and offers the same storage capacities. As far as we know, these two models differ in their flash memory chips only. We will be able to compare them now as the Vertex 2 is also included into this review.
Our Agility 2 has firmware version 1.11.
We have to reference our earlier review again. We once tested a Colossus LT and now we’ve got its full-featured rather than lite version. As opposed to the rest of the SSDs, the Colossus series comes in the 3.5-inch form-factor and incorporates several Indilinx Barefoot controllers (two in our sample and four in larger-capacity models) combined into a RAID0 array. The latter fact prevents this SSD from supporting TRIM.
Our sample’s firmware is version 1.000.
Here is one more SSD which is based on a pair of controllers but the key feature of the RevoDrive is its form-factor. It is designed as an expansion card with PCI Express x4 interface. It’s easy to understand the replacement of SATA with PCI Express: the four PCIe lanes ensure a peak data-transfer rate of 1000 MBps into either direction, which is far above the capabilities even of the newest SATA interface version. This huge bandwidth is given to two SandForce controllers combined into a RAID0 array on a Silicon Image 3124 chip. By the way, this RAID controller has four channels and there is a seat for a daughter card on the RevoDrive. Having two more SATA channels and two SandForce controllers, the daughter card can be used to transform the RevoDrive duo into a RevoDrive x2 quartet.
There is one thing we don’t like about this SSD’s design. The interface of the Silicon Image 3124 is PCI-X rather than PCI Express. We can make guesses as to why this controller was chosen (perhaps it could be obtained from suppliers immediately and in large quantities or it might just have a highly appealing price), but the fact is that the SSD developer had to install an additional Pericom chip as a bridge between the PCI Express and PCI-X interfaces.
This SSD design cannot support TRIM but promises unprecedented speeds: up to 540 MBps at reading and up to 400 MBps at writing. We like the way this SSD is identified by the OS: as a Silicon Image 3124 controller with two disks connected to it. Thus, you can access the RAID controller’s BIOS if you need to. There are no special problems with booting up from this SSD. You can install your OS on it just like on any other RAID controller.
The last runner today is called Vertex 2. We’ll use it for comparing to the others (although we don’t think the race of SSDs based on identical controllers is going to be very exciting) as well as for checking out the new firmware. During our first tests of this SSD we used firmware version 1.0, but now it is 1.1. We are most interested to see what changes the new firmware brings about.
The following testing utilities were used:
The SSDs are tested with the generic OS drivers. We format them in NTFS as one partition with the default cluster size (for FC-Test we create 32GB partitions), connect them to a mainboard port and enable AHCI. We have transitioned to a new method of benchmarking storage devices.
IOMeter is sending a stream of read and write requests with a request queue depth of 4. The size of the requested data block is changed each minute, so that we could see the dependence of the drive’s sequential read/write speed on the size of the data block. This test is indicative of the maximum speed the drive can achieve.
The numeric data can be viewed in tables. We will be discussing graphs and diagrams.
Now let’s see what all these lines have to tell us (by the way, note how high the speeds are – hundreds of megabytes per second!). The two pairs of SandForce controllers are competing to be the leader in terms of sequential read speed: one pair is united using the chipset and the other resides on the RevoDrive card. Both pairs deliver a max speed of about 400 MBps. The dual-SSD version is somewhat faster with small data blocks but the RevoDrive behaves more consistently, without sudden changes in speed. The Crucial C300 with SATA600 is trying hard to catch up with the leaders but has to content itself with third place, showing a terrible performance hit on 64KB data blocks. Judging by the graph of the same SSD connected the chipset’s SATA300 controller, this slump is due to the Marvell controller’s problems. The Crucial is actually the best among the classic-design SSDs. The single-chip SandForce-based products are all inferior to it.
The OCZ Colossus is somewhat ambiguous. It offers a higher top speed than the SandForce-based models when reading large data blocks, yet its pair of Indilinx chips falls behind the single SandForce controllers at small data blocks.
The 40GB G.Skill is slower than its larger-capacity cousins but faster than the Intel X25-V.
We’ve got completely different standings in the sequential writing test. The Crucial shows its best, competing with the RevoDrive for top position when connected via SATA300. The same Crucial performs worse on the Marvell controller but that’s the controller’s, not the SSD’s, fault. The OCZ Colossus feels good here whereas the RAID0 pair of Corsair Force SSDs connected to the ICH controller loses its ground and is similar to the single-chip SandForce-based products in speed.
Another unpleasant surprise is that the Agility 2 and the Vertex 2 with firmware 1.1 cannot deliver even 100 MBps. The 40GB G.Skill is even slower, although ahead of the Intel X25-V which performs in the same category.
We can also note the sudden growth of speed of the more advanced models when switching from 2KB to 4KB data blocks.
For 10 minutes IOMeter is sending a stream of requests to read and write 512-byte data blocks with a request queue of 1. The total of requests processed by each SSD is much larger than its cache, so we get a sustained response time that doesn’t depend on the SSD’s buffer size.
There is nothing to talk about at reading. Every SSD is very good and the difference between them is negligibly small. Yes, the best model is twice as fast as the worst, yet the numbers are all so low that such things like a luckier distribution of load among flash memory banks may be a factor. The RevoDrive may have got its first place thanks to its low-latency interface, but the pursuers are very, very close to it.
Dual-controller solutions feel best at writing as they can alternate write requests between what is actually two SSDs if the requests are arriving in a proper order. The Intel X25-V joins the leaders without having such an advantage, though.
The Crucial C300 is obviously slower than the others when writing 512-byte random-address data blocks irrespective of the interface we connect it with.
Now we will see how the performance of the drives in random read and write modes depends on the size of the requested data block.
Most of the SSDs behave in the same way at random reading. We can only name four products with somewhat lower speeds: the G.Skill, the Vertex 2 with firmware 1.1, the Agility 2 and the pair of Corsair Force drives. The Intel X25-V and the Colossus with two Indilinx chips are ahead when reading very small data blocks but slow down when reading medium-sized ones.
The new testing method offers detailed results for small data blocks, so we can see a tremendous performance leap as the SSDs switch to 4KB chunks of data (the Intel X25-V does not count in because, for technical reasons, its results are approximated from those we achieved with our older method). Why does it happen? Because it is in 4KB blocks that data are written into flash memory chips. Interestingly, we can see this peak not with every SSD but only with the Corsair (both the single SSD and the duo), the OCZ RevoDrive, and the Crucial C300 (with both interfaces).
Another interesting fact is that the chipset-based Corsair array is excellent at writing small data blocks whereas the Crucial C300 is poor then. The latter shows its best when writing random-address data blocks of medium and large sizes and is followed by the RevoDrive.
In the Database pattern the drive is processing a stream of requests to read and write 8KB random-address data blocks. The ratio of read to write requests is changing from 0% to 100% with a step of 10% throughout the test while the request queue depth varies from 1 to 256.
You can click these links to view the tabled results:
We will build diagrams for request queue depths of 1, 16 and 256.
There is a couple of SSDs that deliver outstanding results at a request queue depth of 1. One of them is the highly consistent OCZ RevoDrive which copes neatly and confidently with every load. The other is the Crucial C300 whose behavior betrays very advanced deferred writing algorithms. It also takes first place at pure reading. However, The Crucial C300 only shows such results when connected via SATA300. Its performance is far from outstanding on the Marvell controller.
Among the remaining SSDs, the RAID0 array and the Colossus, being dual-controller solutions, have higher results. The Vertex 2 with new firmware 1.1 is fast, too.
As the request queue gets longer, the leaders make their preference clear: the RevoDrive wins at reading while the Crucial C300 breaks all records at writing. It is a mystery to us how the C300 can accomplish such wonders, but its performance – tens of thousands of write operations per second! – is truly impressive. Among the other products we can single out the RAID0 array built out of two Corsair F120 drives. Alas, the scalability of the array is not 100%.
The Intel X25-V, G.Skill and Colossus are the losers in this test. The 40GB SandForce-based model doesn’t slow down much, though.
Winding up this part of our tests, we will build diagrams showing the performance of each SSD at five different request queue depths.
This pair of diagrams makes it clear what benefits can be achieved by changing a single SSD with a RAID array. The write speed improves considerably, which also has a positive effect on the performance under mixed load. And if you can load your disk subsystem so high that there occurs a requests queue, the reading performance will grow up with the RAID array as well.
If the first diagram were not annotated, one might thought it to refer to some SCSI or SAS drive with all those characteristically shaped graphs indicative of read request reordering and aggressive deferred writing. It seems a small miracle that Crucial with Micron managed to achieve that from flash memory.
The second diagram suggests that we must give up using Marvell controllers. SSDs behave awfully when connected via them.
These SandForce-based devices behave in the same way, even the two Vertex 2 drives with different firmware.
The Colossus produces characteristically shaped graphs, too. This time around, the graphs are typical of Indilinx controllers.
As for the Intel X25-V, its behavior is highly specific, too. It features an excellent performance at reading (especially for this, low-capacity model) but has only mediocre results at writing.
The RevoDrive is similar to the SandForce-based products in this test but copes with mixed loads better. It slows down when there are only write requests to be processed.
The drives are tested under loads typical of servers. The names of the patterns are self-explanatory. The results are presented as performance ratings which are calculated as the average speed of the drive at every load.
If there are only read requests to process, you may want to consider the RevoDrive in the first place. It is just excellent under such a load. The Corsair Force-based RAID and the Crucial are fast, too, but you should not connect the latter to the Marvell controller which performs awfully at high queue depths.
The low-capacity models based on the SandForce controller and the OCZ Colossus are the worst performers in this test.
The addition of a small share of write operations into the test pattern doesn’t change the standings.
The multithreaded tests simulate a situation when there are one to four clients accessing the disk at the same time – the clients’ address zones do not overlap. We will discuss diagrams for a request queue of 1 as the most illustrative ones. When the queue is 2 or more requests long, the speed doesn’t depend much on the number of simultaneously running applications. You can also click the following links for the full results:
Our multithreaded tests often prove to be an extremely difficult trial for hard disk drives, but SSDs find them rather easy. They don’t care what exactly memory cell to access. Moreover, the multiple threads produce a kind of a request queue, allowing to read simultaneously from two different flash memory modules. As we can see, most SSDs increase their performance at multithreaded reading rather than otherwise. The Intel X25-V and the OCZ Colossus are the only two products to slow down.
So, the RAID0 with two Corsair drives is in the lead, followed by the Crucial (which performs better with SATA600 than with SATA300 in this particular test). Next goes the Colossus and the rest of the SSDs.
The RevoDrive is slow with one thread but gets closer to the leaders when reading two data threads. It looks like the load is not heavy enough for this product to show its best.
As the number of data threads grows up, the RevoDrive climbs higher in the standings. The opponents help, too: the RAID array built out of two Corsair SSDs slows down at four data threads and the Crucial is irresolute as to what interface is better for it.
Multithreaded reading is difficult for HDDs but multithreaded writing is such for SSDs. There are two SSDs that suffer the most when writing two threads: the Crucial slows down with both interfaces, giving way to the single-chip SandForce-based products. The OCZ Colossus has an even bigger performance hit, though.
The RevoDrive is not troubled by the increased number of data threads and the Crucial is trying to catch up with the leader (this SSD now prefers the ICH controller again). By the way, this is one of the few tests where SSDs do not deliver high speeds and do not have an advantage over hard disk drives.
As for the other products, the Vertex 2 is undecided as to what firmware version is better for it. The Agility 2 is as fast as the other SSDs with the same controller (its smaller capacity must be the reason for the small gap). The low-capacity G.Skill is worse, even though ahead of the Intel.
For this test two 32GB partitions are created on the drive and formatted in NTFS. A file-set is then created, read from the drive, copied within the same partition and copied into another partition. The time taken to perform these operations is measured and the speed of the drive is calculated. The Windows and Programs file-sets consist of a large number of small files whereas the other three patterns (ISO, MP3, and Install) include a few large files each.
You should be aware that the copying test not only indicates the speed of copying within the same disk but is also indicative of the latter’s behavior under complex load. In fact, the SSD is processing two data threads then, one for reading and another for writing.
This test produces too much data, so we will only discuss the results obtained in the Install, ISO and Programs file-sets. You can use the following link to view full results.
Creating files in FC-Test is a real-life load. The speeds are high but none of the SSDs is faster than 225 MBps. The peak speeds can be observed in the Install pattern, i.e. not with the largest files, which means that the OS’s data caching is most beneficial for SSDs.
When creating large files, the dual-controller solutions and the Crucial C300 are in the lead. The latter SSD is superior to the single-controller SandForce-based products but is let down by the Marvell controller on small files.
The Vertex 2 performs worse with the new firmware again. The G.Skill, on the contrary, looks good and faster than the Intel X25-V.
We can see the most advanced solutions compete fiercely in the reading test. The RevoDrive snatches the win by outperforming the RAID array of two Corsair Force drives. The only problem is that the speeds are still far from the promised 500 megabytes per second. We guess the SSDs can’t reach that mark under such load. Reading a single file is too easy for them.
Once again we can note that the pair of Indilinx controllers in the Colossus doesn’t look good in comparison with the new-generation products. The Crucial with a single Micron controller is even a little bit faster.
We’ve got two winners at copying: the RevoDrive and the Crucial. The latter is once again let down by the SATA600 controller, though. We can also note that the new firmware is not good for the Vertex 2.
Compared with the previous versions, the Vantage version of PCMark is more up-to-date and advanced in its selection of subtests as well as Windows Vista orientation. Each subtest runs ten times and the results of the ten runs are averaged.
Here is a brief description of each subtest:
Basing on these subtests, the drive’s overall performance rating is calculated.
The RevoDrive is ahead of its closest pursuer by a third in this test! The faster interface, even though via a PCIe—PCI-X bridge, ensures an advantage to its pair of SandForce controllers. The second place goes to another such pair but joined via the chipset-integrated RAID-controller. Take note that the latter solution enjoys a terrific advantage in the Media Center test which is sensitive to caching. This caching is performed by the driver.
As for the other SSDs, the Vertex 2 performs worse with the newer firmware. The Crucial isn’t good compared to its opponents. The OCZ Colossus takes last place, being unable to compete with the new-generation products.
Next goes our homemade test of defragmentation speed. We created a very defragmented file system on a 32GB partition of a disk by loading it with music, video, games and applications. Then we saved a per-sector copy of it. Now we copy that partition to the disk we want to test. We run a script that evokes the integrated defragmenter of Windows 7 and marks the time of the beginning and end of the defragmentation process. For more information about this test, you can refer to this article.
We must remind you that defragmentation is useless and even harmful for solid state drives due to their operation principles. However, we use this test as it allows to benchmark their performance at a rather peculiar load that involves both reading and writing of small data blocks.
Caching rules. That’s why the Crucial C300 wins this test. With its not very high results in the random write test when processing small data blocks, it could only achieve this victory through advanced caching mechanisms. The RevoDrive looks expectedly excellent.
It is low-capacity SSDs that are inferior in this test. The test zone occupies a large part of their storage space, so they have less space for the free block pool. As a result, they have to sort data more when writing.
Now we are going to show you one more interesting test in which we use WinRAR version 3.91 to compress and then uncompress a 1.13GB folder with 8118 files in 671 subfolders. The files are documents and images in various formats. These operations are done on the tested drive. This test depends heavily on CPU performance, but the storage device affects its speed, too.
After being nothing but mediocre through most of our tests, the Colossus shows its good side here. The rest of the SSDs have very similar results, indicating that we are limited by the CPU performance. The Intel X25-V is the only drive to fall behind its opponents, although the same-capacity G.Skill is just as fast as its higher-capacity counterparts.
Most of the SSDs have similar results when unpacking the archive, too. We can only name the losers: the two low-capacity models, especially the X25-V, and the Crucial (with both interfaces).
You can refer to our Hard Disk Drive Power Consumption Measurements: X-bit’s Methodology in Depth for details on this test. We will just list the specific modes we measure the power consumption in:
We’ve got fewer participants here because we can’t measure the power draw of a RAID array or PCIe devices. We also do not want to look for any differences in the power consumption of the Vertex 2 with its two firmware versions.
Let’s check out each mode one by one.
The G.Skill and the Agility 2 have the lowest power requirements when staring up. The latter is especially interesting because the similar Corsair Force and Vertex 2 need somewhat more power. The Agility 2 may have fewer memory chips, we guess. The Crucial, with an ARM processor inside, needs a lot of power, but the OCZ Colossus, being actually two SSDs in one case, needs the most.
The Colossus needs twice as much power as the others in idle mode. Among the other products, the Intel and the Vertex 2 have somewhat better results, but such small differences may even be due to inaccuracies in our measurement methods.
The Vertex 2 is the best under random-address load but the others are rather economical, too, even the Crucial. The Intel SSD acts up: its power consumption is almost twice higher at random writing than at random reading. Anyway, it is the OCZ Colossus that takes last place: having two SSDs inside, it just can’t perform any better in this test.
Every SSD needs more power for sequential writing because there are more data to process. The standings haven’t changed, though, except that the OCZ Vertex 2 enjoys a bigger advantage here. If you want a fast and economical SSD, this is the model to choose among SandForce-based products.
It’s time for us to sum up this test session.
Let’s start with two solutions that receive X-bit's Ultimate Innovation title today:
The first is the OCZ RevoDrive. We had expected excellent results from this RAID0 array of two SandForce controllers hidden behind a PCI Express interface and we are not disappointed. This SSD is brilliant. If you need the fastest SSD available and have an empty PCIe x4 slot, you should definitely consider the RevoDrive. Take note, however, that it will only show you its best under really heavy loads. The RAID0 array we built out of two Corsair Force SSDs suggests that, too. The performance benefits are obvious but not twofold. So, you should think twice before doubling the number of devices.
The second winner is Crucial RealSSD C300. It is even ahead of its time somewhat because it lacks a decent SATA600 controller to show its very best. We’ve got an impression that this SSD, with a unique controller which is based on an ARM processor, is especially good at server loads. We are absolutely amazed at its writing performance in our server-oriented tests.
We are proud to award OCZ RevoDrive and Crucial RealSSD C300 with our Ultimate Innovation title:
The OCZ Vertex 2 is a disappointment because its firmware update has lowered its performance rather than otherwise. The changes are not dramatic, yet annoying anyway.
The OCZ Agility 2 looks like a weaker version of the Vertex 2. Perhaps the difference is due to the twice lower capacity. We should compare same-capacity models to be sure.
The G.Skill Phoenix Pro leaves a good impression. Despite its low capacity, it is rather fast and often ahead of the Intel X25-V, which is a strong opponent. So if you don’t need more than 40 gigabytes, low-capacity SandForce-based SSDs may be a good choice.
The OCZ Colossus with two Indilinx controllers is not quite competitive against newer products that have only one controller. OCZ should think about updating its 3.5-inch SSD series. In fact, they have already begun that as we’ve recently got a 3.5-inch Vertex 2.
As we wrote in the Introduction, things are changing very fast in the SSD world. During the time it took us to prepare this review, we received a lot of new SSDs of various capacities. So, we’ve got enough devices for a couple of new reports which are going to be published very soon. We will talk about SSDs of unusual storage capacities and about other SSDs with PCI-E interface. Stay tuned!