by Aleksey Meyev
07/08/2010 | 03:44 PM
We will be talking about solid state drives from Corsair today. The company’s SSD line-up has expanded considerably since we tested them last.
After our recent tests of SSDs from OCZ we are especially interested in comparing same-series products of different storage capacities. The capacity has but a small effect on a hard disk drive whose performance depends mostly on the recording density, heads actuator and firmware algorithms. But when it comes to SSDs, we have such performance-determining factors as the controller, firmware and capacity. We might have inferred that from the top speeds the manufacturers write into their product specs, yet we were indeed surprised to find different-capacity SSDs to differ that distinctly both at sequential operations and at server loads which included even a small share of write requests.
Every fact needs rechecking, of course. Therefore we are going to add the results of Corsair products to our database now.
Here is a whole bunch of products from the Extreme series. We’ve been lucky to get nearly each existing model, from the humble 32GB to the serious 256GB one. Some time ago we tested the 128GB model, but now we’ll check out the whole series using our updated testing method.
Each Extreme series model is based on a slightly modified Indilinx Barefoot controller (IDX110M00-LC) and, like Barefoot-based products from other brands, is equipped with 64 megabytes of cache. The following speeds are specified for the two top-capacity models: up to 240 MBps at reading and up to 170 MBps at writing. The 64GB model is promised to deliver 230 and 135 MBps, respectively. The 32GB model is expected to be as fast as 210 and 100 MBps.
Sad but true, Corsair does not offer firmware updates for these SSDs and says explicitly that they do not support TRIM.
Interestingly, the firmware versions are different: the 32 and 64GB models have firmware version 2.1. The 128GB model has firmware version 1.0 and the 256GB one, version 1.1.
However, it is still possible to update the firmware with the new one with TRIM support. Although, there are so many notes to take into account that you will definitely think twice before you decide to actually do it. Take for instance the fact that Corsair supports firmware updates and considers them perfectly “legal” from the warranty standpoint, but doesn’t recommend doing them. You can read more about it in the company forums.
The model name makes it clear that this SSD is based on a SandForce controller. Indeed, it is a SandForce SF-1222TA3-SBH chip, the same as in the OCZ Vertex 2. This controller ensures increased data reliability and does not need external cache memory. It is fast at both sequential and random-address operations. Like its competitors, Corsair offers two subseries within its SandForce-based series. The earlier one had round storage capacities, namely 100 and 200 gigabytes, whereas the more recent series comes in capacities of 60, 120 and 240 gigabytes. As far as we know, this was achieved by reducing the number of memory cells accessible by the controller but not accessible by the user. In other words, it is the redundant capacity which is present in every modern SSD so that the controller could have some free memory cells to process write operations more efficiently. Thus, the transition of a model from a formatted capacity of 100 to 120 gigabytes leads to a certain reduction in performance. Corsair doesn’t mention this, however, and specifies top speeds of 285 and 275 MBps for every model in the series. Perhaps such speeds are indeed possible in the ATTO Disk Benchmark, which is mentioned as the source of the specified numbers, but we can hardly have more than 250-270 MBps using the SATA300 interface these SSDs are equipped with. We’ll check this out shortly, though.
This series supports TRIM. Our model has firmware version 30CA13F0.
The Nova series is the Extreme series reinvented to lower the price. It includes four models of the same capacities and based on the same Indilinx Barefoot controller. The flash memory itself is different: the Nova series uses Intel modules whereas the Extreme series has flash memory from Samsung. The specified speeds are somewhat different, too. Our 128GB model can deliver up to 270 MBps at reading and up to 195 MBps at writing. The senior model has a top read speed of 250 MBps. The 64GB model has lower specified speeds: 215 MBps at reading and 130 MBps at writing. The 32GB model is declared to deliver 195 and 70 MBps, respectively.
As opposed to their predecessors, these SSDs support TRIM.
Our sample has firmware version 1.0.
We tested a 128GB representative of this series about a year ago. Perhaps not a long period of time, but the SSD market has seen a lot of changes. SSDs with Samsung S3C29RBB01-YK40 controllers (also known as PB22-J) have steadily sunk from the category of very good products into the bottom market segment. Our two samples have specified read and write speeds of 220 and 180 MBps and differ from the 64GB model which has a top write speed of 120 MBps only. The series has expanded upwards, too. It now includes a 512GB model, designed in the 3.5-inch form-factor, with read and write speeds of 240 and 200 MBps. It is for this series that the company offers firmware to update old models and add TRIM support. Our two samples have it: it is firmware version VBM19C1Q.
Finally, here is another new series from Corsair targeted at the low-end market segment. The Reactor series is based on the JMicron JMF612 controller we saw, in a modified version, in SSDs from Western Digital. Now we’ll have a look at this controller from a different side, especially as it is indeed different: Corsair’s series includes other storage capacities (60 and 120 gigabytes) and each model is equipped with 128 rather than 64 megabytes of cache. We are promised a read speed up to 250 MBps and a write speed up to 170 MBps (110 MBps for the 60GB model). The Reactor series supports TRIM. Unlike Western Digital, Corsair has implemented the controller’s support for USB 2.0 (alas, not USB 3.0): you can see the USB connector next to the pair of SATA ones.
Our model has firmware version 1.0.
The following testing utilities were used:
The SSDs were tested with the generic OS drivers and formatted in NTFS (wherever formatting was required) as one partition with the default cluster size. 32-gigabyte NTFS partitions with the default cluster size were created for FC-Test (if the drive is smaller than 64 gigabytes, it is partitioned in two halves). Each SSD was connected to a mainboard port and worked with enabled AHCI. We want to remind you that we are now using new methodology for testing 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 changes each minute, so that we could see the dependence of the tested drive’s sequential read/write speed on the size of the processed data block. This test is indicative of the maximum speed the drive can achieve.
The numeric data can be viewed in the tables below. We will be discussing graphs and diagrams.
It’s simple with the top speeds of reading. The two SSDs from the Performance series and the junior (32 and 64GB) models of the Extreme series are about as fast as 200 MBps whereas the others are close to 250 MBps. There is no real competition when the SSDs process small data blocks: the new SandForce controller accelerates faster than the others as the data block size grows up.
Interestingly, its performance is so high on small (up to 4 KB) data blocks that it loads our testbed’s CPU by 95% with requests processing (there are about 30 thousand requests per second!). Well, such loads are rather theoretical ones because there is no reason for any application to read sequentially located data in such small chunks, yet we are really surprised to see our CPU almost suffocated by disk subsystem requests.
Another surprise is the performance of the Reactor. It outperforms many opponents on all data blocks, leaving the Performance series behind, for example.
The results are much more varied at writing. Like the OCZ Vertex 2 in our previous review, the SandForce-based SSD does not meet our expectations and stops far from the specified top speed, delivering a write speed of only 133 MBps.
Then, the Extreme series shows it clear enough that there is a difference in performance between different-capacity products which is roughly equal to what the manufacturer declares. We should also note that the Indilinx-based products reach their top write speed on 16KB data blocks and then slow down a little on larger blocks. The Nova copes better but we don’t know if this is due to its different flash memory chips or some changes in firmware.
The Performance and Reactor series drives are competing for top places. Their performance on large data blocks depends on the particular size of that block, which is due to some imperfections in their controllers. Or rather the performance depends on when and how the controller erases the memory cells.
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.
The SSDs all have very low read response time. The Reactor and the 256GB Performance are somewhat worse than the others, but the difference is negligible at less than 0.1 milliseconds.
The results differ more at writing. The Reactor and the senior Extreme series models are close behind the leading Force whereas the Nova, despite its 128GB capacity, is as slow as the junior Extreme series models. The Performance series performs poorly, the 128GB model having a response time of over 1 millisecond (that would be a fantastic result for an HDD but only a mediocre one for an SSD).
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.
We can note two facts about this random-address reading. First, the Reactor is slower than its opponents on small data blocks (by the way, when looking through the results of this test, you should be aware that hard disk drives are about 50 times slower here). Second, once again (after the OCZ Vertex 2) we see the SandForce being slower than the other controllers on data blocks of 4 KB and smaller.
Here, the results correlate with the write response time of the SSDs with minor exceptions that can be observed in the diagrams. For example, the JMF612-based Reactor prefers very small data blocks, coping just perfectly with them, but slows down on large data blocks, sinking from a leading position to the group of outsiders. We can see the pair of Performance drives with Samsung controllers suffer a similar, if not so dramatic, performance slump.
Take note that the Extreme series drives go in pairs: “large & fast” and “small &… not so fast”. The Nova should belong to the first group with its capacity, but falls into the second one with its performance.
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 the following links to view the tabled results for IOMeter: Database.
We will build diagrams for request queue depths of 1, 16 and 256.
When the request queue has the shortest depth possible, the Performance series is ahead at pure reading and at 10% writes. At high percentages of writes the large-capacity SSDs of the Extreme series take the lead. The Extreme drives enjoy a larger advantage at reading than the Performance ones at writing, though.
As for the other products, the Nova is comparable to its predecessors at very high percentages of reads but loses to them, even to the 32GB model, at writing. The Force is always in the middle between the leaders and losers, and the Reactor is poor at both reading and writing. The latter SSD is more or less good at some mixed loads only.
When the queue grows longer, we have a completely different picture. The SandForce controller wakes up, taking the leading position at high percentages of reads and being one of the best at writing. The Samsung controller in the Performance series does not like long queue depths and can only beat the Reactor. The Indilinx-based SSDs fall into two groups again basing on their capacity, and the Nova is worse than the Extreme series at writing.
Winding up this part of our tests, we want to show you diagrams that illustrate each SSD’s performance at five different request queue depths.
The graphs are drawn to the same scale and serve as good illustrations of the SSDs’ writing performance. We are now absolutely sure that not only sequential speeds but also random writing performance of Indilinx-based SSDs depends on their storage capacity. The graphs are very stable: the firmware is flawless and predictable but there is no performance scalability at long queue depths, although these SSDs do some reordering of read requests.
The Force model’s controller can work with queue depths up to 16 requests long. It improves its performance at longer queue depths irrespective of the ratio of reads to writes. So, the controller either reorders them regardless of their type or employs some other mechanism that helps improve writing performance.
The Indilinx controller performs differently in the Nova: at longer queue depths, it loses performance sooner, i.e. at a lower percentage of writes. If this is due to its firmware, then the firmware is not good. And if it is due to the different flash memory type, the performance tradeoff is rather too large.
We know what to expect from SSDs based on the Samsung controller and the diagrams illustrate its behavior well enough. It can do read request reordering, but this leads to a sudden performance hit at mixed loads. Interestingly, the larger-capacity model delivers higher performance than its smaller-capacity cousin.
The JMF612-based Nova has a special character. It behaves not unlike the Samsung controller but has higher performance at low percentages of writes. Alas, when the share of writes is high, its speed plummets to 150-200 operations per second. On the other hand, this is as fast as HDDs can do and much faster than the previous JMF602B controller’s 2-3 operations per second.
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.
The Force wins this read-only load thanks to the performance growth at long queue depths. It would be just perfect if it had a smooth graph, indicating a lack of firmware flaws. The Nova is again somewhat slower than its cousins with the same controller but also faster than the Performance series. The Reactor is the slowest again.
This pattern introduces some write requests but the Force is still unrivalled. This SSD can scale its performance up depending on the request queue both at reading and at writing. The Nova improves its standing, now being ahead of the small-capacity Extreme series models although behind the large-capacity ones.
The rest of the SSDs slow down at long queue depths. The Reactor is faster than the P256 which in its turn is ahead of the P128. We wouldn’t recommend using these SSDs for applications that involve a lot of writing.
The multithreaded tests simulate a situation when there are one to four clients accessing the hard 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 applications. You can also click the following links for the full results:
The Force and Reactor drives are somewhat better than the others at multithreaded reading. The Indilinx-based models accelerate at four data threads. This is not the first time we note this peculiarity of the Indilinx controller.
The Reactor is surprisingly the best at multithreaded writing. Based on a simple JMicron 612, it beats the SSDs with much more advanced controllers. The Force is more or less good, yet can yield more than 100 MBps at two threads only. In the Extreme series the SSDs are ranked according to their storage capacities, the Nova taking an in-between position, whereas the junior Performance model is always ahead of the senior model from the same series.
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 disk is processing two data threads then, one for reading and another for writing.
This test produces lots of results, so we will only discuss the Install, ISO and Programs patterns. Click the following link for the full results obtained in FC-Test.
We don’t see any unexpected results here. The P256 is the most remarkable drive in this test, delivering excellent performance in the ISO and Programs patterns but failing in the Install one. Its junior cousin doesn’t behave like that and takes a leading position in each pattern. The Force doesn’t show anything exceptional while the Nova and Reactor are good and perform as fast as the best (i.e. the large-capacity) models of the Extreme series.
It’s hard to name winners and losers here because the drives all deliver similar performance of about 100 MBps with small files and 200 MBps with large files, even though it’s somewhat unexpected that the Nova and Reactor are slightly better than the Force and X32. We can note that the Performance series fall behind the other SSDs when processing small files.
The Performance series is somewhat better than the others at copying files, though. The rest of the SSDs have similar results, the Force and Reactor being just a tiny bit faster. The small-capacity Extreme series models have the lowest speeds here.
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 Force is first in this test just as we might have expected. As we know from our previous review, this controller is excellent at such loads. Second place goes to the P128, which is a surprise, especially as its cousin P256 has the lowest overall score. We don’t know the reason for the twofold difference between their scores. The Extreme series models pass this test with similar results, the Reactor and Nova following them closely.
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 the disk and now copy it to the disk we want to test. Next 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 harmless 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.
Our test image is exactly 32 gigabytes large, so we cannot benchmark the Extreme X32 model in this and next tests.
And once again the P128 performs brilliantly, beating the F120, while the P256 takes three times as long to complete the task! We just can’t explain this. The rest of the SSDs are more predictable: the Extreme series are ranked up according to their capacity. The Reactor and Nova are somewhat slower than the others.
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.
The P128 is again in the lead while the Reactor and Nova take second and third places, rather to our surprise. The Extreme series are now at the bottom of the diagram, retaining their respective standings. The P256 is slow, too.
One more win for the P128 that shares top place with the Force. The other Performance series product is last, having spent half as much time as the others to unpack the archive.
You can refer to our article called 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:
Let’s check out each mode one by one.
The P256 needs the lowest current to start up, only half the current required by its lower-capacity cousin. The Extreme series are ranked up according to their storage capacity again: the larger the capacity, the higher the power requirements. The rest of the SSDs don’t show anything extraordinary here.
The SSDs are ranked up according to their controllers, indicating who is the main consumer in idle mode. The Samsung controller proves to be the most economical while the Indilinx-based SSDs require 50% more power. The SandForce needs twice and the JMicron 612 needs three times as much power as the Samsung.
There are no clear leaders at random operations. But we can see which SSDs are not economical in this mode. These are the Force and Reactor at reading and the Performance series at writing. The Reactor is also voracious at random writing, requiring over 2 watts of power.
Seven out of the nine products pass this test with close results. And there are two losers: the Force needs a lot of power for sequential writing and the Reactor for both sequential reading and writing.
The new SandForce controller and the Corsair Force drive based on it are a success. It offers high speed of processing files and has good performance at server loads, especially at really heavy loads. It would also be excellent as a disk for a home computer or workstation. We only wish it could deliver the specified speeds which are close to the limit of the SATA300 interface. The controller seems to be capable of that, the problem is in the firmware.
Corsair’s Indilinx-based Extreme series shows that their writing performance, both with sequential and random-address data, depends on the drive’s storage capacity. So, saving on the capacity also means that you save on performance. The Corsair Nova series is indeed somewhat slower than Extreme although has the typical advantages of the Indilinx controller. Considering the lower price and TRIM support, the Nova is an appealing option for wise users who want to find the most optimal price/performance ratio.
The Corsair Performance series has the following advantages: low price, good performance at low reads-only server loads (by SSDs’ standards; such loads would be too much for modern HDDs), rather fast processing of files, and low power consumption. This series is not good for writing. The inexplicable difference between the 128 and 256GB models of this series confuses us, too.
Finally, the Corsair Reactor series includes inexpensive SSDs for users who can’t afford (or just don’t need) the more expensive and advanced products. As opposed to SSDs with JMicron 602B controllers, the JM612-based models are not too bad at writing. They are as good as HDDs at that and much faster than HDDs at reading. Unfortunately, besides low performance, this series also has high power consumption. It needs about as much power as 2.5-inch HDDs, which is not what we might expect from an SSD.