by Aleksey Meyev
01/18/2010 | 02:02 PM
At the current moment the performance of an SSD drive is largely determined by its controller and firmware and is almost unaffected by the brand of its flash memory chips. It is the controller, the number of channels it supports, its ways of interacting with memory and its load optimization algorithms that determine the SSD’s top sequential speeds as well as its performance at random writing, which is the most difficult operation mode for flash memory.
In our previous SSD-related reports we paid much attention to products with controllers from Intel (both of the first and second generations) and Indilinx (with different firmware). We also tested SSDs with Samsung’s controller, e.g. the OCZ Summit. Up till now, we have only omitted to test widely available SSDs with a JMicron 602B controller. We will do this test now and will also throw in an Indilinx-based Corsair X128 and a couple of earlier-tested products.
As soon as SSD controllers had got more or less mature, there appeared a lot of various products based on them. Many companies seized the opportunity to enter a new and promising market. Quite expectedly, the foremost of them were the makers of other types of flash memory products as they only had to adapt to the new controllers and order/manufacture new product casings. Thus, Apacer came up with a series of MLC-based SSDs with JMicron’s controller. We’ve got a 64GB sample.
We already tested a Corsair-branded SSD. It was the Corsair P128 model from the Performance series which had a Samsung controller inside. Today, we’ve got an Extreme series, Indilinx-based, 128GB model called X128.
Like Apacer, PQI is a well-known maker of external flash storage products. At two times the capacity as its opponent, the PQI S525 also features a USB 2.0 connector besides a SATA one. The additional interface is not PQI’s innovation, though. All SSDs based on the JMicron controller may be equipped with a USB connector as the controller itself supports that interface. Some makers just don’t think it necessary to install it.
Transcend is a well-known name in the flash memory industry, too. As opposed to the previous MLC-based products, this SSD is based on the more expensive SLC variety of flash memory. This ensures a much longer specified service life and a higher speed of writing.
We will also take a couple of SSDs we tested earlier for the sake of comparison. One of them is the brilliant second-generation Intel X25-M and the other is the above-mentioned Corsair P128.
We did not change the Intel SSD’s firmware with the newer version that supports the Trim command in Windows 7 and provides a certain performance boost. Alas, Intel had failed twice in the same way: the firmware they had made available for download at the time of our tests caused a data loss under certain conditions. Therefore we decided that we would check it out and the efficiency of the extension to the data transfer protocol later on. Perhaps the Trim command itself will gain broader support. Right now, it is supported by only a few products (by Intel’s second-generation SSDs after firmware update and products with the Indilinx controller that can be updated to firmware version 1.4) and only by those Intel chipset drivers that are bundled with Windows 7. We are still running Windows XP but considering switching to the new OS.
We use the following method of benchmarking SSDs:
We guess this method will help compare the SSDs under identical conditions and yet produce real-life results.
The following testing utilities were used:
We formatted the drives in FAT32 and NTFS as one partition with the default cluster size. For some tests 32GB partitions were created on the drives and formatted in FAT32 and NTFS with the default cluster size, too. The drives were connected to the mainboard’s SATA port and worked in AHCI mode.
We will use WinBench 99 for low-level tests. This benchmark goes first and we “organize” the flash memory cells before it (by a stream of read requests), so it is indicative of the maximum speed you can have with your SSD.
The SSDs based on the JMicron controller are absolutely identical when it comes to sequential reading. Unfortunately, their read speed is only about two thirds that of the other SSDs. They are roughly equal to modern 7200rpm HDDs, but not faster. Yes, we expect SSDs to outperform HDDs now. While early SSDs used to be far inferior to HDDs at linear operations, many modern multi-channel controllers for SSDs are almost as fast as to utilize the full bandwidth of the SATA-300 interface. HDDs are still a long way to delivering such performance.
The Corsair X128 is good. It successfully outperforms the second-generation X25-M and is indeed very close to the interface bandwidth. Perhaps it would be a little bit faster if it supported SATA-600.
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 processed data block. This test is indicative of the maximum speed the drive can achieve.
The numeric data can be viewed in tables by clicking the links below. We will discuss graphs and diagrams.
The PQI falls behind the other JMicron-based products when reading large data blocks. It obviously has the same number of channels as the others, so this must be the influence of the flash memory chips installed in it. The SSDs based on the JMicron controller are overall slower than their opponents with both small and large data blocks.
The X128 is inferior to the X25-M as well as to the P128 on small data chunks but outperforms the latter on data blocks larger than 16KB. It cannot catch up with the Intel SSD, though.
Corsair’s SSDs are ahead at sequential writing. Despite different controllers, they have almost identical speeds on most of data blocks. The Intel is about as fast as the leaders on small blocks but reaches its top speed sooner. The low speed of sequential writing is a characteristic trait of Intel’s SSDs with MLC flash memory.
The JMicron-based products go neck and neck up to 64KB data blocks. The controller and the page-based memory access mode obviously limit the performance of the Transcend drive. It does take a spurt on large data blocks, though. While the same-controller models can only catch up with the X25-M (the Apacer is a little bit ahead of the PQI again), the Transcend with SLC memory is 50% faster! On the other hand, it cannot overtake the leaders as the controller proves to be more important than the memory type even under such simple load.
In this test IOMeter is sending a stream of requests to read and write 512-byte data blocks with a request queue of 1 for 10 minutes. The total amount of requests processed by the drive is much more than its cache buffer, so we get a sustained response time that doesn’t depend on the drive’s cache.
Flash memory storage features a very low response time at reading. However, we can note that the three JMicron-based models are much slower than their opponents. The Corsair X128 is somewhat slower than the earlier-tested products, too.
And when it comes to writing, we are lost for words to describe this picture. The Corsair X128 is competitive. Its Indilinx controller boasts a much better write response than the Samsung controller of the P128 model. The newcomer can challenge Intel’s second-generation SSDs which are currently among the fastest in terms of write response.
The JMicron controller is quite a different story. It brings us back to the times of the first generation of solid state drives which did not have any firmware optimizations and had a tremendously poor write response. Just consider: these SSDs take one fifth of a second – an eon from the CPU’s or OS’s point of view – to write 512 bytes! This leads to a total failure under Windows XP which sends most of its service requests in such blocks and does not do anything else until it gets a confirmation of successful writing. As a result, the computer works very sluggishly.
The Transcend feels somewhat better as its SLC memory writes faster and does not suffer that much from the lack of caching. Still, its write response is about five times as high as that of typical HDDs. Do you want to know what an SSD with SLC memory and good controller can do? Just check out our review of the Intel X25-E.
Now we will see the dependence between the drives’ performance in random read and write modes on the size of the processed data block.
We will discuss the results in two ways. For small-size data chunks we will draw graphs showing the dependence of the amount of operations per second on the data chunk size. For large chunks we will compare performance depending on data-transfer rate in megabytes per second. This approach helps us evaluate the disk subsystem’s performance in two typical scenarios: working with small data chunks is typical of databases. The amount of operations per second is more important than sheer speed then. Working with large data blocks is nearly the same as working with small files, and the traditional measurement of speed in megabytes per second becomes more relevant.
Everything is neat and nice at reading. The processing of small and large data blocks by the SSDs depends on their response time and sequential speed, respectively.
When the SSDs are writing data in small chunks, the JMicron-based models are affected by their terrible response time. The difference is huge: the graphs of the SSDs with MLC memory and JMicron controller merge with the X-axis if shown in the same diagram with the graphs of the other SSDs. Take note that the Transcend, for all its theoretical writing capabilities in the way of SLC memory, does not differ from the other JMicron-based products. Alas, nothing can be done with the controller’s algorithms here. Frankly speaking, the Corsair P128 isn’t much better: it cannot compete with the SSDs based on Intel’s and Indilinx’s controllers at writing small data chunks.
Comparing the X25-M and the X128, Intel’s controller is faster with very small data blocks but loses its advantage quickly as the data blocks grow larger, eventually falling behind the Indilinx-based Corsair.
As a result, the X128 becomes an unrivalled leader of the test on large data blocks. Take note that the Corsair P128 overtakes and outperforms the X25-M on very large data blocks as it has a higher speed of linear writing. The Transcend’s advantage shows up finally, yet this SSD is still inferior to the products based on other controllers.
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 this link to view the tabled results for the IOMeter: Database pattern.
We will build diagrams for request queue depths of 1, 16 and 256.
At the shortest queue depth, the X128 is slower than the X25-M or P128 at very high percentages of reads. It is, however, better than the other Corsair and competes with Intel’s product at high percentages of writes.
The JMicron-based models are good at reading only. Although not as fast as their opponents, their reading performance is much higher than what you can get from an HDD. But as soon as there appear write requests, these SSDs slow down almost to a halt. The speed of less than 10 operations per second is a disaster. The Transcend with SLC memory delivers 20-30 operations per second at high percentages of writes, but this is still much lower than with any modern HDD.
When the request queue gets longer, the X25-M goes ahead, the X128 taking its deserved second place.
Next we will show you diagrams with the performance of each new SSD at five different queue depths. Such diagrams are most illustrative when it comes to SSDs.
One look at this diagram would tell us what controller is installed in this SSD because we saw a very similar diagram with the OCZ Vertex Mac Edition. So, this model has got an excellent controller with superb firmware that can effectively process write requests and make a good use of the request queue.
These three SSDs produce very similar graphs. And we have to note that their controller does not see the depth of the request queue and can only perform fast enough when processing read requests. The use of SLC memory in the Transcend model does not save the day: the drive only shows a somewhat higher performance at high percentages of writes and is faster under mixed loads with a larger share of reads.
The drives are tested under loads typical of servers and workstations.
The names of the patterns are self-explanatory. The Workstation pattern is used with the full capacity of the drive as well as with a 32GB partition. The request queue is limited to 32 requests in the Workstation pattern.
The results are presented as performance ratings. For the File-Server and Web-Server patterns the performance rating is the average speed of the drive under every load. For the Workstation pattern we use the following formula:
Rating (Workstation) = Total I/O (queue=1)/1 + Total I/O (queue=2)/2 + Total I/O (queue=4)/4 + Total I/O (queue=8)/8 + Total I/O (queue=16)/16.
We guess these four diagrams make it clear that the JMicron-based SSDs can only be recommended for read operations and only if they are considerably cheaper than SSDs based on other controllers. They almost come to a halt as soon as there are write requests to be processed. Even the Transcend, for all its SLC memory, is no better than today’s HDDs. You can easily find this out by comparing its performance ratings with those of the 2TB HDDs. The JMicron-based SSDs are absolutely unsuitable for a top-performance workstation due to extremely low performance.
The Corsair X128 is always second to the Intel X25-M and ahead of the other model from Corsair.
The multithreaded tests simulate a situation when there are one to four clients accessing the drive 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:
Multithreaded reading is an easy load for SSDs due to the way they store data. An SSD does not care what memory cell to read data from if the requested data block is large enough for multi-channel access. There are some differences in the practical test, though. The Intel X25-M delivers the highest performance and even gets faster when reading multiple data threads. Corsair’s models slow down at two and three data threads. When reading four threads, the X128 goes ahead suddenly as if this is just the right kind of load for it. The P128 does not accelerate that much.
The JMicron-based products slow down with each added thread although not as quickly as HDDs usually do.
When doing multithreaded writing, the SSDs behave in the same way as in the multithreaded reading test. The only exception is Intel’s X25-M which maintains the same speed irrespective of the number of data threads. The Corsair SSDs slow down with the addition of a second thread but accelerate when more threads are added. The JMicron-based products are steadily losing their already low speed as we add more and more data threads.
For this test two 32GB partitions are created on the tested SSD and formatted in NTFS and then in FAT32. A file-set is then created, read from the SSD, copied within the same partition and copied into another partition. The time taken to perform these operations is measured and the speed of the SSD 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.
We’d like to note that the copying test is indicative of the drive’s behavior under complex load. In fact, the SSD is working with two threads (one for reading and one for writing) when copying files.
This test produces too much data, so we will only discuss the results achieved in NTFS. You can click this link to view the full results for FC-Test FAT32.
You may guess the results of this test from the previous ones. We have three obvious leaders among which the X25-M copes best with small files, the X128 with large files, and the P128 always keeping in second place. The three JMicron-based models are only half as fast as their opponents. Take note that the Transcend SSD is but slightly faster than the other JMicron-based products.
The X25-M wins the test of reading, being pursued by the Corsair X128. The JMicron-based models show an interesting peculiarity of their controller: their speed with small files is higher than with the huge files of the ISO pattern. However, this is one of the few tests where their performance is not downright depressing.
Write operations kill the JMicron-based SSDs again. The X128 is inferior to the X25-M in every pattern. The latter’s algorithms are somewhat better here.
PCMark 2005 has the same tests as the 2004 version (not only in names, but also in results as we have seen a lot of times), so we only discuss one test from PCMark 2004 which is not available in the 2005 version. It is called File Copying and measures the speed of copying some set of files. The other results can be learned from the table. The PCMark 2005 tests are:
The final result of the average of ten runs of each test.
Click this link to view PCMark 2004 results table.
There is no point in commenting upon each individual test. The X128 and X25-M are fighting for top place. The X25-M is ahead at loading operations while the X128 wins the rest of the tests. Third place goes to the P128. The JMicron-based models are roughly half as fast as the leaders, which is reflected in the overall scores.
To make this part of our test session complete, we are going to run the latest version of PCMark called Vantage. Compared with the previous versions, the benchmark has become more up-to-date and advanced in its selection of subtests as well as Windows Vista orientation. Each subtest is run 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.
This portion of our tests goes without detailed comments as well. This is because we have the same three leaders, which change places throughout the individual tests. One thing is clear: the JMicron-based SSDs, including the Transcend with its SLC memory, are not competitive to the SSDs based on other controllers.
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. The tested disk is connected to the mainboard’s SATA controller whose operation mode (AHCI/Standard SATA) is controlled from the mainboard’s BIOS. Next we run a script that evokes the console version of the Perfect Disk 8.0 defragmenter and marks the time of the beginning and end of the defragmentation process. AHCI is turned on. You can refer to this article for details about this test.
Strictly speaking, this test makes no practical sense for solid state drives because there is nothing to defragment on them. Every memory cell is equivalent to any other, so defragmentation won’t have any effect. However, this test will allow us to compare how much time SSDs spend reading and writing the same amount of small data blocks.
The X25-M is ahead as it is generally better than its opponents at processing small data blocks. The X128 and P128 take twice the amount of time to perform the task, the P128 being somewhat better of the two: the Samsung controller must be better under this load although the Indilinx-based X128 has been generally faster at writing throughout this test session.
The three JMicron-based SSDs take far more time for defragmentation and differ with their results. The Transcend is ahead, obviously due to its SLC memory. It is hard to explain why the PQI is ahead of the Apacer here. Perhaps it is due to its larger capacity. Anyway, these SSDs both show very bad results here.
Now we are going to show you one more interesting test in which we use WinRAR version 3.8 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 SSDs split up into two groups again. Interestingly, the Transcend makes use of its faster memory type here and enters the leading trio. The Samsung controller’s firmware is not perfect as is indicated by the low result of the P128 in this test.
Unpacking the archive proves to be a highly sensitive and illustrious test. First place goes to the X128. It is 2 seconds faster than the X25-M, probably due to its higher write speed. The P128 is third, once again following the leaders closely. The Transcend is fourth, but its faster memory type still cannot make it competitive to the SSDs based on better controllers. The PQI is fifth, being somewhat faster than the Apacer, although we cannot find out if this difference is due to the different storage capacity.
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.
Funnily enough, the weakest controller is not the most economical. The JMicron-based products need less power than the top-performance X128 and X25-M to start up, but are not as economical as the Corsair P128.
This peak startup consumption is not crucial for flash memory, though.
What is more important, the three JMicron-based models are the most voracious in idle mode. This controller seems to be as simple as to lack any reduced-consumption modes.
Corsair’s SSDs are the most economical here: the P128 needs a third of a watt, and the X128 needs half a watt.
Note that this power consumption is not very low by the standards of HDDs of the same 2.5-inch form-factor. 2.5-inch HDDs need about as much power in idle mode as the more voracious of these flash-memory drives.
We’ve got similar standings when doing random-address reading and writing, but the X128 takes first place. We wonder how it manages to deliver the same writing performance as the X25-E while being as economical as the P128.
As opposed to HDDs, SSDs consume more power when writing. The difference between reading and writing is almost twofold. As a result, the SSDs reach a power draw of 2.5W (and the PQI needs even more) which is the maximum amount that a single USB 2.0 port can yield. SSDs and 2.5-inch HDDs are comparable in power consumption in this operation mode.
When data is sequential, the SSDs need more power for reading (multi-channel access puts a load on the controller as well as memory chips) but less for writing it. The only exception is the Intel X25-M whose power consumption grows up at writing, too. Like the low speed of writing, this must be the tradeoff of its architecture which is so effective at processing small data blocks. Corsair’s drives win this test with almost identical results.
This is going to be a joyless conclusion. The widely available SSDs based on the JMicron controller can only disappoint you with their performance, especially if compared with alternative products. We can only recommend them if your SSD is going to work under read-only loads. The JMicron-based SSDs we have tested are very slow at writing even if they employ SLC memory. Considering that their performance at reading is far from record-breaking, buying a JMicron-based SSD would only make sense if it is cheaper than SSDs with other controllers and you cannot use ordinary HDDs due to circumstances such as a high ambient temperature or vibration.
Otherwise, we’d recommend you some other product, e.g. the Corsair X128 based on an Indillinx controller. Coupled with good firmware, the controller makes this model competitive to the second-generation Intel X25-M under most loads, including random writing, and beats the latter at sequential reading.
Meanwhile, we have already seen hard disk drives with SATA 600 and SAS 600 and now expect these interfaces to be implemented in solid state drives as well. Additionally, SSDs have recently got a strong support in the way of the new operating system Windows 7 which can correctly erase data in flash memory. It looks like this is just the right time for the manufacturers to roll out some seriously updated SSD products!