ADATA SX900 128GB SSD Review

The ADATA SX900 128GB SSD came to us with a surprise under the hood, the new B02 version of the SandForce SF-2281 controller. This new stepping is designed to provide revolutionary improvements in power efficiency with no loss of speed. We test the SX900 and the SF-2281VB2-SPC controller to see how it stacks up against the competition.

continued...

Synthetic Benchmarks

ATTO Disk Benchmark

ATTO Disk Benchmark is one of the industry standard benchmarks. ATTO is used to generate many of the sequential specifications on SSD packaging and marketing statements. ATTO is a very simplistic tool that generates easy to understand results for most users.

One of the key complaints about ATTO from many of the enthusiast crowd is that it uses easily compressible data. For drives that rely upon compression (LSI SandForce controllers), this benchmark will only tell half of the story, and the rosy half of it at that. Tests with incompressible data will show much lower scores at the very same measurements with the LSI SandForce controllers. Used in conjunction with other tools that test with incompressible data, ATTO can provide a good view of overall storage subsystem performance.

ATTO serves to verify that the rated specifications on the box are in fact attainable, and that the device is in basic good working order.

Article Image

ATTO testing reveals very solid sequential performance with fully compressible data. We observe a top read speed of 559MB/s and 528MB/s write speed.

Again we note the lower minimum speed during the test, and the Kingston V300 shares this characteristic. Without current firmware it is impossible to determine the impact of the new stepping on this metric.

HDTUNE PRO

HDTune Pro provides an in-depth look at many different aspects of storage system performance, and allows for many customizations in testing that are beyond what other benchmark programs offer. One key aspect of HDTune pro testing is that it tests over the entire surface of the drive. This eliminates 'burst' results, providing a more accurate representation of the overall sequential speed. This can reveal certain characteristics in drives which are normally overlooked.

Article Image

HDTune Pro Read results find the ADATA SX900 falling slightly behind the similar capacity Samsung TLC 840 and Kingston V300. The minimum speed is also significantly lower than other SSDs in the test pool.

Once we have working firmware for the B02 controller the minimum speed will be an important metric. With the controller going into advanced idle modes with several NAND packages turned off, there might be a slight pause as it ramps up under load.

Article Image

HDTune Pro Write tests the sequential write speed of the SSDs, and the SX900 easily bests the 120GB TLC 840. It holds its own with its SandForce-powered peer, the Kingston V300. The Samsung Pro continues to lead the charts in average speed, and the Intel falls into fourth place.

Access Time – Iometer

Iometer is a comprehensive utility that can be an intimidating testing tool for the uninitiated. One of the most relevant metrics that it can test is also the simplest, the measure of the access time of the attached storage. Latency is one of the key measurements that can relate directly to a fast, responsive computer. Latency is the measurement of how fast the device can respond to a single request, and many utilities measure access time with a number of different approaches. Unfortunately not all measure correctly, so we will test with the industry standard 4k random access at a QD of 1. We attain our numbers by testing at 4kQD1 with a 5 sec ramp time for 30 seconds. Three runs are conducted and the results are averaged. The SandForce-powered SSDs relies upon compression for part of its performance, and is tested with both random (incompressible) and repeating (compressible) data to highlight the two levels of performance.

Article Image

Average Read and Write Latency results fall within the same range as the other SSDs in the test field. This points out that there will probably not be a significant performance gap with the older SandForce stepping.

Article Image

Maximum Read Latency takes into account the longest single read access (file) requested during the duration of the test. By measuring this "Longest I/O" it is possible to see the efficiency of the controller and firmware of the SSD. If there is a correlation between a very high maximum latency and an overall higher average latency, that can be indicative of a device/host issue. Even with an SSD kicking out an appreciably higher maximum latency the result would have to be in conjunction with higher overall latency to indicate a problem.

The average latency is important, but having a very fast SSD is only good if there are not long hangs for individual file requests.

The SX900 scores well in this test, and the 120GB TLC 840 towers above the rest, with a very high maximum latency which is undesirable.

Article Image

Maximum Write Latency takes into account the longest single read access (file) that is requested during the duration of the test. The SX900 also performs well at this test.