|G.Skill FM-25S2S-64GB SATA-II MLC SSD|
|Reviews - Featured Reviews: Storage|
|Written by Olin Coles|
|Tuesday, 16 December 2008|
Page 6 of 11
ATTO Disk Benchmark Results
EDITORS NOTE: Please read Solid State Drive (SSD) Benchmark Performance Testing to understand how the benchmarks used in this article should be interpreted.
The ATTO Disk Benchmark program is free, and offers a comprehensive set of test variables to work with. In terms of disk performance, it measures interface transfer rates at several different user-specified intervals and reports read and write speeds. The drives seek and access times are not statistics made available in this application, which makes this a considerably basic tool, although the adjustable test settings allow for a large range of differing results. Manufacturers seems to like this tool because it offers them the ability to reduce the test length load size to produce high benchmark results. Conversely, Benchmark Reviews uses this tool with the largest (32 MB) test chuck used to produce our test results.
Bandwidth results from our tests are illustrated as the transfer rate in the images below, showing the performance curve for the G.Skill FM-25S2S-64GB from 512 Bytes up to 1 MB test chunks. G.Skill International Enterprises specifies bandwidth expectations as 170 MBps read and 100 MBps write for their MLC solid state drive, and so far our own bandwidth test results indicate that these are optimistic figures.
Testing with the JMB322 SATA controller, our results are shown in the first chart pictured below. Read performance for the G.Skill SSD begins to plateau from 128.0 KB to 1024 KB and generates a 123.7 MBps bandwidth around 1024 KB, indicating a high performance throughput for most file size chunks. The same is true for the write bandwidth, which sustains a 88.5 MBps bandwidth beginning at the 32 KB test chunk range. The ATTO Disk Benchmark tool reports bandwidth results close enough to G.Skill's 155/90 MBps rated maximum to believe there's a possibility under 'perfect' test conditions.
Since ATTO Disk Benchmark offers test settings for each transfer file size from .5 KB to 1 MB, Benchmark Reviews decided on using the largest (1 MB) statistic to build the chart shown below. In this illustration, I have organized the products using the sum of their read and write bandwidth speeds to determine position rank.
Once again, the MemoRight GT SSD demonstrates that an optimized generation 1.0a SATA controller can outperform even the latest generation 2.0 SATA controllers (if you care to pay the high price tag). The Mtron Pro 7500 MSP-SATA7525 trails behind, with the Patriot Warp v2 SSD just ahead of the Seagate 7200.11 hard drive and G.Skill FM-25S2S-64GB to finish out the top-level performance section. The Silicon Power SATA-II SSD, OCZ CORE Series SSD, and Mtron MOBI 3500 SSD all lead the upper-mid performance section.
The mid-range of performance in our chart is filled by the OCZ 64GB SATA-II SSD, Mtron MOBI 3000 SSD, and Western Digital Raptor HDD, which aren't really all that far off from the products positioned above them. On the other hand, the low-end of performance is a noticeable distance away. The Crucial / Lexar CT32GBFAB0 SATA-II SSD takes a major dip in write-bandwidth performance, which is mirrored by the Super Talent MasterDrive MX SSD. I'm not entirely clear on the technology each of these lower-end SSD's contain, but considering that two of them are SATA-II and match the performance of our slowest SATA-I SSD's I would contend that a design improvement is in order.
Originally I hadn't planned on including the ATTO Disk Benchmark results in this article. While the software is decent enough to mention, it was merely included because almost all SSD manufacturers test with it... and for very good reason I have learned. After several tests had been completed, I began to see why they decided on this particular software for benchmarks. What I like least about ATTO Disk Benchmark is how you can manipulate the settings to produce extremely wide range results from the same product.
As an example, if you reduce the total test length size from 32 MB (used in our testing configuration) to one of the smaller sizes the benchmark results are more than 30% different (as in higher bandwidth). This reason alone is enough for manufacturers to tweak their own test configurations for self-benefit. Nevertheless, in regard to our ATTO tests the entire range of SSD products is very well represented with our configuration. It should be noted that with cache buffers growing larger and larger, perhaps there's a reasonable middle ground that will provide the best of both worlds. I suspect that hybrid drives could play an important role in this argument very soon.