|OCZ Apex 120GB SSD OCZSSD2-1APX120G|
|Reviews - Featured Reviews: Storage|
|Written by Olin Coles|
|Thursday, 29 January 2009|
Page 8 of 14
EVEREST Disk Benchmark
Many enthusiasts are familiar with the EVEREST benchmark suite by Lavalys, but very few are aware of the Disk Benchmark tool available inside the program. The EVEREST Disk Benchmark (version 2.05.36) performs linear read and write bandwidth tests on each drive, and can be configured to use file chunk sizes up to 1MB (which speeds up testing and minimizes jitter in the waveform). Because of the full sector-by-sector nature of linear testing, Benchmark Reviews endorces this method for testing SSD products, as detailed in our Solid State Drive Benchmark Performance Testing article.
The SSD products tested with EVEREST Disk Benchmark are connected to the Intel ICH10R SATA controller resident on the Gigabyte GA-EX58-UD4P motherboard. Using the 1MB block size, our read performance measured an average 167.7 MBps and forms a perfect line as it scans across all sectors. At some invisible point in the waveform, maximum read performance peaked at 167.8 MBps, proving that the dual-JMicron JMF602B SSD controllers were not at all strained by this test. Linear write-to tests were a different story...
Linear disk benchmarks are superior tools in my opinion, because they scan from the first physical sector to the last. A side effect of many linear write-performance test tools is that the data is erased as it writes to every sector on the drive. Normally this isn't an issue, but it has been shown that partition tables will occasionally play a role in overall performance. The real issue we found with the OCZ Apex, wasn't really an issue at all.
Although the waveform chart below makes the linear write performance appear horribly low, this is a synthetic result of using a linear write test in conjunction with two internal drive controllers that reserve the first 64 sectors for a shared cache buffer. The maximum linear write performance peaked at 71.7 MBps at the start of the test, and went on to sustain 46.8 MBps average for the remainder of the test.
This actually makes sense, when you consider that the JMicron JMF602B controller offers a rather small 16KB of on-die cache buffer per chip. When the linear write test first begins, it actually overflows the controller's buffer because it writes to the reserved sectors. Please understand that in real-world scenarios the end-user would never be capable of writing data to an un-partitioned disk, so because of this particular architecture our linear write performance benchmark of the OCZ Apex SSD is rendered unreliable for comparison.
EDITORS NOTE 02/18/2009: After receiving the final/retail firmware for the OCZ Apex SSD, we upgraded our test sample and prepared it for testing with the Microsoft DiskPart utility. Unfortunately, even with the recommended preparation and latest available firmware, the OCZ Apex still offered very inconsistent (and obscenely low) EVEREST linear write speeds. Although the OCZ Apex and G.Skill Titan SSD share identical architecture, there is a difference in the firmware that does not result in the low sustained write performance from the G.Skill version.
The chart below shows the average linear read and write bandwidth for a cross-section of drives attached to the Intel ICH10 Southbridge. Although I personally consider this the single most important comparison of SSD products, by my own admission the OCZ Apex is at a disadvantage because of incompatible linear write testing with the JMicron JMF602B chips and internal RAID-0 architecture.
Linear bandwidth certainly benefits the Solid State Drive, since there's very little fluctuation in transfer speed. Hard Disk Drive products decline in performance as the spindle reaches the inner-most sectors on the magnetic platter.