Results
ATTO Disk Benchmark
ATTO has been around for over 20 years and manufacturers storage controllers and adapters. Disk Benchmark is software created by the company to evaluate performance of storage devices in a number of cluster sizes ranging from .5KB to 8MB. By default ATTO uses 4 threads and a file size of 256 MB. For testing on a RAID card with dedicated cache the file size was increased to 2 GB No other changes were made.
Single drive performance was excellent and pushes the limits of SATA 6Gbps transfer speeds with 562 MBps read and 517 MBps write speeds! When 8 drives are used transfer speeds hit a whopping 2.18 GBps read speeds, writes increased further and broke 2.25 GBps performance. Unfortunately in both cases, 4K filesizes do not really show the capabilities of the drives themselves and are limited by the RAID controller.
Crystal DiskMark X64
CDM x64 is a storage performance evaluation software from Crystal Dew World. The software by default tests random data at a 1 GB file size, and runs it through 4 separate read and write scenarios. Each test is run 5 times. the sequential test uses a 1024K file, the 512K and 4K tests are both random. Options exist in the setting to use a 0 fill test data, this will improve performance scores on SandForce controlled SSDs. For testing on the RAID card, random and 0 fill data is used, and file size is increased to 4 GB the maximum allowed in the software.
Single drive performance was excellent sequential performance hit 467/457 read and write respectively. 4K random performance was respectable and easily brushed over 60K write IOPS. With RAID testing the results for sequential performance sky rocketed. Once again Writes far surpassed read speeds. Hitting 1.8 GBps reads and almost 2.2 GBps in writes. Single threaded 4K performance saw only a minor bump and 4K multithreaded performance results increased well on reads but write speed was almost cut in half.
AS SSD
A program very similar to CDM random data test. AS SSD was designed solely for testing Solid State Drives. It uses a 1 GB file size and measures sequential and random 4K performance. The Multithreaded test, is an attempt to measure NCQ performance and is also an asynchronous test as no current consumer drives handle more than 32 threads. AS SSD provides read/write and total scores however this are not really important and are based solely on data that is already reported.
in single drive mode, AS SSD shows the Mercury drive to be an excellent performer. The drive scores over 500 MBps read speed and 285 MBps write. 4K single thread performance is decent, and multithreaded performance is outstanding, with well over 200 MBps reads and writes. Access time is about what should be expected with very low read times and write times being much higher.
Once 8 drives are connected and run through the software we see what we have seen before. Sequential writes dominate the results at 2.11 GBps, with Read speed falling behind at only 1.7 GBps. 4K single threaded performance sees a small bump over single drive, and multithreaded 4K performance sees an increase in read but once again a decrease in 4K writes. Access times also get skewed increasing read access to over .1 ms while write access dropped below .05 ms.
PCMark Vantage
Futuremark produced a number of PC benchmark suites over the years. Vantage is used for storage reviews, as it is more up to date than PC05, while the newer PCMark 7 has an inherently flawed storage test mechanism.
Designed for Windows Vista PCMark Vantage Storage test covers a number of end User style tasks such as importing music or videos into photo gallery, windows media player and windows media center. It measure both windows and application startup performance and also measures the performance of anti-virus scanning and playing video games.
Results in PCMark Vantage are extremely hindered by the RAID card. This is due to the method Futuremark uses to perform its tests. Single drive performance scored a total of 43 thousand points which is far less than what the drive is capable of with an onboard SATA 6Gbps port. RAID performance was not any better grabbing on 46 thousand points in best test scenarios.
AIDA64 Storage Tests
Finalwire has revamped the popular Everest software suite from Lavalys (formerly AIDA32) a complete software based diagnostic system that can be used to identify, monitor and benchmark a computer and runs on either 32bit or 64bit setups.
AIDA64 storage tests are divided into two section read and write. Unlike other benchmark software an AIDA64 write test is destructive and does not require a drive to even be formatted. Read Test suites are measured multiple times, while write tests are performed only after a secure erase and only run multiple times if results are aberrant to expectations. Various block sizes are used in all test and is not user controlled.
Single drive results are very good peaking at 527 Mbps in linear read testing. With random reads scoring over 580 MBps. This is normal on SandForce drives in AIDA64 even though it surpasses the actual capabilities of the SATA 6 Gbps interface. Access time follows what was measured in AS SSD and measure .06 ms. In 8 drive configuration the results peak out at 2 GBps exactly this may indicate a limit of the software. Random reads do not scale as cleanly and peak at just over 1.7 GBps. Also seen previously read access time increased a bit to .19ms.
In all linear tests the drives both in single or RAID configuration performed very well up until the 80% full mark. single results stayed linear at 299 MBps until it nose dived to 250 MBps until 99% full where it dropped to 80 MBps. This brought the average down to 257 MBps. RAID performance peaked at almost 2 GBps, but linearly grabbed at 1850 MBps. Instead of dropping down in performance however the RAID array actually corrupted and the controller in an effort to protect itself, tossed a drive in every linear test.
Random performance hit 285 MBps in single drive testing but averaged out to almost 260 MBps overall. In RAID 1.1 GBps was the maximum although the average came in at just under 1 GBps. Access time results were a bit more erratic in single drive testing, starting off decently at .11 ms but quickly escalating to .21 ms were it finished out the test. RAID was a complete straight line at the .05 ms mark
