As we had just completed similar testing on the Kingspec 1TB MultiCore PCIE SSD a few days back, we thought it important that we tried to copy that testing scenario as close as possible and elected to use the same bench, adding into our test stream the OCZ RevoDrive 3×2 PCIe SSD that we also have on hand. SSD Testing at TSSDR differs slightly depending on whether we are looking at consumer or enterprise SSDs. For consumer SSDs, our goal is to test in a system that has been optimized with our SSD Optimization Guide, although CPU C States have not been changed at all. Benchmarks for consumer testing are also benchmarks with a fresh drive so, not only can we verify that manufacturer specifications are in line but also, so the consumer can replicate our tests to confirm that they have an SSD that is top-notch. We even provide links to most of the benchmarks used in the report.
Enterprise testing is significantly different as we explore performance in steady state, explore drive latency, and do our best to follow SNIA test protocol. For the KingSpec PCIe SSD, this SSD can be a consumer/enterprise varient so we are going to start things along with our typical consumer benchmarks, and then explore some testing at steady state in IOMeter.
This new PCIe Test Bench build was the result of some great relationships and purchase; our appreciation goes to be quiet, Corsair, Crucial, Intel and InWin for their support in our project. Our choice of components is very narrow, in that, we choose only what we believe to be among the best available and links are provided to each that will assist in hardware pricing and availability, should the reader be interested in purchase.
|PC CHASSIS:||InWin D-Frame Open Air Chassis |
|MOTHERBOARD:||ASUS Maximus VI Z87 MotherBoard|
|CPU: ||Intel Core i7-4770K CPU|
|CPU COOLER:||Corsair H100i CPU Cooler|
|POWER SUPPLY:||be quiet Dark Power Pro 10 1000W PSU|
|SYSTEM COOLING:||be quiet Silent Wings 2 PC Fans|
|GRAPHICS CARD: ||MSI Radeon HD 7870 Hawk Gfx Card|
|MEMORY:||Crucial Ballistix Tactical Tracer 1600Mhz Memory|
|KEYBOARD:||Corsair Vengeance K95 Mechanical Gaming Keyboard|
|MOUSE:||Corsair Vengeance M95 MMO/RTS Laser Mouse|
|ROUTER:||NetGear R6300 AC1750 Dual Band Gigabit WiFi Router|
|HBA||HighPoint RocketU 1144C 4 x USB 3.0 20Gb/s HBA |
The software we will be using for today’s analysis is typical of many of our reviews and consists of ATTO Disk Benchmark, Crystal DiskMark and IOMeter. Our IOMeter testing was conducted both in a fresh drive state (at least in the case of the Scorpion Deluxe) as well as testing in steady state and testing was conducted to determine read and write IOPS and throughput, as well as file server, web server and workstation pre-configured tests comparing all drives. This report also contains true-to-life transfer tests of all three cards.
In our analysis of both the KingSpec and Mushkin SSDs, there were serious compatibility concerns with AS SSD Benchmark software as well as Anvil Storage Utilities, in that both were incapable of properly testing write throughput. We initially thought it might be a system issue and tested both PCIe SSDs on three different systems where we observed the same results from each. To go one step further, we pulled out our tried and true OCZ Revo 3×2 PCIe SSD and tested it on all three systems where we observed results that matched manufacturers specifications.
In examining both the KingSpec and Mushkin SSDs, we found a common feature which was their use of LSI’s own Fusion MPT2 firmware. Believing that this could be the only possible reason for such, we contacted LSI and we are awaiting return. I quick search of the internet seems to indicate that other have experienced similar with the same software. Having said that, the incompatibility was all the reason we needed to conduct true to life testing, as well as enterprise testing in IOPS
ATTO Disk Benchmark is perhaps one of the oldest benchmarks going and is definitely the main staple for manufacturer performance specifications. ATTO uses RAW or compressible data and, for our benchmarks, we use a set length of 256mb and test both the read and write performance of various transfer sizes ranging from 0.5 to 8192kb. Manufacturers prefer this method of testing as it deals with raw (compressible) data rather than random (includes incompressible data) which, although more realistic, results in lower performance results.
Understanding that we can set ATTO to suit our needs with respect to testing incompressible data samples as well, we choose to leave it as a typical person might if they just bought the card and were checking it out quickly. Understanding that this method represents a very unrealistic data sample, it is still the method used by manufacturers to set their specifications. A quick look below gives us our first comparison of the Mushkin Scorpion Deluxe, KingSpec Multicore and Revo 3×2 PCIe SSDs:
Although the KingSpec is ‘King-of-the-Hill’ at 2.5GB/s read throughput, the Mushkin Scorpion Deluxe handily surpassed it in write performance at 1.9GB/s which was totally unexpected. Considering we have had (and used) this OCZ Revo 3×2 for a few years now, it still provides amazing performance. Just for good measure, we also included results of the Samsung 840 Pro SSD as a point of comparison.