Tom's Storage Charts 2009: A New Test Environment

Test Setup Overview and Settings

The following table lists all of the system components we use in our new storage reference system. In addition to the Supermicro motherboard and Intel Core i7 processor, we picked Corsair DDR3 memory and two rather old components: an ATI Radeon HD 3450 graphics card, and a Seagate NL35 400 GB hard drive. Neither is particularly fast, but they don’t have to be. The graphics card doesn’t matter at all for storage testing, and the hard drive is only important for system and benchmark startup. Since neither of these components affects benchmark performance on test drives, we could use essentially any component we wanted.

We picked the Radeon HD 3450, as it requires relatively low power and is passively cooled. The Seagate hard drive was one of the first business drives validated for 24/7 operation. Since we happened to have three of these drives in our labs, we decided to go with this one, so we can quickly replace a faulty drive should a problem arise, restoring the benchmarking state within a few minutes without changing the type of hard drive.

Swipe to scroll horizontally
System Hardware
CPUIntel Core i7 920 (45 nm, 2.66 GHz, 8 MB L2 Cache)
Motherboard (LGA 1366)Supermicro X8SAX, Rev. 1.0 Chipset: Intel X58 + ICH10R BIOS: 1.0B
RAM3 x 1 GB DDR3-1333 Corsair CM3X1024-1333C9DHX
HDDSeagate NL35, 400 GB (ST3400832NS) 7,200 RPM, SATA/150, 8 MB cache
GraphicsATI Radeon HD 3450 256 MB DDR2
Power SupplyOCZ EliteXstream 800W OCZ800EXS-EU
Power Measurementsh2benchw 3.12PCMark Vantage 1.0
I/O PerformanceIOMeter 2006.07.27Fileserver-BenchmarkWebserver-BenchmarkDatabase-BenchmarkWorkstation-Benchmark
Throughput TestingIOMeter Streaming Read TestIOMeter Streaming Write Test
System Software And Drivers
Operating SystemWindows Vista Ultimate SP1
Intel Chipset DriversChipset Installation Utility
AMD GraphicsCatalyst 8. 12
Intel Storage DriversMatrix Storage

Component Settings

We installed Windows Vista Ultimate with Service Pack 1 and turned off all maintenance and visual options, as well as SuperFetch, to make sure that performance doesn’t vary due to any of the built-in Windows features. In addition, we switched off all power saving mechanisms for the same reason: performance has to be reproducible. As we found recently, using power saving options does have an impact on performance of the fastest SSDs, such as Intel’s X25-M.

The memory runs at default CL9 timings, which isn’t fast, but again, it doesn’t have to be; reliability was our key priority. We installed all drivers that are typically required to get good performance: ATI’s 8.12 Catalyst release, Intel’s Chipset Installation Utility version and the Matrix Storage drivers from Intel, version