Page 1:Getting Older By The Minute
Page 2:Drives On The Block
Page 3:How We Tested
Page 4:Benchmark Results: CrystalDiskMark
Page 5:Benchmark Results: HD Tune
Page 6:Benchmark Results: PCMark Vantage
Page 7:Benchmark Results: PCMark Vantage, Continued
Page 8:Really Simple
Page 9:Really Loaded
How We Tested
As you know, there are two kinds of benchmarks: synthetic and real-world. Synthetics seek to obtain absolute values. “All other things being equal or isolated, the product can do this.” A real-world benchmark usually involves taking some sort of stopwatch (mine is a cheap three-button unit from Sportline) and measuring the time it takes the system or application to perform a given task set. Real world benchmarks may isolate some variables, but a lot of the point is to emulate ordinary usage conditions. In our case here, this is particularly important. I expected there to be a difference between the conclusions drawn from synthetic and real world tests. The question was how much of a difference there would be.
Here is the test platform we used:
|Processor||Intel Core i7-965 Extreme (45nm, 3.2 GHz, 8MB L3 Cache)|
|Motherboard||Intel DX58SO (Intel X58 Express/ICH10R), BIOS SOX5810J.86A|
|RAM||3 x 2GB DDR3-1600 OCZ3P1600LV6GK|
|Hard Drive 1||Kingston SSDNow V-Series, 64GB (SNV125-S2BN/64GB), SATA 3 Gb/s, 64KB In-controller cache|
|Hard Drive 2||Transcend SSD25D, 60GB (TS60GSSD25D-M), SATA 3 Gb/s, 64MB cache|
|Hard Drive 3||Intel X25-M G2, 160GB (SSDSA2M160G2GC), SATA 3 Gb/s, 32MB cache|
|Hard Drive 4||Western Digital VelociRaptor, 300GB (WD3000HLFS), 10,000 RPM, SATA 3 Gb/s, 16MB cache|
|Graphics||EVGA GeForce GTX 280 (01G-P3-1280-AR)|
|Power Supply||PC Power and Cooling Turbo-Cool 860W|
|Performance Measurements||PCMark Vantage 1.0.2|
|HD Tune Pro 3.5|
|System Software And Drivers|
|Operating System||Windows 7 Ultimate Edition (fully updated with Windows Update on 12/29/2009)|
|Intel Chipset Drivers||Chipset Installation Utility 184.108.40.2069|
|Nvidia Graphics Drivers||GeForce 195.62|
I like CrystalDiskMark and HD Tune for their very quick assessments of raw performance. HD Tune in particular can reveal interesting drive attributes in its throughput graphs. Historically, we’ve leaned toward the Productivity tests within PCMark Vantage, and I actually ran that suite on these drives. However, I threw out the data in favor of the HDD Suite, which focuses more precisely on the type of common Windows tasks we want to highlight here.
For real world testing, I developed two scenarios. The first focused on a simple boot into Windows 7, just looking at going from a cold boot to the Windows desktop. The second took this first scenario and then piled a ton of apps and data on top of it. This is closer to being a true real world situation. After a system reset, I inevitably find myself gravitating back to having two or three spreadsheets, many browser windows, multiple .doc files, Outlook, and other apps open throughout the day. If I were smarter, I’d just have these constant companions pre-loaded from my Windows Startup folder. For this test, I assumed I was actually that smart.
Note that boot times were measured starting after the motherboard POST period. Specifically, the DX58SO will show a splash screen followed by a couple of alphanumeric codes in the bottom-right corner of the screen. The last of these is “94" and then the screen goes black. This is the point when I would start the timer.
Final note: I burned a lot of hours repeating tests because I didn’t have patches and drivers properly set. If you want top results from your SSD, flash your motherboard's BIOS first, then the SSD firmware, then install Windows, then get the most current storage drivers installed. In the case of my test setup, I kept getting inexplicable drop-offs in two specific PCMark Vantage tests that resulted from running Microsoft’s default storage driver rather than Intel’s Matrix Storage driver (version 8.9). I’d also missed how a motherboard update had reset my disk mode to legacy IDE rather than SATA AHCI. This resulted in semi-normal looking benchmark scores but bizarre hibernation resume times of up to five minutes.