Brian's Brain

Continued from 'Solid-State Drives Challenge Hard Disks: SiSoftware Speaks, And Testing Qualifiers'...

Some you might have been surprised to discover that I didn't tackle other tests in this study...off the top of my head, candidates include:

• System boot and application launch metrics, and
• Power consumption measurements (along with battery life estimates)

I'll first remind you, dear readers, that I am a staff of one here (the dog and cats aren't of any help with techie projects...if anything, they're counterproductive in this particular regard), that I have daily blog post and other editorial responsibilities in addition to my periodic feature articles and cover stories, that I unfortunately require a modicum of occasional sleep, and that I continue to operate under the delusion that some semblance of a personal life is also desirable. ;-) With that rant out of the way, here are other thoughts:

• I consistently strive to focus my hands-on project efforts on tests that are reasonably repeatable and portably meaningful, both iteration-to-iteration and system-to-system. System boot and application launch time, for example, are heavily dependent not only on the 'raw' storage device performance characteristics but also on how many applications (and per-application features) are installed, along with the degree of drive fragmentation.
• Pragmatically, too, just how does one judge when a system is done booting, or when an application is done launching? So much code and data loading continues to take place in the background that simply 'waiting for the hourglass to disappear' is an unacceptable oversimplification of reality.
• With respect to power consumption and battery life, the storage subsystem (whether HDD or SSD) consumes only a small fraction of the total power burned by a PC. Far more substantial current draw derives from the CPU and its associated chipset, from the standalone GPU if it's present, from the LCD backlight, etc. Not to mention the fact that the (unpublished and nonlinear) inefficiency of power supplies renders the common approach of monitoring power draw at the AC plug a rough approximation at best...and that battery life is dependent both on best-case battery charge capacity and on battery age...

With all those qualifiers said, I still think this project delivered captivating and valuable results. ;-) If you concur, and if you'd like to peruse the source data that transformed into article Figures 5-7, hit this link for a ZIP of all of the ASCII-formatted Sandra report files, and hit this link for an Excel spreadsheet consolidation of their particularly relevant stats.

In closing, I've got good news to share. Several Intel SLC (single-level cell) flash memory-based 32 GByte SSDs, whose impending arrival I forecasted in the print article's sidebar, showed up on my doorstep yesterday. Stay tuned for benchmark test results (both single-drive and dual-drive RAID 0 striped, the latter in the same three configurations I tested the Intel MLC SSDs under) to be published sometime next week here at Brian's Brain. As for those still nonexistent Micron SSDs, on the other hand...