SSD Performance: One Cog In The Wheel
It's easy to lose perspective after getting into the minutia of SSD benchmarking. Our PCs are comprised of multiple components, though. Storage is but one variable in an overall performance picture that is also affected by processor, graphics, and memory performance, too. That’s why upgrading to an even faster SSD won't always cut into the time it takes to boot Windows, launch Photoshop, or load a level in Crysis 2. Given the speed-up from SATA 3Gb/s to 6Gb/s or first-gen SandForce to second-gen, you'd expect massive improvements. That simply is not always the case, though.
The idea of diminishing returns in the storage world makes good sense. In order to illustrate why, let's look at the vital statistics of loading a level in Crysis 2.
| Overall Statistics | Crysis 2: Level Loading |
|---|---|
| Elapsed Time | 00:58 |
| Read Operations | 9 493 |
| Write Operations | 395 |
| Data Read | 755.01 MB |
| Data Written | 26.49 MB |
| Disk Busy Time | 2.27 s |
| Average Data Rate | 345.00 MB/s |
Using a 240 GB Vertex 3, it takes us 58 seconds to load into the last level of Crysis 2 (including the intro scene). Yet, the disk is only busy for 2.27 seconds. Why? Because the computer is doing other things the remainder of the time. Data has to be loaded into memory, operated on by the processor, loaded into the graphics processor's on-board memory, and so on.
This isn't limited to gaming-related scenarios either; it applies to everything. We're not saying there isn’t a noticeable difference between a 64 GB m4 and a 240 GB Vertex 3, because there is. However, looking at boot times and level load times isn't a pure measure of storage performance. If you measured drive responsiveness over the course of a week doing different tasks, you would feel the total effect of a faster SSD. And that's why we focus on long traces. Our Storage Bench v1.0 is a two-week trace, yielding a satisfactory gauge of responsiveness.
