4 KB Random Performance: Response Time
We're trying to dig right into the benchmark analysis, which means skipping over some of the explanation given in the past for the data we're presenting. Again, if you missed any of that before, flip over to page seven of Second-Gen SandForce: Seven 120 GB SSDs Rounded Up.
Again we see the m4s finish in the order we'd expect. However, the margins between read response times are sufficiently small that they shouldn't affect performance, unless you're talking about reading lots of data.
The observed write response time is another matter. The 64 GB m4 has a response time of 0.69 ms. That's 5x slower than the 128 GB m4, which in turn is only 85% slower than the 512 and 256 GB m4s. In the case of the 64 GB m4, not only are you writing data 5x slower than the 128 GB m4, you're also taking 5x as long to finish a write op.
Response time is a measure of the difference between initiating and completing an operation, while throughput is a measure of the amount of data transferred. These two values affect performance in different ways, but they're not multiplicative. So, it's not like the 64 GB drive feels 25x slower than the 128 GB drive. Throughput and response time are usually correlated, in that you get high throughput with low response time.
The maximum response time gives us a look at the extremes. In random reads, the SKUs we're testing fall in the reverse order this time, again with the exception of the 256 GB m4. Of the SSDs, Crucial's C300 actually performs the worst, suggesting a generational improvement.
Drives with more capacity deliver a lower maximum write response time. The 512 GB m4's response time is about 16x lower than the 64 GB model. However, even the entry-level drive is more than twice as responsive as our hard drive (Seagate's 5400.6 has a max write response time of 195.24 ms).
The different max write response times are reflective of the role hardware and firmware play in performance. The firmware in the 256 GB and 512 GB m4 doesn't wait for a back-to-back 4 KB write operation in order to put two 4 KB chunks into one 8 KB page. Every random write operation automatically goes to a separate page. That's why the 512 GB m4 tops the chart, as it has more dies per channel to better leverage simultaneous write operations. And with half the number of dies per channel, the 256 GB m4 falls behind. However, the 128 GB m4 is like the 256 GB m4, because both have four dies per channel. This is where firmware comes into play. The artificially-capped firmware in the 128 GB m4 restricts performance in order to provide better product differentiation. If it wasn't capped, the 128 GB m4 should have a similar max response time to the 256 GB m4.
