Performance Focus – Toshiba RC100 480GB

I'm sticking with the 'burst vs. saturated' plots here, as they do well to show both sustained performance and the more realistic (for client PC usage) burst throughputs.

Sequential

Random

Burst and sustained reads for both sequential and random are nearly identical in the above two charts, which is expected. Sustained random write performance also nearly matched burst performance, but that might have been due to the cached performance not being as high as it should have been (leveling off at ~30,000 falls far short of the rated 110,000 IOPS there). Burst sequential writes see an SLC cache speed of nearly 900 MB/s, with a sustained speed closer to 400 MB/s.

Burst Random %Read Sweep

While QD1-QD2 performance did ok here, we did note that at higher queue depths, we see a dip in throughput in the 10%-40% region of the above chart. Not a huge deal as most client workloads occur in the 70% and higher region.

Caching

Caching performance was inconsistent when it came to intermittent 60 second writes with varying idles. In some passes, the RC100 took a few seconds before seeing SLC speeds kick in, but if given enough idle time before the pass, we saw over double the data rate for 15-20GB worth of sequential writing.

« PreviousNext »