LOSTCIRCUITS

Of all "basic" HDD performance benchmarks, SiSoft Sandra still appears the most robust and we used it to compare the DiamondMax Plus9 with and without MaxBoost enabled. The raw, composite score is of little interest here, rather, it is the cached or "buffered" writes that will be redirected to the system memory instead of having to be funnelled through the SATA interface to the cache. In other words, what happens is really a memory copy from one memory area to another but there will also be an overhead that will cap the transfer rate somewhere. The main result is that the buffered writes increase from 60 MB/sec to 1566 MB/sec. An interesting side effect of enabling Maxboost is a small overall performance hit in sequential transfers as well as in cached reads. This appears logical as one additional step is interposed in the data path.

ATTO uses different test file algorithms than SiSoft and appears to read from the "RAM disk" as well as write to it. We deliberately set the test file size to 32 MB to force a buffer overflow, that is, to fill up the allocated memory space. As a result, we see a marginally increased write performance. The slight improvements may be caused by utilization of the "breathing room" between the different test patterns, which are used to empty out part of the cache and, thus, result in a minor boost. If we decrease the test file to 16 MB, the I/O performance is the limiting factor at the smaller block sizes but the "RAM Disk" is working much more efficiently for the highest read / write performance in the order of 1 GB/sec.