just brew it! wrote:Something is definitely fishy with the HD Tune results people have been posting. The reported Burst Rate is sometimes lower than the Minimum and Average rates, which should be impossible?
Both HDTune and HDTach are legacy spinning disk programs that don't really have updates for the modern SSD era. You should probably be worrying more about the -1% CPU usage
The exact nature of the details is not really that relevant - what matters is that in real-world usage, users are reporting awful performance; Pre-Diskfresh (or Samsung's new firmware doing the same thing), the HDTach/HDTune benchmarks are all over the shot with some staggeringly low numbers to back up the lousy real-world performance that users are whining about. After the Diskfresh or firmware-based refresh, real-world performance is good and the HDTach/HDTune benchmarks are a nice straight line again.
All these old benchmark tools do is pick a random series of blocks at that position on the drive and measure how fast the read of those blocks is. With SSD's it all over the place because the physical location of the blocks no longer corresponds to a single physical location - it's up the SSD controller where in the NAND cells the virtual block pointers are pointed at.
The graph plot tools are just secondary evidence to support awful read speeds on these defective drives. The concern is not with the exact output of the tools, just that the tools are an easy visual representation of an obvious problem that goes away after a Diskfresh. Straight line = good, drive is working within spec. Erratic line = unmistakable performance inconsistency to back up user claims of slow SSDs.