Quote:
Originally Posted by MickMcGeough
You probably haven't written to the entire thing yet, then. I can't see how reading the block, storing the block in memory, erasing the dirty cells, modify the block in memory, then write the block back to the cells can be as fast as writing directly to the cells under any circumstance.
|
It’s also possible that OS X doesn’t produce enough native throughput to expose the write cycle as a bottleneck.
The drive itself could also maintain a rotating set of cells that it pre-trims in the background and re-maps on demand. This would eliminate the up front need to perform the erase cycle on demand. I’m not sure if any drives do this, but it seems reasonable to assume this is a trick that drive controller manufacturers have already exploited to improve general write performance.