The last few replies seem to miss the mark in a few ways.
First, about SATA SSD's versus spinning disks: everyone who doesn't understand storage gets hung up on misleading bandwidth claims between the fastest of the spinning disks (250MB/sec for a current-gen 8-platter 16TB Seagate Exos drive for example) versus a SATA SSD (~550MB/sec for a generic drive from ten years ago.) If you truly understand the storage stack and how applications pull data, this bandwidth number is functionally meaningless. Instead, you need to understand details around random read access times if you want to talk about the application start and level load duration. Even the fastest access times of a spinning disk will be around ~5msec, predicated on a physical 15,000 RPM 2.5" drive with 4 msec of rotational latency plus another msec or so in controller and head placement latency. Compare this to even a cheap SATA SSD which will have access latencies in the hundreds of microseconds or less -- a full order of magnitude (in excess of 10x) faster than the spinning competitor.
So yes, even a SATA SSD can be 10x or even faster than the uber-fastest SATA spinning disks where it truly matters for reads, which may still be enough to be impactful for the load times of a properly storage-threaded game. I'll go back to something I've mentioned in this thread (or one of the half-dozen equivalent storage-complainer threads in the forum): the Windows kernel is capable of literally millions of IOPS, which would equate to dozens of gigabytes/sec in throughput, even on a single CPU core. At the commodity levels we're talking about, the limit isn't the kernel or controllers or drivers, it's intelligent application logic. This new DirectStorage opportunity provides a way for the application developers to "simply" leverage known-good APIs for accessing storage, versus writing their own code and apparently sucking at it.