If you really want to compare to nMP it should probably be a PCIe based SSD with comparable speeds.
Really this is only true if your workflow will notice the difference.
All of the good 6 Gbps SATA SSDs are in the 500 MB/sec to 550 MB/sec range for large streaming sequential reads and writes..
The Samsung PCIe SSD disk in the Apple Tube is roughly twice as fast. If you read and write a 100 MB PSD file it will take the SATA SSD disk 400msec, and the PCIe SSD disk 200msec. You could take a vacation on the time savings
.
For smaller IO sizes the PCIe disk has less of an advantage, since there are per transfer latencies in the disk and in the OS. The smaller the payload, the more the per-transfer costs dominate the overall time.
The only clear advantage to SSD over spinners is that SSDs never have any rotational or head movement latency. Both SATA SSDs and PCIe SSDs are far better than spinners in that regard. But, if something is 20 times faster is it worth it to pay twice as much to get to 24 times faster?
I would expect to see that examining many meaningful workstation applications and workflows, both SATA SSDs and PCIe SSDs are far faster than spinners. I would also expect to see that SATA SSDs and PCIe SSDs are much closer together in actual workflow performance than you'd expect from the simplistic sequential benchmarks.
If the real world difference is seconds per day, there's no reason to needlessly bump the price by using components that won't pay for themselves.
(And to show that I seriously look at applications and performance, today I ordered $22K of Intel 910 PCIe SSDs for one of our applications. They were absolutely the best $/performance addon for the app.)
