Interesting....I ran CyrstalDiskMark on my drive at 9% full and again at 93% full, which I would consider to be 'emptyish' and both 'fullish'.....results are similar:
Note: I didn't run until I got 'golden figures'....just ran it off the bat.
good to see user's real world results. but in itself, as you say, interesting...lol
just having a quick gander at the results...the sequential queue depth 32 results are expected of a nvme drive.
but the 4kb queue depth 1 results are vastly different to anandtech's:
vs 47-60mb/s and 146-148 mb/s on yours, respectively.
making some long-drawn assumptions (yeah ass-u-me, i know)...it may be possible that anandtech didnt report on empty/full results on synthetic tests as these don't show up any difference, and it's only on their simulated workload tests that the ssd's controller/nand start to show up their deficiencies.
i guess one can say it's like having a 1/4 mile drag race vs a gran prix...lol...who knows...
To be fair, I think in the real world would we even notice a difference between the two running at optimum speed? Depending on the use case I suppose.
negligible at best. zero difference at worst.
even going from sata 3 to nvme yields very little real-world difference...unless hammering the storage subsystem, in which case then the mp510 makes the most sensible choice as it has vastly superior write endurance - 1700tb (mp510) vs 650tb (sx8200p)