oh, you were refering to the intel drive SPECIFICALLY not to SSD tech in general..
there was a bug in the intel firmware that caused SOME drives to have severely degraded performance (which occurred after they were exposed to some rare writing situation)... so the new firmware fixes that bug.
While steady state is involved its not the cause, the cause is controller bug. So its not that the drives will "degrade forever", but that if you have and older firmware intel drive it will degrade to steady state (where it is still fastest on the market), at which point it stops degrading unless it gets into this weird write situation and gets all messed up.
"It should be noted that any of our SSDs will see periods of reduced performance after significant random write fragmentation (white noise random fragmentation, not what Windows generally does with it?s ?random? writes) as the drive cleans this all up mixed with additional data being written. This new firmware does not change this fact. What it does do is prevent the drive from getting into a state where further sequential writing will not recover the drive. You should see this with HDTach, or a large file copy, or just general use, etc. So, if a drive is in what previously seemed to be a permanently degraded state (as discussed, we still feel this is highly unlikely for a client PC user), and a user installs the new firmware they will feel an instant improvement for any sequential operations, which will get better in time as the drive cleans itself up further. This firmware will also prevent the user from getting into such a drastic state of fragmentation, and generally help ensure the sequential write performance is as good as it can be at any moment. This change really has no significant impact on random performance."
This "white noise" writing caused the firmware to crap out pretty much.
steadystate is an approximate figure, its not an EXACT figure of write speed, steady state should no exist in a controller with trim, in a controller without it SHOULD remain around a certain figure, getting worse and then GETTING BETTER as you write, that is because there is no physical degradation of the drive involved.
The problem is that you can write to individual 4kb blocks, but can only erase them in groups of 128. So if you want to write 4kb to the first group, you have to read 127 blocks, erase all 128, then write back those 127 + the 1 you wanted to write... need to make another write to same group? repeat.
the reason steadystate fluctuates is that sometimes you tell it "overwrite 16 blocks" that happen to be on the same group of 128 that erase together. weather they are on the same group or are on different groups makes a difference and depends on the kind of data you have written to it before.
The controller doesn't KNOW that 60 out of the 128 blocks in that group contain data that was erased... to save time on a spindle drive, the OS never tells the drive to erase something, it just says "change block # from containing data Y to data Z"
Trim will actually TELL the drive "user deleted blocks #, # and #" and the controller can choose to erase them NOW leaving them ready for a full speed write, or just keep that data for later and keep X% of the drive "clean" or whatever it is that they choose to do. (varies by firmware)