Yeah thats absolutly right. Still doesn’t change the fact, that this won’t help with sequential speeds (since this is limited to flash and controller hardware anyway). Its laughable to see some reporters claiming 1.5GB/s speeds purely based on this 300% improvement.

This improvment will defenetly help with lowend controllers. Too bad nobody is going to implement on old hardware anyway. My old v4 (which is based on phison S5) could really use this >:)

Not sure of who’s got it wrong.

Normally a single write (a byte or a KB or whatever) means the SSD-controller reads an entire block that the data is stored in into a buffer, changes the data and then saves it to a NEW block (if it has been used before it must also be erased before it can be written to) and invalidates the old original block which TRIM will eventually take care of. It’s very wasteful in read/writes as well as copying entire blocks to change a few bytes or kilobytes here and there. Most SSD’s become fragmented (which isn’t a problem in itself since seek-times are practically out of the equation), but having to copy 2000 blocks and then rewrite them to clear-flagged areas when they change.. That’s very time consuming (relatively speaking and very inefficient.

Garbage collection is normally about changing flags on blocks that are no longer used to make them available for new writes (but they have to clear the blocks before new writes).

By reading and then writing the change back to the same block several time consuming steps are removed and write amplification and wear level is severely diminished. This obviously affects power usage (added bonus) since the SSD unit will only have to do half the job to reach the goal as current controllers with normal GC & TRIM.

It’s not really a bottleneck, since current controllers are capped with SATA6gBIT speeds anyway. And since we have TRIM, garbage collection is already pretty efficient. So this won’t help with seqiential speeds not even a little bit. Its a hard limit of controllers and flash that no middleware can fix.

What this will do (if ever implemented) its gonna reduce write amplification (less wear) and increase write speed, when there are little fresh blocks left (heavy workload). This will be especially apperent, when drive is nearly full.

Just expecting 3x improvement in sequential speeds is just plain naive and only indicates, that person expecting this doesn’t know much about controller and flash architecture.

It’s not fancy, it’s simple yet very clever by removing the single most time-consuming bottle-neck in current SSD-controllers. Whether it’s 300% in reality is an entirely different matter since current SATA-generations can’t handle that bandwidth. SATA 6Gbit maxes out at a theoretical 750MB/s (in reality it’s lower due to overhead from OS (file system among other things), hardware, drivers etc.