Originally Posted by Sir Beregond
Wow I am apparently so far out of the current knowledge of SSDs. My mind is a bit boggled to hear that certain modern SSDs have a slower write than a hard drive.
Guess I need to do some reading. All we really had when I built my last rig was SATA 3 2.5" drives.
Sent from my SM-G965U using Tapatalk
QLC is usable - just - if you realise its limitations and don't use it in any scenario where you might have sustained writes (>50GB) or sustained combined read/writes or want to keep the drive 90% full. But at the prices being demanded for them currently? Avoid like the plague because better drives using TLC can be had for similar or cheaper prices.
I was early on the QLC train, because they were about 30% cheaper than the TLC drives at the time where I was shopping and they were so new I optimistically thought, "well, how bad can they be, really
Answer: pretty freakin' bad. I have seen sustained writes in the 90MB/s range with nothing else happening on the drive (2TB Samsung 860QVO), and the 65-70MB/s range if doing sustained read/writes at the same time. I will admit this is something of a specialist edge case, but nonetheless, it's rather troublesome.
Lesson learned, I refuse to buy them again, but if
the price is right, they can be acceptable depending on your usage scenario. I wouldn't use them as an OS drive, I wouldn't use them as a scratch drive, but I might think about using one as a game drive if the price was right
. By that I mean about half the price of any other drive.
Originally Posted by Liranan
The 3TB DT01ACA300's I have in my server (XFS) have a sustained read/write of about 180MB/s, while QLC has higher access times that's not enough for me to say that these drives are acceptable, they're horrendous.
I don't see where in what you quoted I said they were acceptable?
I did say, however, that you can cherry pick a HDD to make even a QLC drive look good: just go with an SMR drive.
Seagate did a sneaky and "upgraded" their 8TB Barracuda Compute HDDs to SMR last year, and obfuscated the drive tech in their documentation in the process so it was hard to tell without actually having a drive to test. End result? I have four lemon drives where if I write more than about 10GB to them at once, they will choke and write at something crazy like <1MB/s. That is absolutely unacceptable, and even QLC is better than that.