I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
1–10 of 261 posts
Re: I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
#2Well, yes, but which were those 2 out of 4 vendors?
Re: I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
#3Name the offenders please.
I am sure it might be easy to see visually - a lack of substantial capacitor on the board would indicate a high likelihood of data loss.
Re: I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
#4Name the offenders please. I am sure it might be easy to see visually - a lack of substantial capacitor on the board would indicate a high likelihood of data loss.
SK Hynix and Sabrent: https://x.com/xenadu02/status/1496290369184874497?s=20
Re: I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
#5Well, yes, but which were those 2 out of 4 vendors ?
SK Hynix and Sabrent: https://x.com/xenadu02/status/1496290369184874497?s=20
Re: I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
#6Re: I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
#7Earlier quoted context omitted.
SK Hynix and Sabrent: https://x.com/xenadu02/status/1496290369184874497?s=20
Which did "pass" the test?
Re: I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
#8Re: I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
#9Name the offenders please. I am sure it might be easy to see visually - a lack of substantial capacitor on the board would indicate a high likelihood of data loss.
I'd be curious how well it actually correlates. It would be hard to make the most performant system that's always consistent with flushed data but there are probably a lot of firmwares out there with untested performance ideas, etc.
Re: I tested four NVMe SSDs from four vendors – half lose FLUSH'd data on power loss (2022)
#10I guess it's time for `fsync_but_really_actually_sync_it_please(2)` (and the lower level equivalents in SATA, NVMe etc.)?