|
Thanks for the quick replies makere, Klyith and repiv! I'll go for the Kingston then, appreciate the help.
|
# ? May 14, 2022 14:28 |
|
|
# ? May 27, 2024 02:53 |
|
Welp I’m twelve hours into chkdsk /r on a four month old Samsung 970 Pro after one block started sending CRC errors to Windows. System’s running fine, but Macrium Reflect’s backup errors out and Samsung Magician’s various health checks could recognize the bad block and couldn’t do anything with it. (No SMART errors declared either) Genuinely wondering if it’s just Windows being stupid after some updates the other day. Love this, I totally don’t have things I need to do on my computer today.
|
# ? May 15, 2022 00:18 |
|
see if crystaldiskinfo has a better diagnosis for you. doesn't need to run a long analysis either. it was also the only thing i could use to identify the weird behaviour of my drives
|
# ? May 15, 2022 00:32 |
|
Arivia posted:Welp I’m twelve hours into chkdsk /r on a four month old Samsung 970 Pro after one block started sending CRC errors to Windows. System’s running fine, but Macrium Reflect’s backup errors out and Samsung Magician’s various health checks could recognize the bad block and couldn’t do anything with it. (No SMART errors declared either) Genuinely wondering if it’s just Windows being stupid after some updates the other day. CRC errors on drives can also be caused by a cable, but that wouldn't apply to a NVME disk.
|
# ? May 18, 2022 14:14 |
|
future ghost posted:I had to RMA a 970 EVO that was doing something similar. Was getting Media and Integrity errors listed in SMART logs, and HDTune's error scan tab looked like a christmas tree with all the red block errors. Each time HDT found an error it created a corresponding disk error in event logs. Sent all of that to Samsung and they approved the RMA. Whatever was hosed up it was a Windows problem. Ended up restoring from backups, there were never any issues with the drive itself in SMART or a chkdsk afterwards. Somehow it just hosed its own filesystem up. Yay!
|
# ? May 18, 2022 14:16 |
|
Arivia posted:Whatever was hosed up it was a Windows problem. Ended up restoring from backups, there were never any issues with the drive itself in SMART or a chkdsk afterwards. Somehow it just hosed its own filesystem up. Yay! I think that drive actually hosed up. Windows does NOT do this bro.
|
# ? May 18, 2022 14:51 |
|
If a drive develops a bad block that had data on it, it will continue to be reported as bad by chkdsk until that data is either recovered or no longer relevant (ie the file is deleted or re-written). If the number of bad sectors reported by chkdsk stays the same, it's just the one bad sector over and over again. The drive's smart system, as reported by crystaldisk, will show just the one error related to the event. If that is just one sector the drive is still fine. The drive can of course reallocate the bad sector, but without the data that's pointless. So it just keeps saying "bad sector bro" when the OS tries to read it. Because that's bad data and maybe the user wants to do something about that. HDDs have some potential to recover a bad sector -- a read attempt during the correct phase of moon gets the data, the drive writes it to spare capacity and remaps the sector. On a SSD a dead block is dead. (Which is why SSDs come with lots of spate capacity and preemptively self-check for weak blocks.) Or at some point you overwrite the data on that bad sector. That's implicitly saying that you no longer care about whatever was in that bad sector, so the drive writes it to spare space and remaps. At that point the next chkdsk will show zero bad sectors, but smart will still show 1 reallocated sector. Klyith fucked around with this message at 16:46 on May 18, 2022 |
# ? May 18, 2022 16:42 |
|
Here's the smart info for the drive that had an error reported. After I restored from backup I ran the integrity check in Samsung Magician and it reported no errors when it did previously. I haven't bothered doing chkdsk again, but I could if you think it's really necessary.
|
# ? May 18, 2022 18:16 |
|
Arivia posted:Here's the smart info for the drive that had an error reported. After I restored from backup I ran the integrity check in Samsung Magician and it reported no errors when it did previously. I haven't bothered doing chkdsk again, but I could if you think it's really necessary. So what I would do is every week or two, run a chkdsk /r followed by checking crystaldisk again. If you have more bad sectors in chkdsk, or if the Media/Data errors field increases a bunch, then I would worry a bit more. (OTOH not every "Media and Data Integrity Error" is a data loss error, and it's up to the manufacturer as to how exactly that is used -- there are drives that have huge numbers of those with no apparent problem. But NVMe drives also keep an internal error log on the drive itself. So at that point I would boot up a linux USB stick and use the nvme-cli package, which can read and show the log in a human-readable output. For ex my NVMe drive has nothing but "Invalid Command" errors logged, no problem. If your error log is full of unrecoverable reads or other bad stuff, that's a problem.) And keep doing that for several months until I was satisfied that this was a one-off.
|
# ? May 18, 2022 19:30 |
|
For nvme-cli, sudo nvme smart-log /dev/nvme0 this assumes you only have 1 nvme drive on the host, otherwise it might be nvme1 or 2 etc. Look for media_errors field. quote:Media and Data Integrity Errors: Contains the number of occurrences where the controller detected an unrecovered data integrity error. Errors such as uncorrectable ECC, CRC checksum failure, or LBA tag mismatch are included in this field
|
# ? May 19, 2022 01:23 |
|
FWIW that CDI log looks like mine did. You may also have corresponding event log errors pointing at disk as the cause. Personally I'd RMA it with Samsung.
|
# ? May 19, 2022 04:48 |
|
future ghost posted:FWIW that CDI log looks like mine did. You may also have corresponding event log errors pointing at disk as the cause. Personally I'd RMA it with Samsung. Me too. I had a very similar thing happen with a 850 Pro. RMAd it.
|
# ? May 19, 2022 14:52 |
|
future ghost posted:I had to RMA a 970 EVO that was doing something similar. Was getting Media and Integrity errors listed in SMART logs, and HDTune's error scan tab looked like a christmas tree with all the red block errors. Each time HDT found an error it created a corresponding disk error in event logs. Sent all of that to Samsung and they approved the RMA. I've got a very similar problem with a 980 Pro, already having RMA'd it once and the problem came back. There was likely a bad motherboard involved too that wasn't discovered til after the RMA, but I am not looking forward to dealing with this again.
|
# ? May 19, 2022 18:38 |
|
Aquila posted:I've got a very similar problem with a 980 Pro, already having RMA'd it once and the problem came back. There was likely a bad motherboard involved too that wasn't discovered til after the RMA, but I am not looking forward to dealing with this again. future ghost fucked around with this message at 21:39 on May 20, 2022 |
# ? May 20, 2022 21:25 |
|
makere posted:I would go for the Kingston unless you're made out of money. The reason why Kingston was put on the poo poo list was because they would swap internals (nand/controller) without changing product name, but pretty much all manufacturers do that now. Yeah, I don't think this is true at all. Is there any evidence of WD or Samsung doing this?
|
# ? May 22, 2022 21:25 |
|
WD did it on the SN550 https://www.tomshardware.com/uk/news/wd-blue-sn550-ssd-performance-cut-in-half-slc-runs-out Samsung did it on the 970 EVO PLUS https://arstechnica.com/gadgets/2021/08/samsung-seemingly-caught-swapping-components-in-its-970-evo-plus-ssds/
|
# ? May 22, 2022 21:29 |
|
Whelp, that totally sucks.
|
# ? May 22, 2022 21:33 |
|
WD's swap on the SN550 was bad, but the worst example was the Crucial P2. They changed that to QLC with far worse performance, and less than a year after introduction and giving reviewers the good version. The 970 Evo is IMO not that a big deal: the write speed after SLC exhaustion was slower, but in trade you get a bigger SLC cache to begin with. Plus better read performance in general. Personally I'd rather have the new model. In an ideal world the companies would announce this type of thing even when the swap is defensible as being the same product. But in the world as it, Samsung is doing only a minor offense.
|
# ? May 23, 2022 00:25 |
|
In an ideal world those would just be new models.
|
# ? May 23, 2022 06:30 |
|
I’ve now got an old Samsung 950 Pro nvme drive kicking around. Can I use this as like a really fast portable drive? Any enclosure recommendations? Preferably USB-C for connecting to macs without having to use a dongle. Edit: something like this? http://sabrent.com/products/ec-snve Sabrent are meant to be a decent brand right? track day bro! fucked around with this message at 20:45 on Jun 6, 2022 |
# ? Jun 6, 2022 20:39 |
|
track day bro! posted:I’ve now got an old Samsung 950 Pro nvme drive kicking around. Can I use this as like a really fast portable drive? Any enclosure recommendations? Preferably USB-C for connecting to macs without having to use a dongle. Sure. It's vastly overqualified for the job but if it's a small size drive it probably doesn't have much resale value. Sabrent are fine.
|
# ? Jun 6, 2022 22:22 |
|
Really surprised the controller (RTL9210B-CG) on that sabrent does Gen3 x2 to USB 3.1 Gen2, that's pretty neat. I'm not having any luck finding the device for sale, I'm curious how much they cost. Gotta go direct to realtek though it looks like.
|
# ? Jun 6, 2022 22:37 |
|
Was doing some computer janitoring and checked in on my SSD's. They're all fine as far as SMART can tell, but I started idly wondering about how far you should trust rated endurance. It's only even remotely possibly relevant on my OS drive, a NVMe 500GB Samsung 960 EVO from late 2017. It's apparently done 55TB of writes out of a rated endurance of "up to" 200TB written, and has a bit over 7000 power-on hours. I have backups of course, but since it's the OS drive I'd rather have a planned replacement than an unplanned one where I have to be without my computer for a while. If I can count on half the rated endurance then I'll probably want to upgrade the entire computer before I hit that, which would be a good time to replace it anyway. I realize there's probably no statistics for this, just curious if anyone's had SSD's die on them mid-life.
TheFluff fucked around with this message at 03:47 on Jun 16, 2022 |
# ? Jun 16, 2022 03:42 |
|
TheFluff posted:but I started idly wondering about how far you should trust rated endurance. The few times anyone has tested this the SSDs have exceeded their rated endurance, often by many multiples. TheFluff posted:I realize there's probably no statistics for this, just curious if anyone's had SSD's die on them mid-life. SSDs definitely die from other things besides flash endurance. For the average or enthusiast users, probably 99% of dead SSDs are from non-endurance causes. Very few people write enough data to do that in reasonable time.
|
# ? Jun 16, 2022 04:26 |
|
Is it normal for SSD to SSD copying to be highly variable when going over a certain volume? Here's my setup: Asus TUF GAMING Z690-PLUS WIFI D4 ATX LGA1700 Motherboard Western Digital Black SN850 2 TB M.2-2280 NVME System drive ADATA XPG SX8100 4 TB M.2-2280 NVME Media drive When moving 20GB from the WD to the ADATA, I'll get a burst of multiple gigs a second that drops off sharply to 500MB/sec. I figured that's the SLC filling up and the drive having to write the rest out directly to multi-level flash. But when I copied the same load back immediately, it started at 500MB/sec then shot up to ~1.7GB/sec near the end...? I just want to make sure I don't have some kind of bus bandwidth issue going on.
|
# ? Jun 21, 2022 12:51 |
|
Shumagorath posted:Is it normal for SSD to SSD copying to be highly variable when going over a certain volume? Here's my setup: Copying can be highly variable for a lot of reasons. 1. Copying many small files is slower than few large files (due to filesystems, not the drive itself). If this was 20GB of random stuff, you'd expect ups and downs as it copied files of different sizes. 2. If you're on windows, defender can be a bottleneck for SSD-to-SSD copies. And it's not gonna be consistent at all. So for any sort of benchmark or test you need to turn off defender temporarily. Shumagorath posted:When moving 20GB from the WD to the ADATA, I'll get a burst of multiple gigs a second that drops off sharply to 500MB/sec. I figured that's the SLC filling up and the drive having to write the rest out directly to multi-level flash. But when I copied the same load back immediately, it started at 500MB/sec then shot up to ~1.7GB/sec near the end...? There are pretty much zero hard specs for the SX8100. Also apparently the 4TB models of that drive may be QLC. So if Adata did the same thing as WD did to the SN550 -- down-converted a MLC design to QLC without increasing the cache or doing the variable cache size trick -- then yeah that seems like SLC caching. Particularly if you were copying something like a single 20GB video file, which would have the least problems from 1 & 2 above. Means you have a disgustingly small SLC cache, but it's a cheap adata drive so that's what you paid for.
|
# ? Jun 21, 2022 14:14 |
|
The reported free space in Windows has also been bouncing around and I'm not whipping that many files back and forth. I'm questioning if it's worth just wiping and refunding but the next-closest 4TB is still $300 away and I'm likely beyond the return window. Oh well; at least I have good backups if this thing is a lemon.
|
# ? Jun 22, 2022 03:42 |
|
Shumagorath posted:The reported free space in Windows has also been bouncing around and I'm not whipping that many files back and forth. I'm questioning if it's worth just wiping and refunding but the next-closest 4TB is still $300 away and I'm likely beyond the return window. Any free space weirdness is gonna be a windows thing, not a drive thing. Drives don't AFAIK change their reported size based current conditions like SLC cache or even internally-remapped bad sectors. And knowing used vs free space is the filesystem, not the drive. But yeah 4TB is inflated price territory for NVMe and that adata is just about the only thing I'd call "worth the money" at that size... and even that's a stretch. There's also the Inland Platinum at $470 which is slightly better -- still QLC but with a phision e12s controller, I believe that one has expanding SLC cache. I dunno though, $70 just for better write speed might be ignorable on a data-storage drive. (4TB on a sata drive OTOH is relatively affordable. You can save a lot of money by being ok with 2 wires.)
|
# ? Jun 22, 2022 14:51 |
|
My build went from zero spinning drives to zero cabled drives and I'm sticking to it. Optical is in a caddy on the desk
|
# ? Jun 23, 2022 03:53 |
|
Shumagorath posted:My build went from zero spinning drives to zero cabled drives and I'm sticking to it. I'll give my Bluray burner another decade and if I'm not using it, for sure I'll take it out.
|
# ? Jun 23, 2022 04:02 |
|
I used my change to a Meshify to abandon my 5.25” drives… I used my optical drive maybe twice in the past five years. Bought a USB model and put it in my desk drawer, should I ever need one…
|
# ? Jun 23, 2022 16:06 |
|
movax posted:I used my change to a Meshify to abandon my 5.25” drives… I used my optical drive maybe twice in the past five years. Bought a USB model and put it in my desk drawer, should I ever need one… My optical drives just started working again with my new board, I put too many nvme drives in my old build and the SATA controller they were plugged into turned off due to lack of lanes, haha. I did actually use it, I needed a network driver for the board from the cd because I guess z690 is still so new.
|
# ? Jun 23, 2022 19:21 |
|
Kinda SSD related question - If someone has created a SATA SSD containing a bootable Linux, can they attach an NVME drive to the host and clone it over and have it boot or is there any additional boot (MBR etc) stuff required to get it to boot when going SATA -> NVMe? Just came up at work and I didn't know if it would work or not, needed to order some parts to attach the NVMe drive to the system so can't try it out right now..
|
# ? Jun 24, 2022 00:01 |
|
priznat posted:If someone has created a SATA SSD containing a bootable Linux, can they attach an NVME drive to the host and clone it over and have it boot or is there any additional boot (MBR etc) stuff required to get it to boot when going SATA -> NVMe? Yes, if it's not an ancient PC or an ancient linux. Modern grub and fstab can target using UUIDs so that's how they'll generally be set up by the OS installer.
|
# ? Jun 24, 2022 01:21 |
|
Klyith posted:Yes, if it's not an ancient PC or an ancient linux. Modern grub and fstab can target using UUIDs so that's how they'll generally be set up by the OS installer. Nice, thanks! The AMD PDKs don’t even have SATA ports anymore (since Rome I believe) as my coworker discovered when he went to make a boot drive with a SATA loaded with CentOS
|
# ? Jun 24, 2022 01:25 |
|
movax posted:I used my change to a Meshify to abandon my 5.25” drives… I used my optical drive maybe twice in the past five years. Bought a USB model and put it in my desk drawer, should I ever need one… I have an external USB DVD burner and it's a laptop style one which I don't think is great. I might invest in a decent USB-C enclosure and put a good quality burner in there. I only use it for ripping CDs (I'm a bit Old Man Yells At Cloud when it comes to streaming music) and the compact laptop drives aren't great for that.
|
# ? Jun 24, 2022 23:34 |
|
Thanks Ants posted:I have an external USB DVD burner and it's a laptop style one which I don't think is great. I might invest in a decent USB-C enclosure and put a good quality burner in there. I only use it for ripping CDs (I'm a bit Old Man Yells At Cloud when it comes to streaming music) and the compact laptop drives aren't great for that. IMO, the ‘best’ (i.e., least amount of SW layers) is probably Thunderbolt so you can effectively get an external PCIe <-> SATA thing going… or if eSATA ports / add-in cards are still a thing. I like trying to get everything as native as possible, but I also generally distrust USB / try not to use it for anything outside of keyboard / mouse / webcam.
|
# ? Jun 27, 2022 08:01 |
|
movax posted:IMO, the ‘best’ (i.e., least amount of SW layers) is probably Thunderbolt so you can effectively get an external PCIe <-> SATA thing going… or if eSATA ports / add-in cards are still a thing. I like trying to get everything as native as possible, but I also generally distrust USB / try not to use it for anything outside of keyboard / mouse / webcam.
|
# ? Jun 27, 2022 14:19 |
|
movax posted:IMO, the ‘best’ (i.e., least amount of SW layers) is probably Thunderbolt so you can effectively get an external PCIe <-> SATA thing going… or if eSATA ports / add-in cards are still a thing. I like trying to get everything as native as possible, but I also generally distrust USB / try not to use it for anything outside of keyboard / mouse / webcam. A PCIe to SATA controller chip is gonna have a driver just like a USB to SATA controller does? UASP solves the whole CPU hogging problem of USB mass storage. External HD docks consuming tons of CPU wasn't because they had too many software layers. It was because the old protocol was in-order, unqueued, and I think had limited DMA ability. The overhead was because it was inefficient and bad, not because software layers were doing anything. Imagine you somehow hooked a fast SSD up to an old parallel ATA interface, and set it to run in classic ATA mode rather than UDMA. But then you somehow overclocked the ATA bus to run at 5gbps. It would generate tons of CPU overhead too. (But also, in the context of a DVD burner none of this matters because optical drives are lmao slow.)
|
# ? Jun 27, 2022 14:22 |
|
|
# ? May 27, 2024 02:53 |
|
I have a 2tb Inland m.2 that looks like its dying or dead. It is a secondary storage i use for games and backups of pbotos. I was playing games when my cpu crashed. It then cycled through crashes a few times and was slow to boot up. When if finally did boot, this secondary drive wasnt accessable. Windows shows it in the disk management (though i cant do anything), but crystaldisk doesnt show it at all. Is it just the drive failing? Or more? Would a spare drive cause instability on boot up as it fails?
|
# ? Jun 30, 2022 08:57 |