General NVMe Drive Problems (Fatal)

[c-o-pr]

This post is an update to warn anyone trying to use Samsung 980 Pro that there may be serious bugs with it and macOS that can cause both localized and total data loss.

I'll summarize my experience as a few bullet points:
• Long Boot Trim delays even if SetApfsTrimTimeout=0
• Trim stalls occur when Parallels VM performs Trim, such as when resizing a virtual drive image.
• Drive access occur at arbitrary times.
• Deleteing APFS snapshot causes HW faults.
• When attached to a Sabrent USB EC-SNVE drive, some effect of Monterrey system installed on the external enclosure will cause macOS to hang for long periods when accessing the drive, freezing any part of the system that touches the drive and necessitating an forced drive removal to regain control of the system.
• Running a standard CCC incremental to the same USB-attached will run slowly and eventually cause disconnect and a corruption of the APFS container for the external system and loss of data.
• Once corrupted, the drive will be in a state when Monterey cannot access the drive, and Disk Utility will hang at startup. By taking to the drive to an older system running Catalina, it can be accessed but not repaired. However it can be erased. Taking the erased drive back to Monterey, the drive is fully access and a full system clone can be made with the drive as a target, and will again work normally.
• Samsung Magician reports 9TB of writes used out of 1200TB and no hard errors. Most of those writes have occured because of boot Trim cycles, where it appears that Trim is issued and processed for the full extent of the unused portion of the drive. I can't say for sure but no other part of the system is generating a T of writes between boots.

Caveat emptor: My warning is just a list of effects. I don't know why any of this happens, but I have gone through these problems repeatedly over several months and these are the patterns I see.

Especially notable is that something goes wrong when putting a working Monterey on the 980 Pro into the external enclosure. I've considered it could be a power draw issue, with USB being able to supply a nominal 4.5 watts and the drive reported to use 4–6 watts peak. But I can use the same drive, port and enclosure to write at 600 MB/s sustained for 20 minutes and end up with a working copy of the drive, then put it in the board NVMe and it runs (notwithstanding other issues). Later I take the same drive out and put it in the external enclosure and it goes bad as described.

I have two of these enclosures, have previously updated their firmware to latest version (2021) and have used them successfully with Samsung 970 Plus, WD SN750. I had bad problems with Sabrent Rocket 4 in 2021, and these Sabrent enclosures were also used, but I don't knw what to think if Sabrent is incompatible with Sabrent.

OTOH a WD SN750 has always worked flawlessly with this same build and external enclosure.

Last week I took a gamble and bought an SK Hynix P41 2T (gen4) copied my full install to it without issues, installed, and been running it for several days, perfectly smooth as I hoped. I''m writing this post using it.

My advice is that 980 Pro should be completely off the table for any hack user because you don't know when it's gonna blow and if it does, you stand to lose everything.

HTH and anyone with other experiences please report.

 

[ramazarusx]

Ventura does the same thing? How's SN850?

 

[c-o-pr]

I went for new gen4 SK Hynix P41 instead of SN850 and it's so far been great: zero isuues, top notch perf, my kit has never felt smoother. I will be careful to not recommend because I don't have enough experience / knowledge to say it's gold, but after a week of daily use everything feels perfect.

 

[dredhorse]

does somebody know why OC 0.7.8 doesn't see my EVO970Plus? my OC 0.6.9 version does see it (and it is working as I'm just typing this with catalina booted from it)

 

[c-o-pr]

does somebody know why OC 0.7.8 doesn't see my EVO970Plus? my OC 0.6.9 version does see it (and it is working as I'm just typing this with catalina booted from it)

Your report is a bit vague

• Are you moving devices around?
• M.2 slots often trade PCI lanes with full width slots and SATA controllers, so they can only be used either/or. If you are moving devices around you must consider lanes.
• Is the 970 detected by the BIOS? Is the slot reported as populated?

There's whether:
• the board sees the drive,
• the board BIOS (UEFI) sees a bootable EFI folder (OC or other),
• whether OC sees a loadable SW installation, eg macOS

If you can start OC from the 970 drive then it is seen, so to speak.

It's too easy to get confused about which drive is booted and which EFI is running OC and chainloading where OC loads another loader which may or may not be able to load macOS

I have never heard of a case where OC doesn't find a macOS installation based on drive model, so this idea is a surprise. Just a version update of OC should never cause such a situation — afaik. But OC is it's own little OS with zillions of config options...

 

[maxxx]

does somebody know why OC 0.7.8 doesn't see my EVO970Plus? my OC 0.6.9 version does see it (and it is working as I'm just typing this with catalina booted from it)

have put this drivers on? openhfsplus.efi.

 

[dredhorse]

Your report is a bit vague

Some more detail as I have progressed further:
3 drives: main macOS (samsung evo970 plus), Windows, backup macOS
both macOS drives have an EFI, if I use OC 0.6.9 (main) I see all 3 drives, if I use OC 0.7.8 (backup) I did only see the backup macOS after upgrading that to monetery.
BUT... as I said I progressed further.
Deleted all partitions on backup macOS, recopied man macOs over again.
Able to boot from main EFI into both macOS versions
booting from the backup EFI shows me.... WINDOWS.
reset nvram, replaced efi folder with original 0.7.8 version... no change.
So it looks like OC 0.7.8 has some issues with finding my drives for whatever reason or better said the macOS on it.

 

[c-o-pr]

Please help others help you with additional clarity:

3 drives: main macOS (samsung evo970 plus), Windows, backup macOS

Are all three drives in the system at once?

both macOS drives have an EFI,

You are choosing the desired drive EFI via the BIOS boot picker (F8)

if I use OC 0.6.9 (main) I see all 3 drives,

IOW all three installations appear in the OC picker?

if I use OC 0.7.8 (backup) I did only see the backup macOS after upgrading that to monetery.

Here is where the language gets hard to follow: what does "use" mean in context?

To use... You F8 into OC 0.7.8 on Backup and, if I read this right, no macOS install detected on Backup. Then at some point, somehow maybe via OC 0.6.9 on another drive, you upgraded macOS on Backup after which OC 0.7.8 can find it and it boots.

Are you both running different EFIs and swapping out the contents of EFIs at once? I ask because I caused confusion for myself by not keeping careful track of such convoluted configs. I resolved an approach where I keep a gold config on my main drive and try out alternates on a thumb drive to stay sane. I carefully hover over F8 when booting because when drives are changed out my BIOS reorders the default boot list.

—I'm not saying you are confused, but the combinations you have arranged could lead to confusion add your writeup is not clear.

Plus at this point you are changing out macOS versions too so a lot is going on!

BUT... as I said I progressed further.
Deleted all partitions on backup macOS, recopied man macOs over again.

OK— Here I am struggling and lost because it sounds like you are running versions of macOS that span pre-APFS mac to signed-system-volume mac.

You can't just delete partitions and recopy over macOS to clean parts since middle of Big Sur, unless my "recopy" you mean wipe the drive, run the macOS installer, migrate and add an OC EFI...

So many questions about your setup and evolution...

As a previous poster mentioned, if you have an older macOS that's on HFS+ make sure you have the HFS+ driver correctly installed in your OC 0.7.8 EFI.

Rather than trying to get into your head, I'll suggest we narrow the possible configs down to the desired target installation and getting the desired OC to see it. Set aside the other drives, Windows etc which are potential red herrings, and focus on the

Able to boot from main EFI into both macOS versions
booting from the backup EFI shows me.... WINDOWS.

Here the language has switched up to "main" vs "backup" EFI: you're expecting the reader to keep track of what config is where after you made points that you've been "progressing" your configs through major charges.

If I follow this from OP, I'm not sure if you upgraded macOS and found an OC EFI stopped working for you, or if you changed OC to a later version and found it doesn't detect a macOS install or both?

reset nvram, replaced efi folder with original 0.7.8 version... no change.
So it looks like OC 0.7.8 has some issues with finding my drives for whatever reason or better said the macOS on it.

As these forums attest, zillions of people are using many versions of OC to boot many versions of macOS on both hacks and real Macs, so whatever you are facing has got to be a local config problem.

+++Not to be rude, but you probably have not discovered a latent defect of OC 0.7.x!

If you've never really addressed your OC habitat, I hate to break it to you but it's time to take stock of this chore. There are many writeups in the forums.

Consider just moving on to the latest OC version. Your previous config should work (if it's basically correct) after you account for config.plist schema changes that are a normal consequence of updating OC. You'll want to review / update the companion kexts too, Lilu, WEG, etc because these have had to evolve to support later versions of macOS and you seem to be trying to move forward to latest.

As to the particular problem of having an older version of OC that works and a newer that doesn't, it's not likely to be a general NVMe problem.

You'll get more traction on a solution by posting on one of the other forums here pertaining to your class of HW, simplifying your boot config for testing purposes, and narrowing down the variables in your writeup to make it easier for readers to follow. And as you do this, you may solve the issue along the way.

 

[dredhorse]

hi, thanks for helping out.. to clarify it more but also to close the discussion in this thread:

if I boot from a OC 0.7.8 EFI I only see my windows drive now (that OC 0.7.8 EFI was upgraded with OCAuxillary, I updated OC and all other drivers, kexts etc).
if I boot from a OC 0.6.9 I see all my 3 drives: main macOS (Catalina with AFS), Windows, backup copy of macOS (Catalina with AFS)

So something is wrong with OC 0.7.8 not finding my macOS drives (both of them work when booting from 0.6.9) I already did troubleshoot via the OC documentation and the config.plist should be correct.

what is weird is the order in which I lost my drives

- updated my backup EFI to OC 0.7.8 and rebooted from that EFI
- I did see 3 drives
- updated my backup macOS to Monterey
- after rebooting I only see Windows and Monterey from the OC 0.7.8
- after booting from my OC 0.6.9 I see all 3 drives
- deleted all partitions from my backup drive (the one with Monterey on)
- copied my main macOS (Catalina) again to the backup drive
- boot with the OC 0.7.8 and I only see Windows
- boot with the OC 0.6.9 and I see all 3 drives

So OC 0.7.8 and drivers etc does work (I could upgrade from Catalina to Monterey with it), but is is missing drives in the picker.

And a final note: I did see my missing main macOS in Monterey after booting into and could access the drive without any issue.

But IIRC I already created a different thread, let this one rest.. except if somebody has a good idea why the picker doesn't recognise my 2 macOS installs.

 

[c-o-pr]

hi, thanks for helping out.. to clarify...

- copied my main macOS (Catalina) again to the backup drive

This is the only point that stands out: Catalina was last macOS Apple let run from HFS+. If you are using HFS+ you need a HFS driver in OC or the volume won't be delected. A older EFI from that era will have this. A newer EFI might not have it installed (overlooked config?). There are two versions of OC HFS driver around, one that ships with OC and another called OpenHFSPlus or something per Dortania — I don't recall the history that led to the split or why its important, but I believe either should work for you.

If this isn't the problem, no other detail stands out to me. Good luck