Contribute
Register

General NVMe Drive Problems (Fatal)

To revive it I put it in a NVME to USB caddy and power on and off for various durations, e.g. 10mins on for 30 seconds off. I got the idea because old SATA SSDs can be revived in same way. So you could try that with your XG5.
Wow, thank you very much!
It stops detecting exactly after at least 3 force shutdowns and same firmware problem with dirty shutdown for the XG5 I found.
 
I have a Toshiba XG3 THNSN5512GPU7 that dies after hard shutdowns (e.g. powercuts or hold power button) despite being on Firmware 57DA4105 which was supposed to fix this problem: "- Fixed issue encountered in some system where drive is unable to detect after dirty power down".

To revive it I put it in a NVME to USB caddy and power on and off for various durations, e.g. 10mins on for 30 seconds off. I got the idea because old SATA SSDs can be revived in same way. So you could try that with your XG5.
Hm... How to tell between a drive that is dead, versus one that is just really distracted?!

Thx for interesting tip... My mind boggles at what makes this work...

Truly random access!
 
I know the SN750 is the most recommended NVMe SSD, but has anyone tried the newer model, the SN770?

According to this chart, it uses a WD proprietary controller, not one of the problematic Phison controllers. It's also priced similarly to the SN750 and seems to have better availability. My motherboard (ASUS Prime Z490-A) doesn't support PCIe 4.0, so I wouldn't be able to get the full advertised speeds, but if the price drops on the 2TB model, I might consider getting it.
 
I know the SN750 is the most recommended NVMe SSD, but has anyone tried the newer model, the SN770?

According to this chart, it uses a WD proprietary controller, not one of the problematic Phison controllers. It's also priced similarly to the SN750 and seems to have better availability. My motherboard (ASUS Prime Z490-A) doesn't support PCIe 4.0, so I wouldn't be able to get the full advertised speeds, but if the price drops on the 2TB model, I might consider getting it.

After a year of paying attention to this issue, I've never seen an explanation that claimed to understand why some drives are problematic and others aren't, and no one here is doing any research. The idea that certain controllers are a problem in and of themselves is a conjecture with no clear evidence.

Publishing a list is not research. To the extent that know-how is codified into certain well-read posts, it ages poorly, and is never backed up by any reference to deep understanding.

It comes down to some drives are known to not work well, and others are known to not have gone wrong yet.

So you will have to try.

The catch is I have personally the startup Trim malaise arrive across macOS releases, and also seen a 980 Pro go from seemingly fine wrt Trim to problematic with a firmware update. So even if you try and it works, this could change.

With hackintosh, the community however strong is not in a technical position to actually make things work, and never will be, because it lacks access to key engineering information to understand and advance approaches.

So a question such as "can XYZ drive work cause it's got a different controller" is not meaningful. It's at the level of should I sacrifice a chicken on Tuesday to ensure the gods will look after my hack?

—Don't be tempted to overindulge lore such as for compatible wifi/BT modules per my point: the case of a component with known support in macOS panning out is an edge case. There's never been any such for NVMe drives (plus I have seen a SATA disk from Apple that doesn't play right with sn external enclosure; it takes writes without error and throws them away). When considering clever workarounds, such as adapting Intel AX, or TB, the story always comes with caveats on functionality and later surprises.

At this point, with each release of macOS / 3rd party hack etc, it's half something-given, half something-taken-away. Something that doesn't get mentioned is the risk surface associated with the whole approach to hackintosh. Is my system burning half-a-core on kernel_task because it's been co-opted into a mining farm, or pumping spam? Anything can go on beneath the covers. After reading 10,000 posts I've ever seen one that says "after reviewing the code for kext XYZ...". Pls do not take this as sowing doubt, take it as appropriate concern.

The cataloging approach to gear lore used on these forums is based on a not-yet-known-to-fail precedent. That's powerful for cobbling together a running system, but it should never be confused with understanding.

Such approach can eventually lead to a ghetto.

My nagging thought is that the energy here needs to find a way over to Linux: help make it more mac-like in whatever way you think that's important.

Or better yet, work on something really new that puts maclike to dustbin.

Blah blah
 
Do I will lose the data after updating the NVMe Samsung 970 Evo Plus firmware?, do. I have to reinstall the OS from scratch?
I've done that before with them - it works after the update so no you shouldn't have to.
 
Hello

Is there any chance to use SetApfsTrimTimeout configuration on Clover or i need to move to OpenCore?

P.S.
There are two modifications of Samsung 970 evo plus with different controllers inside (sources:
https://www.ixbt.com/news/2021/08/27/ssd-samsung-970-evo-plus.html).
That is why somebody could install Monterey successfully and somebody not.
 
I know the SN750 is the most recommended NVMe SSD, but has anyone tried the newer model, the SN770?

According to this chart, it uses a WD proprietary controller, not one of the problematic Phison controllers. It's also priced similarly to the SN750 and seems to have better availability. My motherboard (ASUS Prime Z490-A) doesn't support PCIe 4.0, so I wouldn't be able to get the full advertised speeds, but if the price drops on the 2TB model, I might consider getting it.
yup, I've been using SN 770 for couple months, no problem at all. Speedy as always, SMBIOS MacPro7,1 and iMacpro1,1, boot time and runtime.

Now I'm switching to Kingston KC3000, sweet read and write (not recommended, every now and then kernel panic NVMe command time out).
Screen Shot 2022-05-25 at 19.53.02.png

Patriot Viper VP4300 (not recommended, unstable during sleep), WD SN750/SN770/SN850 still the most reliable.
Screen Shot 2022-06-05 at 11.56.15.png

Code:
panic(cpu 0 caller 0xffffff800c175776): nvme: "3rd party NVMe controller. Command timeout. Write. fBuiltIn=1 MODEL=Viper VP4300 2TB FW=V1.3B CSTS=0x1 US[1]=0x0 US[0]=0xa VID=0x1dbe DID=0x5236 CRITICAL_WARNING=0x0.\n" @IONVMeController.cpp:6053
Panicked task 0xffffff953ffc0670: 269 threads: pid 0: kernel_task
Backtrace (CPU 0), panicked thread: 0xffffff90750be540, Frame : Return Address
0xfffffff58d253a20 : 0xffffff800987fe7d mach_kernel : _handle_debugger_trap + 0x41d
0xfffffff58d253a70 : 0xffffff80099e0fa6 mach_kernel : _kdp_i386_trap + 0x116
0xfffffff58d253ab0 : 0xffffff80099d0313 mach_kernel : _kernel_trap + 0x4d3
0xfffffff58d253b00 : 0xffffff800981fa70 mach_kernel : _return_from_trap + 0xe0
0xfffffff58d253b20 : 0xffffff800988024d mach_kernel : _DebuggerTrapWithState + 0xad
0xfffffff58d253c40 : 0xffffff800987fa06 mach_kernel : _panic_trap_to_debugger + 0x2b6
0xfffffff58d253ca0 : 0xffffff800a114af3 mach_kernel : _panic + 0x84
0xfffffff58d253d90 : 0xffffff800c175776 com.apple.iokit.IONVMeFamily : __ZN16IONVMeController14CommandTimeoutEP16AppleNVMeRequest.cold.1
0xfffffff58d253da0 : 0xffffff800c158ac3 com.apple.iokit.IONVMeFamily : __ZN16IONVMeController13FatalHandlingEv + 0x141
0xfffffff58d253dd0 : 0xffffff800a049ea5 mach_kernel : __ZN18IOTimerEventSource15timeoutSignaledEPvS0_ + 0xa5
0xfffffff58d253e40 : 0xffffff800a049da8 mach_kernel : __ZN18IOTimerEventSource17timeoutAndReleaseEPvS0_ + 0xc8
0xfffffff58d253e70 : 0xffffff80098d2735 mach_kernel : _thread_call_delayed_timer + 0x505
0xfffffff58d253ee0 : 0xffffff80098d3802 mach_kernel : _thread_call_delayed_timer + 0x15d2
0xfffffff58d253fa0 : 0xffffff800981f19e mach_kernel : _call_continuation + 0x2e
      Kernel Extensions in backtrace:
         com.apple.iokit.IONVMeFamily(2.1)[67F397CD-C08B-3400-9F9E-04EFE8395C97]@0xffffff800c150000->0xffffff800c17cfff
            dependency: com.apple.driver.AppleMobileFileIntegrity(1.0.5)[AFBE32F3-B2D9-3B48-97F5-1D216E8D1F4B]@0xffffff800afa2000->0xffffff800afc3fff
            dependency: com.apple.iokit.IOPCIFamily(2.9)[135B75F2-5A3F-368C-A114-83A0AE6A10E2]@0xffffff800c425000->0xffffff800c450fff
            dependency: com.apple.iokit.IOReportFamily(47)[EA2F49CB-563B-331E-B686-371A181C09D5]@0xffffff800c462000->0xffffff800c464fff
            dependency: com.apple.iokit.IOStorageFamily(2.1)[B16ABD42-7E05-387B-940C-5077CBD0BACE]@0xffffff800c567000->0xffffff800c57dfff

Process name corresponding to current thread (0xffffff90750be540): kernel_task
Boot args: agdpmod=pikera keepsyms=1 -wegnoigpu chunklist-security-epoch=0 -chunklist-no-rev2-dev

Mac OS version:
21G5037d

Enter the new champion, PNY CS3140 2TB.
1655803382607.png

It is rocksolid, no kernel panic so far, running 2 of them on the system.
 
Last edited:
980 Pro update... Boot Trim delay issue disappeared.

Last January, I installed a Samsung 980 Pro firmware update, and the common problem of APFS Trim boot delay appeared. I wasn't working on any other OpenCore / kit tuning at the time, the firmware update triggered the issue. This was OC 0.7.9, macOS 12.3 as I recall.

I switched to an SN750 I had on hand and this resolved the problem. But if I installed the 980 Pro in another slot, it would also delay at a later point in the boot, right before the desktop comes up as long as the 980 was present, and the delay went away when the 980 was removed. SetApfsTrimTimeout option appeared in OC 0.8.0 and I didn't review it because I was hosted on SN750.

Recently I've been working on replacing the i9-10900K with an i9-11900K. After I got the 11900 up, I returned to the 980 Pro to benchmark the drive over PCIe 4.0 (The 10900 is PCIe 3.0) — I confirmed what others have reported, that the benchark performance increase for the 980 Pro is substantial at almost 2x.

Surprise, the Trim boot delay has disappeared.

I've verified that SetApfsTrimTIeoout = -1

I pared down my EFI a bit getting the 11900 running, but there aren't major changes. I did some review and can't find any change I think would be pertinent.

Maybe 12.4 update fixed it?

I'm sorry I can't offer more specific information, but maybe this helps someone.
 
After a year of paying attention to this issue, I've never seen an explanation that claimed to understand why some drives are problematic and others aren't, and no one here is doing any research. The idea that certain controllers are a problem in and of themselves is a conjecture with no clear evidence.


My nagging thought is that the energy here needs to find a way over to Linux: help make it more mac-like in whatever way you think that's important.

Or better yet, work on something really new that puts maclike to dustbin.

Blah blah
You can get more info here.
 
Back
Top