Had the exact same issue with the short black Sintech adapter. Only cold boot was reliable, but that issue was resolved when I switched to the long black sintech adapter (ST-NGFF2013-C).
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
NVMe with ST-NGFF2013-C; Vega Internal GPU; Mac Pro 2013 (6,1)
- Thread starter CodeJingle
- Start date
- Sort by reaction score
Had the exact same issue with the short black Sintech adapter. Only cold boot was reliable, but that issue was resolved when I switched to the long black sintech adapter (ST-NGFF2013-C).
On macOS 10.13.4, I sometimes get kernel panics related to NVMe.
On wake from sleep:
panic(cpu 0 caller 0xffffff7f9f9af07c): nvme: "IONVMeController::HandleControllerPowerOff, ShutdownNVMe() completed with status = 0xe00002e9\n"@/BuildRoot/Library/Caches/com.apple.xbs/Sources/IONVMeFamily/IONVMeFamily-356.50.26/IONVMeController.cpp:2011
On system halt:
panic(cpu 0 caller 0xffffff7f9bfab261): nvme: "systemWillShutdown, ShutdownNVMe() completed with status = 0xe00002e9\n"@/BuildRoot/Library/Caches/com.apple.xbs/Sources/IONVMeFamily/IONVMeFamily-356.50.26/IONVMeController.cpp:508
Not sure if the panics are new or only the reporting.
Edit: Inserted proper log entries
On wake from sleep:
panic(cpu 0 caller 0xffffff7f9f9af07c): nvme: "IONVMeController::HandleControllerPowerOff, ShutdownNVMe() completed with status = 0xe00002e9\n"@/BuildRoot/Library/Caches/com.apple.xbs/Sources/IONVMeFamily/IONVMeFamily-356.50.26/IONVMeController.cpp:2011
On system halt:
panic(cpu 0 caller 0xffffff7f9bfab261): nvme: "systemWillShutdown, ShutdownNVMe() completed with status = 0xe00002e9\n"@/BuildRoot/Library/Caches/com.apple.xbs/Sources/IONVMeFamily/IONVMeFamily-356.50.26/IONVMeController.cpp:508
Not sure if the panics are new or only the reporting.
Edit: Inserted proper log entries
Last edited:
Ooh juicy. I would recommend to file bug report with Apple even though this is a non-standard hardware configuration.
https://developer.apple.com/bug-reporting/
I submit all the panic reports. But yeah, Apple's track record in supporting third-party hard- and software is definitely less than stellar.
Are you talking about the automated panic submit tool? Yeah of course you do that too, but this needs more thorough reporting, multiple people on this thread are seeing issues. If you can consistently reproduce the issue, then manually file a full report with Apple through the developer bugs site. Assuming you are a developer. You sound like an engineer to me.I submit all the panic reports. But yeah, Apple's track record in supporting third-party hard- and software is definitely less than stellar.
It is a black hole from which nothing meaningful escapes, so don't expect any love from Apple in return. Your contribution is silently held in high regard.
I guess a better bug report would be to reproduce the issue with a passive adapter, not the angelshark. That would put more responsibility on Apple's side. But you're the first guy on the thread to post a literal panic report, so that makes you the guy to file this. If you so choose.
Last edited:
I've filed bug reports with Apple before (bugreport.apple.com) and spent a considerable number of hours to make them easily reproducible. I've stopped doing that. Apple doesn't seem to read the reports or simply doesn't care. I've only ever received a feedback on one report (serious file corruption in Finder) – two years later.
@Kris Kelvin - I've gotten similar responses when calling in for support. It's like they don't have any bug/defect/support ticket management. Also, you have an AngelShark? How do you like it (besides the kernel panics)?
Does Apple have a way to manage support cases via the internet?
Does Apple have a way to manage support cases via the internet?
No I don't have an AngelShark. I was asking if Kris was seeing the Kernel Panic with the AngelShark, if he could reproduce the bug without the AngelShark and file that as a dev bug.
I can't speculate on the quality of the ticket management for Apple when it is at the developer level. Because it is all internal. I could say the secrecy is bad for the support system but also the most profitable company in the world is highly scrutinized so they are saving themselves some headaches.
Apple has several types of managed support. Developer technical support requests, developer bugs, public beta tester feedback, and the 'I have a problem with my computer' support requests.
https://developer.apple.com/support/technical/
https://developer.apple.com/bug-reporting/
https://beta.apple.com/sp/betaprogram/
https://support.apple.com/
Last edited:
Since I am looking for a way to bypass the I/O Wall power button, I started sanding the I/O Board. I got video of it this time. After scraping off the tin and the mask-off for both top layers I bore through a layer in about 45 minutes.
Missing traces from the photo I can extrapolate using data from the video which allows me to be slightly less formal in the process and I can get through the layers a little bit faster. I have to test the theory to make sure I can extrapolate board trace from video footage. Ignore my initial feedback in the video since I still ended up using a formal process. The rest of the video is all vacuum so probably best to mute the audio.
I go back and forth between the metal file and 60, 80, 100, 120, 150, and 220 grit sandpaper. I was going for speed so didn't use the fiberglass.
https://www.dropbox.com/s/l1l1j37tb2q6fu4/20180409_035628.mp4?dl=0
https://www.dropbox.com/s/tpwe3afhkvxzf2e/20180409_043341.mp4?dl=0
https://www.dropbox.com/s/m521x1ar0o48nrs/20180409_050658.mp4?dl=0
Missing traces from the photo I can extrapolate using data from the video which allows me to be slightly less formal in the process and I can get through the layers a little bit faster. I have to test the theory to make sure I can extrapolate board trace from video footage. Ignore my initial feedback in the video since I still ended up using a formal process. The rest of the video is all vacuum so probably best to mute the audio.
I go back and forth between the metal file and 60, 80, 100, 120, 150, and 220 grit sandpaper. I was going for speed so didn't use the fiberglass.
https://www.dropbox.com/s/l1l1j37tb2q6fu4/20180409_035628.mp4?dl=0
https://www.dropbox.com/s/tpwe3afhkvxzf2e/20180409_043341.mp4?dl=0
https://www.dropbox.com/s/m521x1ar0o48nrs/20180409_050658.mp4?dl=0
Last edited:
They do have several, but consider it reasonable to share as little information with the outside world as possible.
I do and I like it – apart from the kernel panics.
I haven't tested again with the Sintech adapter, but I suspect the panics are related to Apple's NVMe drivers and happen to any non-Apple NVMe drives. For example, I've seen reports of them occurring with OWC's Aura Pro X, too.
Apple is at a loss not having you as an active bug reporter, too bad for them. I've gotten fast turn around for a response from Apple for filed bugs but I used to be a tester and test developer and I have an understanding of the politics involved. Most of the effort of a bug report is framing it to not be ignored.
Last edited:
Oddly the first one of these I purchased worked fine. The latest 2 have 50/50 chance of working. I've ordered one of the ST_NGFF2013-C adapters as well as the other ones which I believe to be the 'A' version (pre B & C) as they're local and will get here quicker. If that 'A' version doesn't play either, the 'C' will be the back up plan. I'm really hoping to sort this soon as I have a client waiting for me to deliver this upgraded 10 Core 6,1.
[doublepost=1523370443][/doublepost]
If you're using the Sintech adapter, they mention on the website that their adaptors don't support sleep mode.
Seems like a fully working four lane PCIe 3.0 peripheral would only be one or two pins away from supporting sleep/wake. Like adding sleep/wake support to the NVMe might be as easy as connecting the standby power pin and the WAKE#/PERST# pin.
Anyone on this thread a PCIe developer who can confirm this? Anyone want to blow $4k to become a PCI-SIG member, download and research the documentation to answer my question, and then respond with a correct answer?
http://pcisig.com/membership/become-member
If I'm not mistaken, sleep was working fine for me with both the AngelShark and the Sintech adapter before 10.13.4. Not sure what changed.
Does sleep work for other people using a non-Apple NVMe drive with macOS 10.13.4?
Also, this might be related to the AngelShark. Maybe Apple's drivers just don't expect there to be a second NVMe drive in the tube.
Does sleep work for other people using a non-Apple NVMe drive with macOS 10.13.4?
Also, this might be related to the AngelShark. Maybe Apple's drivers just don't expect there to be a second NVMe drive in the tube.
If some people are seeing an issue with reboot, and others are seeing an issue with sleep/wake, with full shutdown and boot from full shutdown working, I'll bet the underlying issue is the same.
After taking an hour to sand through the second layer of the front side of the IO Board (a ground plane), the third layer appeared and disappeared so quickly ... it was a ninja layer. Even though I was going back and forth between the 150 and 220 grit sandpaper. I'm not sure I've ever see a layer so short lived. I should have taken some footage with it doused in alcohol to get a better shot of the traces. This is why I prefer to not use the metal file when sanding boards down. But a slower process would take 8 or 16 hours instead of 1 hour. In the third layer were some traces with widths so narrow they were practically microscopic. I'm glad I recorded the process on video otherwise there would be no documentation of this layer. The fourth layer (another ground plane) was already showing through the board by the time I stopped.
https://www.dropbox.com/s/wdtbnrlad4x8yaw/20180410_170334.mp4?dl=0
https://www.dropbox.com/s/pxmt5v8dy9uto45/20180410_173649.mp4?dl=0
https://www.dropbox.com/s/mz0c72h68ss1csm/20180410_181003.mp4?dl=0
[doublepost=1523478233][/doublepost]I now have the 400 microscopic probes to involve somehow in further probing the Mac Pro 2013
https://www.dropbox.com/s/wdtbnrlad4x8yaw/20180410_170334.mp4?dl=0
https://www.dropbox.com/s/pxmt5v8dy9uto45/20180410_173649.mp4?dl=0
https://www.dropbox.com/s/mz0c72h68ss1csm/20180410_181003.mp4?dl=0
[doublepost=1523478233][/doublepost]I now have the 400 microscopic probes to involve somehow in further probing the Mac Pro 2013
Last edited:
Platform controller should map most of the GPU pins that don't already have a straight mapping back to the CPU.
https://www.intel.com/content/www/us/en/chipsets/c600-series-chipset-datasheet.html
[doublepost=1523555413][/doublepost]I was looking for IO Wall test points, I found them on the other side of the board. Silly me for not looking there sooner.
https://www.intel.com/content/www/us/en/chipsets/c600-series-chipset-datasheet.html
[doublepost=1523555413][/doublepost]I was looking for IO Wall test points, I found them on the other side of the board. Silly me for not looking there sooner.
I mapped the IO Wall pins to the test points. Still not sure how to replicate pressing the power button, though now I am properly set up to do a logic probe of a functioning IO Wall if I buy a replacement, to eventually figure it out so I can get rid of it again later. I also dived a bit deeper in sanding down the damaged IO Wall flex board.
Also using my fancy wires now yay!
Also using my fancy wires now yay!
Last edited:
After three weeks of trying 4 of the similarly designed Sintech adapters with my Samsung 960 Pro, I can finally come to the conclusion (as did slypete) that the long black adapter (ST-NGFF2013-C) is the only adapter I have tried that so far has absolutely no issues with either cold starts, restarts or wake from sleep. On the other 3 adapters I tried all conceivable tape application options but never received consistent results. The ST-NGFF2013-C requires absolutely no tape, you simply slot it in and you are ready to go.
As to why only one of the identically pinned adapters would work flawlessly, I will leave this perplexing problem to the other clever scientists on this thread.
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Several people have recommended it yes. I'll double check the beginning of the thread references ST-NGFF2013-C and I'll also add AngelShark (and remove PC Parts 239).
[doublepost=1523724561][/doublepost]If I depopulate another PC Server board and only remove the masking of the top layer then i should be able to cross-probe every pin on the board (though I might need to solder some jumpers). Took another three hours but I finally sanded down the top layer of thrle front of the 2U board.
Also I'm fairly confident to finally upload a video demonstration this weekend of continuity testing on an actual Mac Pro 2013.
lmfao ... mac pro 2013 boots after swapping graphics cards ... the internal diagnostics said both gpus were fine it can't tell if the one with the hard drive is in the correct slot or not. no error beep no nothing it just silently laughs at you while appearing to otherwise boot correctly. a mixture of ignorance and stupidity on my part. good thing no one is paying me to do this. laugh and learn.
well besides mounting the cpu heat sink and securing the whole thing on a movable platform i need to properly ground the boards. some of the screw mounts double as drain to ground so without the boards mounted I have to wire up the same drain points to AC earth ground, possibly also an intermediate metal plate. Some of those mounts also tripled as heat sinks but being out of the trash can housing I don't have to worry about that as much.
with the mac pro booting freely out of its prison now i can wire up a replacement internal gpu for testing
https://www.dropbox.com/s/jo1asx8ryxae366/20180414_163810.mp4?dl=0
well besides mounting the cpu heat sink and securing the whole thing on a movable platform i need to properly ground the boards. some of the screw mounts double as drain to ground so without the boards mounted I have to wire up the same drain points to AC earth ground, possibly also an intermediate metal plate. Some of those mounts also tripled as heat sinks but being out of the trash can housing I don't have to worry about that as much.
with the mac pro booting freely out of its prison now i can wire up a replacement internal gpu for testing
https://www.dropbox.com/s/jo1asx8ryxae366/20180414_163810.mp4?dl=0
Last edited:
After going over the GPU pinout for the PC server board I'll have to probe the Mac Pro directly to verify against the PC mapping. Then I need to solder 164 leads to the test GPU. Then I need to wire up a spare flex cable to the broken-out GPU. There is a small chance I get far enough to actually test a different GPU in the Mac Pro this weekend. If it boots I still need to login and check system information to see what the Mac Pro recognized the other GPU as (correctly or in a D300/500/700 bucket list).
Logging in may be difficult while I am manually holding the heatsink to the CPU.
Logging in may be difficult while I am manually holding the heatsink to the CPU.
Maybe I need to try this and see if it fixes my cold boot issues. I have this one at the mo
https://www.amazon.co.uk/gp/product/B073FRCBSP/
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
If I can get more people to confirm that ST-NGFF2013-C is working in Mac Pro 2013 I can have that as the main option in the first post to the thread. Right now I have listed both ST-NGFF2013 and ST-NGFF2013-C.
[doublepost=1523833466][/doublepost]oh wow that is alot of prep work. i think i'm finally ready to probe the mac pro 2013. I also tinned the S9000 pins before attaching wires for testing.
[doublepost=1523833466][/doublepost]oh wow that is alot of prep work. i think i'm finally ready to probe the mac pro 2013. I also tinned the S9000 pins before attaching wires for testing.