Mac Pro (Late 2013) GPU (Driver) Issues

[AidenShaw]

especially because in OS X Window Server crashes and you don´t have something like that in Windows environment. "Display driver stopped responding and has recovered"

The Windows and OSX architectures are actually quite similar from the 10k metre view. Apple calls the subsystem managing the windows the "Window Server", Windows calls it the "display driver".

The salient point is that on both systems the display subsystem can crash and stop responding - but Windows can do a hard reset of the display subsystem and continue after a short pause.

A nice side benefit of the Windows approach is that the display drivers can be upgraded without a system restart. The Nvidia installer replaces the old driver files with the new ones - and then triggers a "recovery". The display driver restarts, using all of the new files.

 

[andrewdon]

I just picked up a 2nd hand nMP.
Looks like the bastard who sold it to me forgot to disclose the frequent Kernel Panics! Subbing to this thread to see if it finds a resolution. I've noticed that it seems to be linked primarily to ihome/iwork apps.


Anonymous UUID: F465C0AB-BF93-DE58-6C96-3EF065EC0AAB

Sun Jul 1 10:12:43 2018

*** Panic Report ***
panic(cpu 2 caller 0xffffff7f98d16577): "displayModeDidChange failed: set_display_mode_and_vram failed!
"@/BuildRoot/Library/Caches/com.apple.xbs/Sources/GPUDriversAMD/GPUDriversAMD-1.66.42/Common/GLKernel/ATI/Accel/AMDAccelDisplayMachine.cpp:299
Backtrace (CPU 2), Frame : Return Address

 

[Martyimac]

I just picked up a 2nd hand nMP.
Looks like the bastard who sold it to me forgot to disclose the frequent Kernel Panics! Subbing to this thread to see if it finds a resolution. I've noticed that it seems to be linked primarily to ihome/iwork apps.


Anonymous UUID: F465C0AB-BF93-DE58-6C96-3EF065EC0AAB

Sun Jul 1 10:12:43 2018

*** Panic Report ***
panic(cpu 2 caller 0xffffff7f98d16577): "displayModeDidChange failed: set_display_mode_and_vram failed!
"@/BuildRoot/Library/Caches/com.apple.xbs/Sources/GPUDriversAMD/GPUDriversAMD-1.66.42/Common/GLKernel/ATI/Accel/AMDAccelDisplayMachine.cpp:299
Backtrace (CPU 2), Frame : Return Address

Others will hopefully respond but here is my take on it. I never used the ihome/iwork apps and it happened regularly with me. In my case, video in Facebook or uTube could cause it. It happened when my screen saver using my photos library would be on. I had one or two occurrences just doing mundane things like word processing. There was no real rhyme or reason to the occurrences.
Sorry you got that computer.

 

[OldMike]

I just picked up a 2nd hand nMP.
Looks like the bastard who sold it to me forgot to disclose the frequent Kernel Panics! Subbing to this thread to see if it finds a resolution. I've noticed that it seems to be linked primarily to ihome/iwork apps.


Anonymous UUID: F465C0AB-BF93-DE58-6C96-3EF065EC0AAB

Sun Jul 1 10:12:43 2018

*** Panic Report ***
panic(cpu 2 caller 0xffffff7f98d16577): "displayModeDidChange failed: set_display_mode_and_vram failed!
"@/BuildRoot/Library/Caches/com.apple.xbs/Sources/GPUDriversAMD/GPUDriversAMD-1.66.42/Common/GLKernel/ATI/Accel/AMDAccelDisplayMachine.cpp:299
Backtrace (CPU 2), Frame : Return Address

Hate to hear that. I just picked up a brand new nMP (6-core/16GB/500GB/D500 - upgraded to 64GB RAM) for a good deal. I have been following this thread, crossing my fingers that this doesn't happen to me. I have been running 24/7 for almost a week, without an issue. I made sure to get AppleCare because of this thread.

Out of curiosity, which model do you have? Any non-factory upgrades?

If you bought it from somewhere like ebay, maybe you can file a claim and get your money back. Good luck!

 

[andrewdon]

Hate to hear that. I just picked up a brand new nMP (6-core/16GB/500GB/D500 - upgraded to 64GB RAM) for a good deal. I have been following this thread, crossing my fingers that this doesn't happen to me. I have been running 24/7 for almost a week, without an issue. I made sure to get AppleCare because of this thread.

Out of curiosity, which model do you have? Any non-factory upgrades?

If you bought it from somewhere like ebay, maybe you can file a claim and get your money back. Good luck!

It was a gumtree find. I’ve been hanging on to my 2010 quad core, for years, but I’ve been loathe to buy the dustbin when a major refresh is around the corner.

Luckily I only spent about ($1600) here in South Africa, that’s considered cheap.

It doesn’t seem to have any upgrades or mods, I have done another fresh reinstall of 10.13.5, and loading the bare essentials in terms of workflow. ;ets see how well she holds together. Maybe Apple is kind enough to stretch the out-of-warrantee support to the D300 crowd.

I’m only using it for adobe illustrator and some basic design stuff. No transcoding or anything like that. M aye it’s somethig I can live with, now that adobe FINALLY figured out that auto save should be a thing.

 

[MackyMcMac17]

Interesting... I'm on 10.13.6 beta now...So my trash bin got the usual panic:
displayModeDidChange failed: set_display_mode_and_vram failed

This time I got the gray screen that said the machine "restarted because it encountered a problem". Better than the machine just locking up, I guess....

Anyone noticed the same?

 

[Martyimac]

Interesting... I'm on 10.13.6 beta now...So my trash bin got the usual panic:
displayModeDidChange failed: set_display_mode_and_vram failed

This time I got the gray screen that said the machine "restarted because it encountered a problem". Better than the machine just locking up, I guess....

Anyone noticed the same?

I got that failure notice when I went to 10.13.5. A little improvement but not by much.

 

[Pardo83]

I'm still on Sierra and I have noticed an increase in windowserver crashes (without the ability to use SSH to reboot the mac) in the last 3 days (3 crashes a day! while before it was 1 crash a week). Temperature is very hot here these days, so it's probably related to it?

 

[Martyimac]

Sorry to report that my D500 machine just had it's first ever freeze. Sigh. Going on vacation shortly but will try and check into this thread as I can. BUT, there is no crash log in Console?????
I am out of ideas.

 

[bax2003]

Sorry to report that my D500 machine just had it's first ever freeze. Sigh. Going on vacation shortly but will try and check into this thread as I can. BUT, there is no crash log in Console?????
I am out of ideas.

Were you able so SSH reboot it ? Was it standard Window Server crash with active mouse ?

 

[Martyimac]

Were you able so SSH reboot it ? Was it standard Window Server crash with active mouse ?

Don't know what you mean by SSH reboot it. Yes, mouse was active except for no clicks registered.

 

[Infrared]

Don't know what you mean by SSH reboot it.

This involves logging in remotely (e.g., from a phone) via SSH and using a terminal command to reboot the Mac. Discussion of reboot methods here:

https://apple.stackexchange.com/que...-command-to-shutdown-restart-and-sleep-my-mac

I would tend to prefer one of the two following conmands under normal circumstances (i.e., when the machine has not frozen):

osascript -e 'tell app "System Events" to shut down'

osascript -e 'tell app "System Events" to restart'

 

[Martyimac]

This involves logging in remotely (e.g., from a phone) via SSH and using a terminal command to reboot the Mac. Discussion of reboot methods here:

https://apple.stackexchange.com/que...-command-to-shutdown-restart-and-sleep-my-mac

I would tend to prefer one of the two following conmands under normal circumstances (i.e., when the machine has not frozen):

osascript -e 'tell app "System Events" to shut down'

osascript -e 'tell app "System Events" to restart'

This goes WELL beyond my comfort/knowledge level.

 

[nglevin]

I am now at five days of unexpected stability with Mojave developer beta 3 on this Mac Pro, fwiw. This is the longest stretch of not-freezing I've experienced, and I've been using it through several hot days.

Prior to that I tried having the restart-your-mac process tied to an Automator action. I'd like to say that it is better than ssh-ing into your Mac Pro to restart it, but the required authentication challenge means that you need to fish for the authentication window blindly while the entire GUI is frozen to enter your password. Which, considering how that can go wrong (surprise, you just accidentally shared your password with _____!), is absolutely not a good solution.

 

[axantas]

Instead of using SSH there is this quite handy app in the AppStore
https://itunes.apple.com/us/app/defrost-pro/id1352993174?mt=8

It is not free, I am not the developer - works very well AND you get the opportunity to send an error report after restarting. The ones I send do not contain very friendly comments...

General situation with my D300: Stunningly stable and no freezes with the latest version of MacOS in the last three weeks.

 

[Infrared]

An advantage of the Applescript above is that authentication is not required.

Does CMD+CTL+EJECT work if the machine is locked up?

Cheers.

 

[nglevin]

I think you're on the money, CMD+CTRL+EJECT and the AppleScript `tell app "System Events" to restart` both politely tell each app to close and have macOS restart. And they're safer than restarting from command line. Still, both are better than a power cycle.

Unfortunately, some apps that take too long to quit, or hold up the process with a modal dialog asking if you're sure you want to discard the current X (Thunderbird...) will stop the entire process. Same as restarting macOS through normal means.

Fortunately I am still experiencing the sweet sweet embrace of stability, for which I hope Murphy's Law does not soon come into play.

 

[FilthyMcNasty]

Things got worse here, seems to have coincided with ambient temperatures in the UK but it's anyones guess!

For anyone who hasn't noticed, 10.13.6 release is out..

 

[ManuelGomes]

Yep, released yesterday. Fixed anything?

 

[Ghafoor]

Unfortunately, no

 

[ManuelGomes]

Bummer

 

[FilthyMcNasty]

Had one crash since installing 10.13.6

Flash screen appeared as mentioned by other posters - no other obvious difference

 

[boli]

Since my last post I've brought my Mac Pro (quad-core, 16 GiB, 256 GB, D300s) to an Apple Store for service.

They found no problems, but because my symptoms were the same as for the D500/D700 repair program, they replaced both D300s anyway (for free), but did not replace the logic board.

I got it back last Friday morning, migrated my stuff back to it during the day, and have been using it normally since yesterday. Unfortunately, it just froze again (that's with 10.13.6, but the Kernel Panic notice only appeared after rebooting via SSH, like with earlier versions).

As a history recap from my first post: Apple completely replaced my Mac Pro (originally bought in June 2014) in December 2016 with a brand new one.

 

[James259]

Hello everyone,
I have a nMP, 32GB ram, D300, and 10-core CPU, and have had system freezes since I bought the machine refurbished. The best performance that I have had in the past 80 days is 3 days in a row, roughly 5-8 hours per day, with no freezing.

In this message, I will share:

1. What goes on when my machine freezes, and
   
   
2. Propose a set of reproducible conditions which I would appreciate someone else trying. Perhaps if we can find a set of testable conditions that can be reproduced then Apple will take note.

1. What goes on when my machine freezes
The freezing I see is:

• no visible response from the keyboard when letters are typed,
• music is still playing,
• the cursor moves around the screen when I touch the trackpad, a
• there is no response to using Command-Tab to switch apps.
• I also cannot use the mouse to move, resize, or operate buttons or scroll bars in the windows.
• When I press Command-Option-Control-Eject-Delete-Eject, then the music pauses, and I hear a beep sound repeated until I take my fingers off of the keyboard.

When the machine is frozen, and I pull the display port cable out of the Thunderbolt2 port, the monitor goes black, and when I put it back into the same port I took it out of, I see a message on the screen that says "Your computer was restarted because of a problem." The machine reboots and a crash report window comes up, the first three lines of which are:

*** Panic Report ***
panic(cpu 0 caller 0xffffff7f99b8464b): "displayModeDidChange failed: set_display_mode_and_vram failed!
"@/BuildRoot/Library/Caches/com.apple.xbs/Sources/GPUDriversAMD/GPUDriversAMD-1.68.20/Common/GLKernel/ATI/Accel/AMDAccelDisplayMachine.cpp:321


2. A set of reproducible conditions
I have found a set of conditions which other users may be able to replicate, that seem to consistently cause the instability on my nMP within 35 minutes of booting-up:

I open:

• Mail,
• Pages,
• Calendar,
• Safari (and play this video:
  ),
• If you have it, open Photoshop too (I opened PS CS6).

Note the time, and leave the computer running with the monitor on, and all app windows open. I turned the volume of the music to low using the keyboard. You can use these apps on the machine if you want to (don't open any more apps), or leave the machine alone. These test conditions have brought on the freezing when I use the computer with only these apps open, and when I leave it unused.

Would someone else be kind enough to try these conditions and reply to say what happened to their machine?
Thanks in advance,

Jim

 

[Kris Kelvin]

I've documented the UI freezes on my MacPro6,1 (with D300) for the last 6 months.
Usage: Light/medium workload every day for ~ 12 hours.

Findings:

• 118 freezes
• 102 of them while decoding video (YouTube or IINA)
• the most reliable way to reproduce the problem has been playing a YouTube video in Safari and using WebGL (e.g. playing Solitaire)
• generally speaking, things have improved considerably

It's hard to attribute the improvements to any single configuration change. However, things got better on february 24th and march 29th – the dates when I removed and cleaned both GPU boards. There was hardly any dust – only lots of thermal paste that I couldn't completely clean out of the resistors.



Still, something in the hardware improved. GPU temperature and voltage didn't change though. My only theory so far is that the thermal paste somehow interacts with the resistors. Maybe the paste's properties change after being "cooked" by the GPU for a few months? There's still way too little data for any real conclusion.

If other users decide to clean their GPU, I'd be interested to know if it helped. btw, you don't have to take the whole thing apart to remove the GPU. When following the iFixit tutorial, you can basically skip steps 3-9.