Mac Pro (Late 2013) GPU (Driver) Issues

[swithdrawn]

I use an HP Z32x via DisplayPort.

I think netkas' tweak is the safest bet to cut the kernel panics... it's clear the issue is hardware related having to do with utilization of the two GPUs, so if you effectively kill one (at least openCL) it should help stability. Everything else is a shot in the dark... not to discredit anyone's efforts but I don't see any real reason why the Eugy84 hack would work (it certainly didn't for a lot of people here), and that old gfxCardStatus utility is for MacBook Pros with a discreet and integrated GPU. I doubt it would do anything on a Mac Pro. But if certain steps seem to help out, then by all means..

That being said I am really curious if anyone else has replicated the issue on Mojave developer beta 4 or newer, as mentioned it's been stable for me and it's been my longest streak yet without any kernel panic.

 

[Levv]

I use 2 Apple Led Cinema Display 27 inch screens

 

[Ghafoor]

Mine was stable for a while, but is now back to crashing again. I noticed that when it "freezes", if I pull out the power cord from my Dell monitor, the computer instantly reboots with a kernel panic.

What monitors are you guys using with your nMP?

I use an LG 34UC87-B monitor connected via Thunderbolt and HDMI. Freeze -> switching between the two inputs forces a kernel panic and automatic rebooting of the system

 

[mr1098]

Mine was stable for a while, but is now back to crashing again. I noticed that when it "freezes", if I pull out the power cord from my Dell monitor, the computer instantly reboots with a kernel panic.

What monitors are you guys using with your nMP?

ViewSonic VX2233WM at the moment, but switching shortly to my everyday -- an HP 27xw.

I will gladly go headless if it equates to no more KPs. Couldn't care less about the GPUs.

 

[jayden]

What monitors are you guys using with your nMP?

I have one monitor, Samsung U28D590. Had this Mac Pro for 3 or 4 years, and it has run mainly 24/7 with no issue. Surprised to see this error when waking the screen today via keyboard-press.

panic(cpu 0 caller 0xffffff7f95e0c64b): "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

As it's the first time it's ever happened, I won't worry about it, but if it becomes as prolific as some of the posters here, I will definitely return.

 

[swithdrawn]

Guys... 25 days of continuous use and I still haven't gotten any crash since switching to Mojave 10.14 developer beta 4 through 8 (I experienced the panic on beta 2). Anxious if anyone can corroborate... or prove this to be a fluke

 

[unvs]

I've had several runs of weeks to months of no crashing since 2015, but it always comes back...

I sort of gave up on the bug reporting after uploading diagnostic after diagnostic for months, and apple is always quick to close the reports as soon as a new OS version is out (which I can understand).

It started crashing again some months ago, and I decided to open a new bug report. It was quickly marked as DUPLICATE OF 37907452. Curious to see what response (if any) there'll be.

It's so frustrating. I don't feel like I could sell the machine either, just so someone else gets stuck with this.

 

[swithdrawn]

It's so frustrating. I don't feel like I could sell the machine either, just so someone else gets stuck with this.

I've thought about that too and won't list it on eBay. But since Apple don't acknowledge the issue, it's "in perfect working order" for their buyback program

 

[nglevin]

Guys... 25 days of continuous use and I still haven't gotten any crash since switching to Mojave 10.14 developer beta 4 through 8 (I experienced the panic on beta 2). Anxious if anyone can corroborate... or prove this to be a fluke

I have had one frightful WindowServer freeze, and it was shortly after my nMP received an upgrade to 10 cores.

Not yet a fluke.

Something about Mojave seems more stable. I've had the occassional WS GUI freeze on every Mac I've owned since Snow Leopard's developer beta; the only time the Mac Pro offended the freezing sensibilities was when it was happening multiple times per day. Definitely not so, now.

 

[spadefist]

I have been lurking on this thread for a while and tried most suggestions offered with little to no effect. However, the Mojave public beta seems to have fixed this for me at least. Prior to the move to the public beta I was seeing crashes approximately once a day. Since upgrade, I have done 2 days without issue (hence ancedotal), for the record I have 12c D500 nMP, with 2 x LG 34UM95 (via TB2) and 1 x U3415W connected (via HDMI to mDP).

 

[iMactouch]

I have no problems until now with a 10core D300 and via Thunderbolt 2 attached a LG 34UM95-P.

 

[swithdrawn]

Something about Mojave seems more stable.

I have been lurking on this thread for a while and tried most suggestions offered with little to no effect. However, the Mojave public beta seems to have fixed this for me at least.

I agree, since beta 4 my system has been very stable. High Sierra was always a dumpster fire but things seem to have improved in Mojave.

 

[Moisie2000]

Hello:

Recently I've been suffering from being unexpectedly booted out onto the login screen on my machine (8-core Mac Pro with D300 GPUs) every 24-48 hours; logging back in, I see messages in Console that the GPU driver was in a state of unreadiness. I even got this on a fresh macOS 10.12.6 install, with no additional software.

I've since tried macOS 10.14 beta 8 - and 24 hours later I suffered the same fate, complete with gpuRestart messages.

I'm obviously glad some of you are having success with Mojave - but it doesn't look like it's a complete cure. I'm beginning to explore the eGPU option... :-(

 

[swithdrawn]

Recently I've been suffering from being unexpectedly booted out onto the login screen on my machine (8-core Mac Pro with D300 GPUs) every 24-48 hours; logging back in, I see messages in Console that the GPU driver was in a state of unreadiness. I even got this on a fresh macOS 10.12.6 install, with no additional software.

Hmm.. this isn't the same issue as the rest of the thread (IOAccelDisplayMachine2::display_mode_did_change) . Have you tried Apple support channels? eGPU would probably help, in any case.

 

[Moisie2000]

Hmm.. this isn't the same issue as the rest of the thread (IOAccelDisplayMachine2::display_mode_did_change) . Have you tried Apple support channels? eGPU would probably help, in any case.

I'm pretty sure somewhere in the last 60+ pages (can't find it right now), someone else has had the same symptoms as I'm currently having. Of course, different versions of the OS are handling these problems differently, and there may be different degrees of GPU problem - but it's still a 2013 Mac Pro GPU issue! For the record, I have also seen the display-freezing error on other Mac Pros - so I am aware of the difference in symptoms, but believe it's the same underlying crappy GPU that's causing it.

As it happens, I'm getting Apple to address this machine because it's still in warranty - but when it's out of warranty, I don't fancy paying out to have the GPUs-and-God-knows-what-else replaced, only for it to fail again a year later (as has happened to some colleagues of mine). In this respect, I think an eGPU is going to be the better investment...
[doublepost=1534955007][/doublepost]

I'm pretty sure somewhere in the last 60+ pages (can't find it right now), someone else has had the same symptoms as I'm currently having.

I think it was listening to me: back in 10.12.6 again, the machine just completely froze - no mouse pointer - and wouldn't come alive by unplugging/replugging the display. On rebooting, I found a gpuRestart message in Console...

 

[swithdrawn]

different versions of the OS are handling these problems differently, and there may be different degrees of GPU problem - but it's still a 2013 Mac Pro GPU issue!

Very true -- different OS versions can manifest the same issue in different ways... even Windows seems to have the issue but recovers gracefully (as perhaps now Mojave does, at least for the display_mode_did_change panic). Hopefully you can make some progress with AppleCare.

PM me if you are interested in a Sonnet enclosure

 

[iigsie]

8 months after Apple eventually replaced the logic board On one of my D300 mac pro’s that was crashing daily before it’s not had a single crash. They’d already replaced both graphics cards, the cpu card and the attaching cables.

Also, it’s running from an image I took of it back in December 2017 when it was crashing daily.

I’ve also got another d300 model that has never had the issue, and I know of several others including the other two models that haven’t had this issue.

 

[swithdrawn]

8 months after Apple eventually replaced the logic board On one of my D300 mac pro’s that was crashing daily before it’s not had a single crash. They’d already replaced both graphics cards, the cpu card and the attaching cables.

Congrats, so I guess the chassis is the only original component at this point?

Still no display_mode_did_change panic since Mojave beta 4 29 days ago.

 

[iigsie]

Congrats, so I guess the chassis is the only original component at this point?

Still no display_mode_did_change panic since Mojave beta 4 29 days ago.

Chassis, io board and psu are the only original parts.


Not one single crash or implication of a crash in 8 months

 

[grunty]

I agree, since beta 4 my system has been very stable. High Sierra was always a dumpster fire but things seem to have improved in Mojave.

Any update on this? I’m just about to install the latest beta to see if it helps.

 

[nglevin]

Any update on this? I’m just about to install the latest beta to see if it helps.

I've seen two instances in the past two betas where the WS freeze would appear to start, then the system recovers successfully soon after. One time while watching YouTube, another time while running a Leopard Server VM.

The VM one was interesting because it briefly kicked me back to the login screen. This is all anecdotal, but I'm starting to really feel like Mojave has a recovery state when the nMP freeze starts happening.

Or maybe I'm just lucky.

 

[grunty]

I've seen two instances in the past two betas where the WS freeze would appear to start, then the system recovers successfully soon after. One time while watching YouTube, another time while running a Leopard Server VM.

The VM one was interesting because it briefly kicked me back to the login screen. This is all anecdotal, but I'm starting to really feel like Mojave has a recovery state when the nMP freeze starts happening.

Or maybe I'm just lucky.

Upgraded to the beta now - will post back how it goes.

 

[tim_tom]

I've seen two instances in the past two betas where the WS freeze would appear to start, then the system recovers successfully soon after. One time while watching YouTube, another time while running a Leopard Server VM.

The VM one was interesting because it briefly kicked me back to the login screen. This is all anecdotal, but I'm starting to really feel like Mojave has a recovery state when the nMP freeze starts happening.

Or maybe I'm just lucky.

same here. mojave beta 8. d300. i was kicked back to login screen while watching vlc yesterday. Luckily i didn't have unsaved files open.

 

[boli]

After my last post my quad-core D300s Mac Pro was brought to Apple once again (they previously replaced both video cards, but the problem persisted).

This time around they apparently replaced the logic board and I/O board, according to a Genius Bar "work confirmation" email last week. As I did not receive a pick-up-notification at the same time, I inquired about the MP's repair status.

They just let me know that they're replacing the MP with a new one – six-core D500s this time, as my original config is no longer available.

My guess as to what happened is that the repaired MP still exhibited the issues – which they may have been able to reproduce with the step-by-step instructions I provided. Bear in mind that this is only speculation based on the email I received.

I'll pick it up soon and hope it'll not have the issue, otherwise I'll give Mojave a go.

 

[JRBv3]

I'm still experiencing this exact scenario almost daily on 10.13.6