Mac Pro (Late 2013) GPU (Driver) Issues

[magic carpet]

Update on my case:

In my case the Service Point found a problem/fault with the RAM-memory which got fixed.
According to the Service-Point the GPUs are ok.
After half a day working I can confirm that no more freezes occured - I hope it stays like this.

 

[SamPotts]

The crashes are getting more frequent again. So for me, replacing both cards didn't fix it permanently and neither does the latest (beta) version of High Sierra unfortunately. Going to ask for a replacement and see how I go. I can't use another machine for work and this my day to day machine. Leaving it with them at the store for ages while they change the graphics cards again only for them to fail again later, isn't really an option.

 

[MackyMcMac17]

My garbage hasn't crashed/froze since upgrading to 10.13.3. It's not clear what's going on. I didn't change anything else. This thing is ridiculous....

 

[loby]

I am also interested in the same.

Don’t know for sure, but my D500’s seem to be ok. Guess I got lucky.

 

[axantas]

My Trashcan has reverted to these nasty unrecoverable crashes/freezes (push the power button as only exit) since upgrade to High Sierra 10.13.3 just a couple of days ago. Fortunately the transition did not damage anything in bootcamp. So: using Windows in most cases eases the situation a bit...

Thinking about leaving OS X and make that thing a Windows-only machine which can run all day long without any hassle.

 

[AidenShaw]

make that thing a Windows-only machine which can run all day long without any hassle

Perhaps Apple will make that the official "fix" for GPU issues on the 6,1 😉

 

[eepyikes]

Can anyone confirm that they're having the problem in cooler room temperatures (maybe below 72F)? I had GPU B replaced and got the crash the very next day, twice (running 10.11.6). Then after that it hasn't crashed in like two months...but that's exactly when it got colder here. So I'm wondering if temperature is a contributing factor.

 

[axantas]

Can anyone confirm that they're having the problem in cooler room temperatures (maybe below 72F)? I had GPU B replaced and got the crash the very next day, twice (running 10.11.6). Then after that it hasn't crashed in like two months...but that's exactly when it got colder here. So I'm wondering if temperature is a contributing factor.

I cannot "confirm" that fact, although - strange - I had a broken heating so around unusual 19 degrees Celsius and three strange freezes (power cycling as only exit) these days. But it did not happen right after booting, so the machine itself was not cold anymore. Furthermore, it happened right after upgrading to High Sierra.

 

[MarkJames68]

A way to test if overheating is a possible factor is to use the Mac fan control program and set the fan to the highest setting.

 

[axantas]

... underheating in this case - highly doubt it, but you never know 😉

 

[mikeboss]

I have a theory that the connectors of the flexcables are to blame. there are 1800 contacts in total! this might explain the more or less randomly occuring crashes.

 

[MarkJames68]

I have a theory that the connectors of the flexcables are to blame. there are 1800 contacts in total! this might explain the more or less randomly occuring crashes.

If you're brave and handy you can partially disassemble the system, blow it out with air and re-assemble.

Any way you look at it something is wrong given all of these issues.

 

[bax2003]

GUI freeze happened again few minutes ago...solutions reboot via SSH

Timestamp Thread Type Activity PID TTL
2018-03-17 18:06:40.370867+0100 0x61a Fault 0x0 0 0 kernel: (IOAcceleratorFamily2) bool IOAccelDisplayMachine2::display_mode_did_change(uint32_t): vendor driver returns false
--------------------------------------------------------------------------------------------------------------------
Log - Default: 0, Info: 0, Debug: 0, Error: 0, Fault: 1
Activity - Create: 0, Transition: 0, Actions: 0

 

[AtomicGrog]

I have the same problems.
Behaviour ranges from a screen freeze, where i can still move to mouse around but nothing else to total system freeze and restart.
On some of these restarts i have seen a red light coming from the bottom of the mac.
Sometimes i get a GPU restart report in console, other times there is nothing. (I assume the system is so frozen it can't write a report anymore.)
I've been in contact with apple several times already, reset nvram, clear caches, etc.
Reinstalled OS X twice, and still this problem is there.

Have seen the screen freeze, oddly much more likely when playing Kerbal (KSP). The oddity being frozen screen but reactive mouse. Dont really want to revert OS. Will need to cope for now (not too frequent...) and wait for patch.

My spec is slightly different to others here, I'm running dual D700's.

 

[voyager77]

Can anyone confirm that they're having the problem in cooler room temperatures (maybe below 72F)? I had GPU B replaced and got the crash the very next day, twice (running 10.11.6). Then after that it hasn't crashed in like two months...but that's exactly when it got colder here. So I'm wondering if temperature is a contributing factor.

My room rarely gets above 72F, even in summer, i never noticed any influence of room temps on the crashing.
[doublepost=1521468979][/doublepost]

Have seen the screen freeze, oddly much more likely when playing Kerbal (KSP). The oddity being frozen screen but reactive mouse. Dont really want to revert OS. Will need to cope for now (not too frequent...) and wait for patch.

My spec is slightly different to others here, I'm running dual D700's.

The graphics crash with a frozen screen but working mouse is a strange one indeed, but it's the one i had happen most of the time.

As you found out, certain apps / games have a tendency to crash the system often. Some were even a good way to cause a crash but then after an osx update suddenly some games became very stable, others remained randomly unstable.

 

[eepyikes]

After I had sent mine in for repair and they replaced graphics board B, it crashed twice the next day (same GPU Hung error as usual). Then it never crashed again for about two months. I was running 10.11.6 at the time. This weekend, I decided it must be safe to upgrade to 10.13.3...well apparently not, it crashed like 3-4 times in the last two days on 10.13.3. [edit]another week has gone by, several more crashes, however none with the beachball...now the issue is when I come back to my desk, the screen will not wake up[\edit]

One thing to note is that on 10.11.6 I always go the GPU driver appears to be hung" error, but on 10.13.3 I get "GPU Reset".

Does anyone know how to more specifically find the errors in the console on 10.13.3? The search is super weird there. Someone posted it on this thread way back but I can't find it.

 

[enieschmenie]

Hello Everyone,

I too have one of these shiny black metal trashcans that freeze undeterministically. To be completely sure that the freezes do not occur due to some devices that are plugged into the Mac I have replaced each and every external periphery that I use with my Mac by now.

I have tested the Mac in various places / different building to make sure that its not the outdated power supply of my home that causes the problems.

I had the machine sent to Apple Technicians at least twice. Even took it even more often to my local Mac distributor. Apparently the hardware is all good.

Since nothing helped and the extent of the problem changes sometimes with the current macOS version I can only conclude that there is a software problem so that macOS is not able to properly handle the two D300s.

What I realised while programming for the GPUs is that it seamed to me that even tho the cards are labeled A and B, the internal order of the GPUs switches while the system is running. I do not have any experience with other dual GPU systems, so I'm not sure if this is uncommon. But still this drew my attention.

Also interesting to me was the behaviour of `top` when SSHing onto the Mac while it is frozen. usually top shows all the processes that are currently running plus how much memory and cpu they are currently using. When the Mac's GPU is frozen it is still possible to SSH onto it and do all sorts of things. But top will not give you the proper information that it should.

So for now I have a solution, not to stop the Mac from freezing, but to calm me at least a little bit with this issue. Since it is possible to SSH onto the device and restart it remotely, I have programmed a tiny App for my phone to do exactly that for me. You would only have to open the app on your phone, select the device in your local network and tell it to restart or shutdown. Since admin names and passwords can be saved in the keychain it is only a matter of three taps to restart your frozen Mac. Whenever I do not have my Macbook running next to my Mac Pro this is a way faster solution than using ssh from a terminal. If anyone is interested you can get it in your local stores near you! 😀. Just search for

DEFROST PRO

I really do hope that you don't mind that I take a minimal fee simply because putting an app in the App Store is not free either and I don't suspect to make profit whit it. I will make it a free app as soon as I have the money back for the app distribution account.

 

[axantas]

I am - in a very positive way - kind of speachless, enieschmenie.
We are speculating about the possible reasons of the crashes for years now. (...the amount of metorites crashing into saturn is one of them) This is the FIRST time ever I hear about someting like "if X (or not X) then crash". I doubt, Apple knows about that.

You have at least one customer.

Question: The crashes in 10.13.3 are very consistent. The GUI (windowmanager ???) stalls and dies. The whole concert in the background happily continues to play, wichout the slightest "hickups". A completely frozen Logic Pro, doing its heavy work in the background - until you powercycle it...

Is there any way to kill or/and restart the windowmanager in your app? I am using Windows 10 in Bootcamp often - the crashes are present in Windows as well, but they are nothing else, but a long screen flicker and restart of the driver/window manager. Obviously it IS possible to do anything else than the Apple way of freezing.

 

[enieschmenie]

I would have answered your question by only mentioning or linking your name axantas – just as thou did with mine. Unfortunately I'm a fool with no idea how to do so and only limited time to learn. Now, marvel the wholeness of your question once again in this post, only so thou canst have proper conversations with one another in this place.

Question: The crashes in 10.13.3 are very consistent. The GUI (windowmanager ???) stalls and dies. The whole concert in the background happily continues to play, wichout the slightest "hickups". A completely frozen Logic Pro, doing its heavy work in the background - until you powercycle it...

Is there any way to kill or/and restart the windowmanager in your app? I am using Windows 10 in Bootcamp often - the crashes are present in Windows as well, but they are nothing else, but a long screen flicker and restart of the driver/window manager. Obviously it IS possible to do anything else than the Apple way of freezing.

First of all it makes me very sad to hear that Windows is crashing as well. I cannot but conclude that it must indeed be the hardware then, that is possessed by the devil.

I too, wondered if there was a way to restart the window manager. Woefully I have to say that I could not find a way to do so. More correctly speaking – There is a way to stop the window manager and all sorts of running programs, but to no effect. If anybody finds the task that must be killed or restarted, I would gladly improve the app to do what must be done. That would be a huge relief for all of us I think.

 

[axantas]

I must admit, that thing with the names ist just based on pure laziness of us nowadays users - if you can copy it - copy it. Copying your name to the left of the messages and pasting, so no mircale. If it would have been "Bob", I would have typed it myself - enieschmenie was a bit too uncommon to start with. Now I know it. 😉

Whether it is hard- or software - I guess even Apple is clueless and coding around - or just ignoring it...

Wat I wanted to say: Windows has its glitches with the MacPro from time to time as well, but it does a brilliant job. If it crashes, it just recovers in the blink of an eye - something I absolutely can live with. So Windows drivers can be restarted, the Apple ones are only to be restarted by letting the Mac crash and restart.

Anyway: Let's look forward to a lot of reactions - maybe ideas for your app.
And: as it really is a brillant idea - don't stop to ask that little fee for it too early... 🙂

 

[AidenShaw]

Windows has its glitches with the MacPro from time to time as well, but it does a brilliant job. If it crashes, it just recovers in the blink of an eye - something I absolutely can live with.

Cupertino needs to hear this message from as many people as possible.

"Innovation" can mean that the software transparently recovers from hardware errors.

So that even if your hardware engineers can't figure out how to design the hardware, your software engineers could hide the problem.

As the orange orangutan-in-chief might tweet - "so SAD that WINDOWS runs better on the MAC PRO than APPLE OSX".

 

[bax2003]

That is the worst thing. I love macOS, but I will not use a half-baked products because of it.
It is obvious that macOS has nothing to do with this. nMP is broken, more or less since 10.9.1.

Cupertino needs to hear this message from as many people as possible.

As the orange orangutan-in-chief might tweet - "so SAD that WINDOWS runs better on the MAC PRO than APPLE OSX".

 

[magic carpet]

Ok, so sad, I'm back in the "my nMP freezes" club.

The machine will run 3 days just fine, then I will get 3 freezes in 2 hours. Then it will work one day, then will freeze again just one or two times in a day... it is absolutely randomly, super frustrating.

Sometimes (not always) I can recover the machine by unplugging a thunderbolt dock (OWC). The only thing connected to the dock are 2 usb cables connected to 2 monitors - to the monitors I just have keyboard and mouse and graphics tablet connected.
Nothing else.

I already tried it with everything connected directly to the nMP - no external peripherals except monitor, mouse, keyboard, everything plugged directly into the nMP. Still got freezes.

However, sometimes I have the feeling that it maybe has to do something with usb - I got quite a few freezes a very short time after connecting external usb-drives over even usb flash drives, doing nothing just copying data. Had it with lots of different usb drives, so the drives are fine. When this happens it will sometimes just loose the connection to the usb drive ("false eject message") or sometimes just "classic freezing".

However, room temperature ist definitely not the solution, we are working in about 18° C (64-65° F) and our nMP developed a huge freezing problem.

Graphics card already got replaced once for some other graphic problems, so no solution here.
The freezing is quite new.

 

[mikeboss]

the quad-core I don't own anymore froze with only keyboard and mouse connected to USB. and just one LCDisplay (1920x1200) connected to HDMI.

 

[ActionableMango]

The problem is that some of us have D300 graphics cards which cause the same issues but aren't covered by the free replacement.

I've seen some people here on MR get their D300 replaced free outside of warranty, so even though what you are saying is true for written policy, there appears to be some wiggle room and so "your mileage may vary".