Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

What do you think is the source of the graphics issues on Mac Pro (Late 2013) ?

  • Hardware

    Votes: 69 53.1%
  • Software

    Votes: 28 21.5%
  • Hardware & Software

    Votes: 32 24.6%
  • Other

    Votes: 1 0.8%

  • Total voters
    130
One of the new Mac pros went down at work today, freezing up during any GPU intensive activity such as video encoding with Adobe Media Encoder. Booting in safe mode seemed to alleviate the issue but obviously without much usefulness. D700's.

Though it may be a thermal issue. After cool down and with the building a/c blasting, the issue is not happening.
 
Last edited:
One of the new Mac pros went down at work today, freezing up during any GPU intensive activity such as video encoding with Adobe Media Encoder. Booting in safe mode seemed to alleviate the issue but obviously without much usefulness. D700's.

Though it may be a thermal issue. After cool down and with the building a/c blasting, the issue is not happening.

Total freeze ? Could you move mouse cursor ?

EDIT: No problems after 72 hours.

EDIT: No problems after 4 days.
 
Last edited:
Just a quick follow-up post. I don't use this machine as much as some people use theirs, but I'm up to just over 37 hours of usage time now (figure from the 'ac' command) and no freezing as yet. That is a longer period of use than was typical between freezing episodes before. So it's looking good so far.
 
Just a quick follow-up post. I don't use this machine as much as some people use theirs, but I'm up to just over 37 hours of usage time now (figure from the 'ac' command) and no freezing as yet. That is a longer period of use than was typical between freezing episodes before. So it's looking good so far.

I really hope that your replacement machine is fine....for at least forever ;)

Mine is fine after 5 days :apple:
 
I've had the same problem, my nearest Apple Centre in Birmingham UK (45 mile round trip, had to go there 3 times) eventually replaced both A and B D300s, but the problem is no better, I'm now going to try going through a local retailer instead.
 
I've had the same problem, my nearest Apple Centre in Birmingham UK (45 mile round trip, had to go there 3 times) eventually replaced both A and B D300s, but the problem is no better, I'm now going to try going through a local retailer instead.

I am very sorry to hear that :(
Did you have problem with brand new machine ?
 
Yes, this was happening from new, before Yosemite it was just crashing an individual app, now it's taking down the whole machine - which (coupled with a GPU swap making no difference) makes me thing it's a driver problem - but I'm willing to try a motherboard swap too.

Has anyone reported this problem with D500 cards?
 
Yes, this was happening from new, before Yosemite it was just crashing an individual app, now it's taking down the whole machine - which (coupled with a GPU swap making no difference) makes me thing it's a driver problem - but I'm willing to try a motherboard swap too.

Has anyone reported this problem with D500 cards?

After crash and reboot, do you have GPU.restard entries in Console or some different error.

EDIT: No problems after 6 days :apple:
 
Last edited:
Hey VirtualRain

Where did you get the nMP block diagram? Cool, I had never seen it complete like that.
Do you have a better original? Full size that is, so that it can be zoomed and readable? :)
Thanx

----------

HP just announced desktops based on AMD 380 Ooops...
No word on specs yet but it could be a sign.
 
Hey VirtualRain

Where did you get the nMP block diagram? Cool, I had never seen it complete like that.
Do you have a better original? Full size that is, so that it can be zoomed and readable? :)
Thanx

----------

HP just announced desktops based on AMD 380 Ooops...
No word on specs yet but it could be a sign.

From Google search....unfortunately there is no available larger image, PDF or more info :(
 
After crash and reboot, do you have GPU.restard entries in Console or some different error.

I'm getting kernel gpu.restart errors, like other people in this thread.

The first Apple centre I tried said they had heard of the problem before, and swapping both the GPUs was Apple's solution... either I got another bad pair, or that's not the solution, because it's still happening, sometimes up to 6 or 7 times a day.

I'm not sure if it's gradually getting worse, or I'm just getting more annoyed by it since the OS upgrade turned it from a single process crash into a near total lock-up (I can move the mouse pointer, and apps playing sound carry on playing, but no controls are responsive apart from the power button.

I've yet to hear any reports of this happening with D500s, only D300s and D700s, so I'm wondering if the different architecture of the D500 might provide a clue to what's going on, and a possible way to get my machine to be reliable.
 
I'm getting kernel gpu.restart errors, like other people in this thread.

The first Apple centre I tried said they had heard of the problem before, and swapping both the GPUs was Apple's solution... either I got another bad pair, or that's not the solution, because it's still happening, sometimes up to 6 or 7 times a day.

I'm not sure if it's gradually getting worse, or I'm just getting more annoyed by it since the OS upgrade turned it from a single process crash into a near total lock-up (I can move the mouse pointer, and apps playing sound carry on playing, but no controls are responsive apart from the power button.

I've yet to hear any reports of this happening with D500s, only D300s and D700s, so I'm wondering if the different architecture of the D500 might provide a clue to what's going on, and a possible way to get my machine to be reliable.

I am asking because a would like to find out for the benefit of all of us: does this happans with gpu.restart console errors even if it turns out that GPUs are not the source of the problem.

In my case, GPU-B was definitely the source of the problem.
 
I'm getting kernel gpu.restart errors, like other people in this thread.

The first Apple centre I tried said they had heard of the problem before, and swapping both the GPUs was Apple's solution... either I got another bad pair, or that's not the solution, because it's still happening, sometimes up to 6 or 7 times a day.

I'm not sure if it's gradually getting worse, or I'm just getting more annoyed by it since the OS upgrade turned it from a single process crash into a near total lock-up (I can move the mouse pointer, and apps playing sound carry on playing, but no controls are responsive apart from the power button.

I've yet to hear any reports of this happening with D500s, only D300s and D700s, so I'm wondering if the different architecture of the D500 might provide a clue to what's going on, and a possible way to get my machine to be reliable.


What monitor are you using? I have a dell u2711 displayport to mini displayport and another newer dell 2414 hdmi. I still have frequent crashes and powering the dell u2711 on and off made it resposive again. With 10.10.3, when it becomes responsive again, the whole system lags and i'll do save my stuff and restart. With 10.10.2 a monitor off on, will solve any issues with no lag. I'm on d300.

I doubt that its a hardware issue, if it is, it might be the inefficient cooling design metioned here, despite what a posh english dude in California touting the best design ever said. Or highly likely, apple want to be a fashion company and making things work is no longer in their ethos.
 
Mac Pro (Late 2013) GPU (Driver) Issues

When it's locked up you may be able to login from another machine (enable remote login). I was as able to login from my phone. There's a free iOS terminal emulator you can use for that. Once in you can issue commands and, e.g., reboot the machine that way rather than holding in the power button.
 
If you have another Mac it may be easier to enable screen sharing before the crash.

The block diagram was interesting as I have been on a quest to burrow through layers of PCI to PCI bridges to get an eGPU working in Windows. There is an amazing amount of complexity with the various logical switches all interconnected like a spiders web.

It is so interconnected that disabling what appears to be an unused port or switch will have unforeseen consequences. I will disable something that appears to have nothing connected to it in EFI shell, exit she'll and boot to Windows. Where I will frequently discover that I have killed all USB ports.

In short, the ability to connect 3 TB ports with s sets of PCIE lanes and 6 video outputs requires much more complexity than past systems. The PCI -I -b output from nMP is 3 times the length of same list of 2014 Mini. And as I have discovered, one tiny mis connect anywhere can cascade on its way down.
 
affects D500's too...

I can confirm a very similar issue with a nMp with D500 cards - issue only manifested when I had to update from 10.10.2 to 10.10.3 (following a hard crash I had to do an internet recovery). The problem first revealed itself when importing a raw photo CR2 file into the new photo's app. Upon trying to edit it - rotate or colour correct the whole machine locks and you have to hard boot out of it - it seemingly corrupts something on the machine that is only cleared after a full erase and re-install. Having done 5 full installs on this machine yesterday I have had the problem manifest using similar files in Preview and Pixelmator. Also the problem occurred using a standard TIFF file. I spent several hours on the phone with a senior Applecare advisor yesterday. We performed a data capture of log files and went through the problem with him screen sharing remotely. He has passed the problem up to the engineering department and I am waiting for him to report back to me. I also experienced GPU restarts in relation to opening Final Cut Pro X on a fresh install of 10.10.3

This Mac Pro is only about 2 months old - it was a replacement unit supplied by Apple after I had 3 months intermittent problems with an 8 month old Mac Pro at the start of this year. This unit was shipped to me direct from the factory in Texas, having already lost 3 months work to nMp problems this year I wasn't particularly pleased to find new ones (on the previous faulty machine literally every component was replaced and it still proved unreliable - random restarts / shut downs / refusing to power on. The faulty machine was shipped to Apple HQ Ireland for further testing and I heard no more after my current replacement arrived)

I will be interested to see if anyone can confirm a definite fix for these GPU resets - it's very hard to determine whether something has been broken in the update from 10.10.2 to 10.10.3 or it is a hardware fault that has only just manifested itself. For what it's worth my monitor is a DVI connection, connected to the nMp by an official Apple adapter.
 
I can confirm a very similar issue with a nMp with D500 cards - issue only manifested when I had to update from 10.10.2 to 10.10.3 (following a hard crash I had to do an internet recovery). The problem first revealed itself when importing a raw photo CR2 file into the new photo's app. Upon trying to edit it - rotate or colour correct the whole machine locks and you have to hard boot out of it - it seemingly corrupts something on the machine that is only cleared after a full erase and re-install. Having done 5 full installs on this machine yesterday I have had the problem manifest using similar files in Preview and Pixelmator. Also the problem occurred using a standard TIFF file. I spent several hours on the phone with a senior Applecare advisor yesterday. We performed a data capture of log files and went through the problem with him screen sharing remotely. He has passed the problem up to the engineering department and I am waiting for him to report back to me. I also experienced GPU restarts in relation to opening Final Cut Pro X on a fresh install of 10.10.3

This Mac Pro is only about 2 months old - it was a replacement unit supplied by Apple after I had 3 months intermittent problems with an 8 month old Mac Pro at the start of this year. This unit was shipped to me direct from the factory in Texas, having already lost 3 months work to nMp problems this year I wasn't particularly pleased to find new ones (on the previous faulty machine literally every component was replaced and it still proved unreliable - random restarts / shut downs / refusing to power on. The faulty machine was shipped to Apple HQ Ireland for further testing and I heard no more after my current replacement arrived)

I will be interested to see if anyone can confirm a definite fix for these GPU resets - it's very hard to determine whether something has been broken in the update from 10.10.2 to 10.10.3 or it is a hardware fault that has only just manifested itself. For what it's worth my monitor is a DVI connection, connected to the nMp by an official Apple adapter.


Give powering off and on all monitors a try next time hang happens. It works as a temp fix for me. My hypothesis is something's amiss with software. Mine starts happening frequently with yosemite. Doubt it's hardware
 
Give powering off and on all monitors a try next time hang happens. It works as a temp fix for me. My hypothesis is something's amiss with software. Mine starts happening frequently with yosemite. Doubt it's hardware

Thanks, I will try that - for the moment I've been getting on with some rendering with Maxon Cinema 4D which fortunately is a largely CPU based task so it's not been causing too many problems, but I'd like the issue sorted asap!
 
Thanks, I will try that - for the moment I've been getting on with some rendering with Maxon Cinema 4D which fortunately is a largely CPU based task so it's not been causing too many problems, but I'd like the issue sorted asap!


I do mograph with c4d and ae also web backend dev with phpstorm. Crashed for me happens mostly with chrome or vlc is active. And many times when flash or vlc is full screen. Im taking a hit with productivity and developed a newmacpro crash anxiety.
 
I do mograph with c4d and ae also web backend dev with phpstorm. Crashed for me happens mostly with chrome or vlc is active. And many times when flash or vlc is full screen. Im taking a hit with productivity and developed a newmacpro crash anxiety.

I get what you're saying about crash anxiety - my first unit earlier in the year went off for repair four times before I finally got it replaced with a new unit. The new machine apparently has a logic board revision which I was hoping meant it wasn't 1st generation tech anymore and that the initial problems had been ironed out - apparently this isn't the case yet.
 
GPU reset

This issue is definitely still persisting.GPU reset this morning just launching Final Cut Pro X - obviously with all the fancy new features offloading to the GPU something has gone wrong somewhere - at the moment I'm unable to edit video or photos, use iTunes - fortunately I'm still able to keep rendering with Cinema 4D at the minute and Logic Pro X managed to work without crashing this morning to finish off a sound mix, but very frustrating overall
 
I'm actually glad it's not just me - having had 3 months of "different issues" with another nMP doing things no one had ever seen it's reassuring not to be alone!
 
For what it's worth, for those experiencing this problem I've just forwarded a load more diagnostic data to my applecare representative who tells me the engineers are aware of the issue and investigating. I would advise anyone who is experiencing this GPU reset problem to contact Applecare and report what is causing the problem in your individual cases. I have sent them the links to this thread, but I'm sure the more information they have the easier it should be to fix this issue.
 
False alarm here, unfortunately. After few (perhaps week's worth?) of daily resumes, I got GPU reset loop this morning. At least no kernel panic, but sigh..

But none since that - so it seems it works as well in 10.10.3 as it did under Mavericks (occassional GPU reset loop on HDMI switch / resume, about once per month). I can live with that, but am not exactly super-happy. 10.10.0/1/2 were essentially unusably annoying (kernel panic once a day on average).

Given the rest of OS X 10.10 quality problems, I am really wondering if it is worth supporting Apple by buying more hardware or not, as I am quite concerned about the direction the quality of their hardware+software combination is going..

Unfortunately (for example) for Lightroom use case, only Windows is the other option. *sigh*
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.