Mac Pro (Late 2013) GPU (Driver) Issues

[axantas]

Or you could just turn off your Thunderbolt display off and then on which gives you access to your computer again but the GPU will not wake up which makes the computer super slow and you will have to reboot anyway. But now that you have access to the cursor, you can reboot properly.

I can confirm that. It works. Just let it finish the job and reboot. It is just the GUI which gets slow.

 

[bax2003]

 

[poematik13]

my 6c/d500 was repaired today. they wouldn't let me read the repair log that was on their ipads. said it was "not for the consumer"

so i really cant confirm what they did, they just said they changed one of the grapghics card. fully expecting this to die again as soon as I render another project.

 

[bax2003]

my 6c/d500 was repaired today. they wouldn't let me read the repair log that was on their ipads. said it was "not for the consumer"

so i really cant confirm what they did, they just said they changed one of the grapghics card. fully expecting this to die again as soon as I render another project.

Did they give you report which components are replaced ?
I always have this info.....parts replaced xxx-xxxx.

 

[gpu-crashman]

Hi Everyone, I'm a little late incoming to this party, but I bring the same tidings. Had a client working in Adobe PremierePro CC2015, with the user reporting sluggish performance. But whenever the user attempted to export his sequence from Premiere, the system would hang / crash when they would attempt to select a destination, or when the export actually began. I was tailing the system.log file when I noticed all of the GPU failure / restarts. So I got to researching...

My main question for the group still checking this forum, is how many people are running dual monitors? Our setup on this system has dual 24" HP monitors. *BUT* they are DVI inputs. So, we're utilizing 2 Thunderbolt to DVI adapters.

Here's my thinking, When looking at the Thunderbolt ports on the back of the MP, the bottom 2 ports & the HDMI port are all using the same bus. Could I be bugging out the GPUs by running Thunderbolt to DVI on 2 monitors?

Thoughts?

 

[bax2003]

Hi Everyone, I'm a little late incoming to this party, but I bring the same tidings. Had a client working in Adobe PremierePro CC2015, with the user reporting sluggish performance. But whenever the user attempted to export his sequence from Premiere, the system would hang / crash when they would attempt to select a destination, or when the export actually began. I was tailing the system.log file when I noticed all of the GPU failure / restarts. So I got to researching...

My main question for the group still checking this forum, is how many people are running dual monitors? Our setup on this system has dual 24" HP monitors. *BUT* they are DVI inputs. So, we're utilizing 2 Thunderbolt to DVI adapters.

Here's my thinking, When looking at the Thunderbolt ports on the back of the MP, the bottom 2 ports & the HDMI port are all using the same bus. Could I be bugging out the GPUs by running Thunderbolt to DVI on 2 monitors?

Thoughts?

Scott Johnson in this topic said that he solved his issue by switching to HDMI display.
Regarding Tb (Displat Port) outs, I think we have here all kinds of possible combos and all of them have issues.

I had my Tb displays connected to Ports in Bus 0 and 2. I connected them now both to Bus 0 and we shall see if this has anything to to with bus bugging.

 

[gpu-crashman]

Scott Johnson in this topic said that he solved his issue by switching to HDMI display.
Regarding Tb (Displat Port) outs, I think we have here all kinds of possible combos and all of them have issues.

I had my Tb displays connected to Ports in Bus 0 and 2. I connected them now both to Bus 0 and we shall see if this has anything to to with bus bugging.

Do you know if you (or anyone else) has the "Displays Have Separate Spaces" option enabled under Mission Control? Maybe is some sort memory resource call in the underlying GPU architecture?

That is to say, 1 monitor, in HDMI is ok?, >1 monitor causes issues WHILE "Displays Have Separate Spaces" option enabled?

 

[bax2003]

Do you know if you (or anyone else) has the "Displays Have Separate Spaces" option enabled under Mission Control? Maybe is some sort memory resource call in the underlying GPU architecture?

That is to say, 1 monitor, in HDMI is ok?, >1 monitor causes issues WHILE "Displays Have Separate Spaces" option enabled?

I don´t want to disable that....if you disable that, you don´t have dock and menu bar on both displays....i need that.

 

[gpu-crashman]

I don´t want to disable that....if you disable that, you don´t have dock and menu bar on both displays....i need that.

I understand, but IF having multiple monitors AND Separate spaces is causing this issue, we should at least positively identify it. Then, (if this should indeed be the smoking gun), we can individually make the decision; is this worth it.

I'm going to follow this plan with my editors if this issue shows up again. If I get any update, I will post to this thread.

 

[flyingfoxuk]

I understand, but IF having multiple monitors AND Separate spaces is causing this issue, we should at least positively identify it. Then, (if this should indeed be the smoking gun), we can individually make the decision; is this worth it.

I'm going to follow this plan with my editors if this issue shows up again. If I get any update, I will post to this thread.

I've had the problem with only one monitor - so can't be all the issue, sadly

 

[axantas]

I don´t want to disable that....if you disable that, you don´t have dock and menu bar on both displays....i need that.

It is not the culprit - i am using only one monitor - disabled it, i did not see any difference concerning the hangs. Leave it on...

 

[edanuff]

Here's my thinking, When looking at the Thunderbolt ports on the back of the MP, the bottom 2 ports & the HDMI port are all using the same bus. Could I be bugging out the GPUs by running Thunderbolt to DVI on 2 monitors?

Thoughts?

As others have pointed out, you can have these problems with a single monitor. That said, there's evidence that passive adapters like Apple's Thunderbolt to DVI adapter don't work well with the nMP GPUs and you should use active DisplayPort to DVI adapters instead.

Separate point, the issues around what you connect to which TB bus for the most part only apply if you're using Apple Thunderbolt displays (or two specific LG models). Otherwise, the TB ports switch to DisplayPort-mode and the TB bus bandwidth considerations don't really apply. The bottom ports are the only ones that have special considerations, don't plug a monitor there if you're using HDMI.

 

[poematik13]

Did they give you report which components are replaced ?
I always have this info.....parts replaced xxx-xxxx.

Yeah, they told me that GPU-B (im assuming this is the compute one?) was replaced. The computer runs fine ATM, its not crashing during sleep anymore. I have a session with a client later today so hopefully it holds up and can actually render things.

 

[bax2003]

Yeah, they told me that GPU-B (im assuming this is the compute one?) was replaced. The computer runs fine ATM, its not crashing during sleep anymore. I have a session with a client later today so hopefully it holds up and can actually render things.

GPU-A is compute GPU, GPU-B is actual "video card", all 6 Display Ports + HDMI are connected to it.

 

[poematik13]

GPU-A is compute GPU, GPU-B is actual "video card", all 6 Display Ports + HDMI are connected to it.

I see. Yeah the computer is working fine now. I've been cutting/grading/rendering a short film on it. It doesn't crash during sleep anymore, and it renders without glitches. Let's see how long this lasts.

 

[chama98]

I see. Yeah the computer is working fine now. I've been cutting/grading/rendering a short film on it. It doesn't crash during sleep anymore, and it renders without glitches. Let's see how long this lasts.

I was thinking that myself. I had to reformat and reinstall the OS and so far everything seems to be ok. Apart from windows 10 on my boot camp partition is crashing with graphics issues so I am thinking of going back to window 8.1. I am thinking the drivers are quite buggy.

I am hoping now that I have mentioned that I haven't had any problems (yet!) I have probably jinx it!!

 

[poematik13]

I was thinking that myself. I had to reformat and reinstall the OS and so far everything seems to be ok. Apart from windows 10 on my boot camp partition is crashing with graphics issues so I am thinking of going back to window 8.1. I am thinking the drivers are quite buggy.

I am hoping now that I have mentioned that I haven't had any problems (yet!) I have probably jinx it!!

yeah it is wise to never ever use a new OS until 2-3 updates/patches. leave windows 10 and os x 10.11 for all the newbies and consumers who have time to experience bugs for us, they are the test subjects. my stuff has to be mission critical.

 

[gpu-crashman]

Issues happen with only 1 display as well you say; well, crap. Okay, well, I suppose all we can do is to stay vigilant with tracking the issue. But, it REALLY does sound related to the OS, doesn't it. Does that makes this issue software? Maybe a firmware version? Then why all the GPU replacements? ugghhh.. this is so frustrating.

 

[MMcCraryNJ]

.

 

[nitrocan]

I really think this is a software/driver/OS issue than anything else...

My crashes have gone down to about 1 or 2 a month, and it's mostly always while using Chrome, especially when playing video in Chrome. The funny thing is, if a currently playing video freezes my machine, the video is still actually playing while the machine is unresponsive, because the audio from the video continues to play until the end of the video. The GUI becomes unresponsive and the video will freeze on a frame, but it's still continuing playback while the system remains unresponsive. I imagine if it were a hardware fault, the whole thing would just stop.

It's the GPU that crashes. The computer itself is completely fine otherwise. The video is actually still playing, you just can't see it.

 

[poematik13]

i would LOVE for this just to be an OS issue that can (eventually) be patched. Thats way easier than hardware. I hate the downtime that comes with constantly having the machine in for service because the GPU's die.

This is totally unacceptable for a "high end" computer regardless. it should be rock solid and render nonstop for years at the price they sell them for.

 

[koyoot]

The amount of failures across the entire lineup of Apple computers points to OS problem, not the hardware. Remember, it does not affect only Mac Pro, but GPU restarts also touched Macbook Pros bot 13 inch with Intel integrated GPUs and 15 inch with Nvidia discreet GPUs.

 

[poematik13]

Yosemite is definitely the worst OS X to ever release aside from Lion. Too many new features piled on and not enough refinement and bug fixes. And now they are pushing El Capitan on us this year. It has been shown that the yearly update cycle for iOS and OS X means less stability.

 

[chama98]

Today, I had another freeze and after following what someone else recommended as in unplugging my thunderbolt display and reconnecting. I had to hit command + Control and eject on the keyboard. My mac restarted. I looked in the console and found these. Can anyone shed a little light and explain what this means?

09/09/2015 19:11:09.000 kernel[0]: [ PCI configuration end, bridges 42, devices 36 ]
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB1
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB2
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB3
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB4
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB1
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB2
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB3
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB4
09/09/2015 19:11:09.000 kernel[0]: pci pause: NHI1
09/09/2015 19:11:09.000 kernel[0]: pci pause: NHI0
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB0
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB0
09/09/2015 19:11:09.000 kernel[0]: pci pause: UFS1
09/09/2015 19:11:09.000 kernel[0]: pci pause: UFS0
09/09/2015 19:11:09.000 kernel[0]: pci pause: BR11
09/09/2015 19:11:09.000 kernel[0]: pci pause: BR10
09/09/2015 19:11:09.000 kernel[0]: [ PCI configuration begin ]


The screen saver was frozen. A process did crash which was adobe Desktop Service.

Many thanks

 

[nitrocan]

Today, I had another freeze and after following what someone else recommended as in unplugging my thunderbolt display and reconnecting. I had to hit command + Control and eject on the keyboard. My mac restarted. I looked in the console and found these. Can anyone shed a little light and explain what this means?

09/09/2015 19:11:09.000 kernel[0]: [ PCI configuration end, bridges 42, devices 36 ]
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB1
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB2
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB3
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB4
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB1
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB2
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB3
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB4
09/09/2015 19:11:09.000 kernel[0]: pci pause: NHI1
09/09/2015 19:11:09.000 kernel[0]: pci pause: NHI0
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB0
09/09/2015 19:11:09.000 kernel[0]: pci pause: DSB0
09/09/2015 19:11:09.000 kernel[0]: pci pause: UFS1
09/09/2015 19:11:09.000 kernel[0]: pci pause: UFS0
09/09/2015 19:11:09.000 kernel[0]: pci pause: BR11
09/09/2015 19:11:09.000 kernel[0]: pci pause: BR10
09/09/2015 19:11:09.000 kernel[0]: [ PCI configuration begin ]


The screen saver was frozen. A process did crash which was adobe Desktop Service.

Many thanks

Why exactly did you have to press Cmd + Ctrl + Eject after reconnecting the display?