Mac Pro (Late 2013) GPU (Driver) Issues

[mikas]

No Monterey for me on my work Mac, thank You. It has to mature, and my softwares need to achieve their supported status for Monterey first.

Anyway, 6,1 6c/D300, uptime 16 days now. No sleep activated, and the thing is stable as it's new. Even with eGPU (by a hack) in use all day long. I am running 11.5.1 at the moment, but will update soon to 11.6.1, I guess. Just wanted to see my uptime grow for a while.

There are reports of crashes with just released Monterey, with 6,1 too, if you look at forum posts. It's not sure if it's the inmature OS, or if the bug would still lurk in those damn drivers or something. BUt let us have some hope those reports have landed on some Apple developer who cares, and has got some abilities to do something about it too. Not very likely, but yeah

BootROM did update itself from 428.140.7.0.0 to 429.40.7.0.1 on my other 6,1 (a D500) with BigSur 11.6.1. That machine has not been tested thoroughly with sleep functions yet, but it's been stable for some time now. But it's equipped with D500, opposed to D300.

I've got third one too, but I am prepping it for surgery, so no results or tests with that one for now. Maybe later.

 

[incumbent]

apple enabled some extra logging in a later build of Big Sur, my crash dumps and panics started giving me a lot more information and i took that as a sign that at least someone in engineering at apple is curious about the d300'ed trashcans, but it's been running so great with Windows 10 i don't have much need to boot into macOS on that workstation anymore. once i get the compatibility manager stuff sorted for cold boots and enable hyper-v it's going to be less likely it ever boots macOS again.

 

[arw]

D300 6-core here, bought it on release day in 2013.
Since a clean install of Big Sur almost a year ago, I didn‘t a single GPU related freeze/KP.
It KP’ed however constantly after only a few hours (sth. about watchdog and windows server).
Since I stopped mounting my network drives via SMB but use AFP, it runs for several weeks (including sleep each day) until I shut it down/reboot by choice.
I‘m only using a 2560x1440 screen, maybe that‘s why I don’t have any problems anymore.
Under Mojave and a 4K TV it constantly froze (reboot via ssh still possible).

 

[hirakujira]

Could you please tell me more about your settings? Such as which Windows 10 version do you install, which Bootcamp version you are using? And the GPU driver version on windows... etc

Just for the record, I bought new hardware and replaced my 6,1. I put the original Apple SSD back in and launched bootcamp and it hasn't crashed in weeks running Windows 10. I needed a Windows machine around anyway, and the 6,1 hauls ass. Won't run Windows 11 which is a shame cuz I rather like the changes introduced by 11, but that computer stopped ***** with me once I stopped running macOS. I'll try Monterey on it, I left macOS 128GB and gave the rest to Windows so I probably won't do any long-term testing or anything, but before you throw your trashcan6,1 into a dumpster considering running an OS that doesn't have a defective driver for the D300.

lol, AMD's software even immediately said "yo dawg i heard u like GPUs" and they're running in Crossfire now lolol something tells me I could probably install macOS via hypervisor and it'd be more reliable than installing macOS on the bare metal. it's embarassing but i'm glad i didn't have to put it on the shelf next to my Mac 512KE.

I've used virtualization for years to run software for oddball radios/scanners and some tools I have been meaning to try out relevant to my work. Windows' subsystem for linux is quite good btw, and these 6,1s are great docker hosts due to having ECC memory and a fast enough bus for storage. I've got an old OWC JBOD via thunderbolt hooked up now and I'm going to check out Storage Spaces because it looked like a cool technology. Lifelong learning for life over here.

Thanks!

 

[axantas]

Talking about Bootcamp/Windows: Do we know the opportunity to restart the graphics driver with "Ctrl(Strg)+Win+Shift+B"? I recently read about it and it was unknown to me. It will probably not recover a frozen Windows, but "on its way to a crash" it might help.

A low beep and a short flicker of the screen will confirm the correct execution.

 

[incumbent]

Could you please tell me more about your settings? Such as which Windows 10 version do you install, which Bootcamp version you are using? And the GPU driver version on windows... etc


Thanks!

i have swapped out the 4TB nvme i had in there with the original apple ssd so it's pretty fresh; i did a boot camp assistan initiated install using a windows 10 iso from microsoft. i told it windows 10 enterprise and it's using the AMD control center that bootcamp designated and haven't touched beyond making sure the settings in the control center are appropriate.

 

[Georgeus]

hi, it looks like this issue is related to high GPU temperature. This is solved by increasing the fan, at least it solved the problem for me.
This is not an advertisement. there is a TG Pro program that can change the standard operation of the cooling fan. I am listing the settings here. Hope this finally resolves this one.
I'm sorry for my not good english

 

[mihai_me]

For me it was not the temperature but Apple / Firefox vs Google.
Once I disabled Hardware Acceleration for both Firefox and Chrome my Mac Pro stopped crashing and it's been running flawlessly ever since.
I also dismantled it once, cleaned the connectors, etc (did not improve anything but it was a nice experience ).

 

[mikas]

moved to Mac Pro (Late 2013) Reastarts Randomly

 

[Levv]

i upgraded my trashcan6,1 with D300s to Monterey on the macOS side of the workstation. i haven't installed much software or anything, just OpenZFS and a bootstrap'ed Homebrew environment. i am curious if the updated extension for the D300s addressed any instability issues. No watchdog kernel panics yet but it's only been running for 20 hours while i do a large migration between pools.

not a single gpu crash dump in `DiagnosticReports` either, though like i said not much going on with it right now, so maybe someone in software engineering at Apple has done us a solid and fixed a bug? g-d knows they've gotten plenty of crashdumps and panics submitted. how many of us are out there still running 6,1s with D300 and tracking current release of macOS? couple thousand? i wasn't expecting anything to ever be done about it but who knows.

kextstat on Monterey:

emory@DreamOn ~ % kextstat |grep AMD
Executing: /usr/bin/kmutil showloaded
No variant specified, falling back to release
  129    2 0xffffff7f96c0d000 0xcb000    0xcb000    com.apple.kext.AMDSupport (4.0.7) 61E40B4A-8B3E-3569-85EB-250CD91E8EB1 <109 108 16 15 9 7 6 3 1>
  130    0 0xffffff7f80e5c000 0x56000    0x56000    com.apple.kext.AMD7000Controller (4.0.7) 7F1162EC-D382-3EF1-AC58-8F873717C45C <129 109 108 16 15 7 6 3 1>
  156    0 0xffffff7f814fc000 0x4000     0x4000     com.apple.kext.AMDRadeonX4000HWServices (4.0.7) B3551921-FA0C-35D7-9B75-65B82699A59F <108 16 15 9 7 6 3 1>
  161    0 0xffffff7f81044000 0x461000   0x461000   com.apple.kext.AMDRadeonX4000 (4.0.7) 45CB4427-5148-335B-8FA2-D5A2002CB6BC <160 108 90 16 9 7 6 3 1>
  163    0 0xffffff7f8103e000 0x3000     0x3000     com.apple.kext.AMDRadeonServiceManager (4.0.7) 6E068EEE-90BD-32C8-82B3-984C59B706D6 <16 6 3 1>
  164    0 0xffffff7f82670000 0xf5000    0xf5000    com.apple.kext.AMDRadeonX4030HWLibs (1.0) 42AF6296-47C2-3FE3-9EFB-D35DFDCBBC6E <16 7 6 3 1>
  165    0 0xffffff7f81013000 0x26000    0x26000    com.apple.kext.AMDFramebuffer (4.0.7) 0E573FD7-F1EF-3237-86D1-12C11C80C1FA <129 108 16 15 9 7 6 3 1>

Is your Mac 6.1 D300 still working flawless on Monterey. Maybe I'll give it a try...

 

[mr1098]

Is your Mac 6.1 D300 still working flawless on Monterey. Maybe I'll give it a try...

6.1 with D300's here. Saw your post and figured I'd roll the dice on Monterey just for ... and giggles.

Uptime ~ 5 hours before my first KP.

Did notice (back in Catalina) that disabling the GPU in my browsers (Firefox && Brave,) does seem to keep me up days, instead of hours... so I'm back to that. (hat tip to mihai_me for that suggestion.)

Fedora 35 KP'd so quickly I sent that tester nvme back to Amazon the next day. Don't even bother with Linux.

I might actually go Win10 on this thing considering a previous post.

One thing is certain though -- I'll never buy nor recommend Apple again in this lifetime.

 

[ikir]

So do we know if it is a hardware or software issue? Those D300 are curse! I’m still on old firmware because third party SSD, I’ve ordered an original one so I can update to Monterey (and update firmware).

Does it happens on Mojave too?

 

[hirakujira]

I did some experiments for my D300 Mac Pro.

1. Install Monterey and play YouTube 4K video: crashes in 1 day
2. Install Mojave and play YouTube 4K video for 5 days: no more crash
3. Tried to install Windows 10 with bootcamp and run 3D benchmark apps to keep high usage of the GPU for over 7 days without any crashes.

And I noticed that the drivers are significantly different between Mojave and higher macOS versions. Perhaps that the driver is the cause?

BTW I'm 100% sure that it's not because of the heat (at least for my case): I tried to replace the thermal paste and set the fan to maximum, it didn't help a bit.

 

[ikir]

I did some experiments for my D300 Mac Pro.

1. Install Monterey and play YouTube 4K video: crashes in 1 day
2. Install Mojave and play YouTube 4K video for 5 days: no more crash
3. Tried to install Windows 10 with bootcamp and run 3D benchmark apps to keep high usage of the GPU for over 7 days without any crashes.

And I noticed that the drivers are significantly different between Mojave and higher macOS versions. Perhaps that the driver is the cause?

BTW I'm 100% sure that it's not because of the heat (at least for my case): I tried to replace the thermal paste and set the fan to maximum, it didn't help a bit.

Thanks! Do you have lastest firmware?
Do you have a 4K monitor? It seems users with low res display don’t have this issue.
Yes I cleaned all Mac Pro, changed thermal paste, also using a 115W CPU and not 130W it is not a temperature issue, maybe for someone yes but not all cases.

Do you have Original SSD?

 

[hirakujira]

Thanks! Do you have lastrate firmware?
Do you have a 4K monitor? It seems users with low res display don’t have this issue.

Original SSD?

Original SSD. But I think SSD doesn't matter.
I don't have a 4K monitor for work, but I tested it on a 4K TV and plug with HDMI cable.
On Windows, I just use windows update to install the latest driver of AMD graphic cards. I didn't download the driver from the official site.

 

[ikir]

If someone with the issue want to try lowering display resolution or using another non 4K display

 

[mikas]

I've got an uptime of 26 days now after full disassembly and maintenance round.
Mac Pro 2013 D300, original SSUAX with Big Sur.
249.60.3.0.0.
Apple 27" TB display (2560x1440)

Sleep disabled. After the maintenance I tried sleep, and it crashed pretty soon, this time garbling the Big Sur SSD non-bootable. I had to change in a fresh Big Sur SSUAX. Temps are pretty low in it, though with not much load on the CPU/GPU.

Previously I have tried 4K and even 5K2K monitors with eGPU, and had a lot of crashes with this D300. With D500 unfortunately it was almost the same. Not that frequent, but still crashes, even with sleep disabled.

Is it the shi**y driver. Is the bootrom. Is it the temps. Is the higher resolution monitor thing. Or is it the combination of any of those? Now it is stable with my config. At least until now. Wouldn't bet on it to suddenly have been overcome with crashes though.

So I don't know anything yet. Pretty hard to troubleshoot this, because it takes so much time and waiting and there are a lot of permutations with all of it.

ps. still love the looks of it, especially after seeing the new Mac Studio. But that's just a personal opinion of course.

 

[ikir]

Yeah temps are not an issue for me, I will try to put the original SSD and update to Monterey and make some tests. Waiting for it to be delivered.

 

[ikir]

this morning no issue with stress test and 4K video, at least for now. Just using a FullHD display instead of 4K.

 

[ikir]

Windows 10 with AMD Crossfire enabled push really good performance even on D300. Made some stress test on Windows and the machine seems stable and pretty fast. Very happy with E5-2695v2

no issue from yesterday in both macOS and Windows 10 with a non 4K display.

 

[mikas]

Would you trust a company who can't correct a sleep crashing problem in a supported machine in a few years time?
Yes, I thought so too.

After 29 days os uptime, light use no sleep, I just decided to put to sleep my 6,1 D300. Let's see what happens tomorrow or in a couple of days. Any guesses?

 

[ikir]

Would you trust a company who can't correct a sleep crashing problem in a supported machine in a few years time?
Yes, I thought so too.

After 29 days os uptime, light use no sleep, I just decided to put to sleep my 6,1 D300. Let's see what happens tomorrow or in a couple of days. Any guesses?

This happens to any other OS and to many machines. There so many unresolved issue on Windows and so many products with hardware flaw that will never be fixed. I trust Apple more than others tech giant, but sure they are not infallible.

 

[richmlow]

So do we know if it is a hardware or software issue? Those D300 are curse! I’m still on old firmware because third party SSD, I’ve ordered an original one so I can update to Monterey (and update firmware).

Does it happens on Mojave too?

My (late) 2013 Mac Pro runs on macOS Mojave 10.14.6. It has an upgraded 1TB OWC Aura Pro X2 SSD. See my signature for more details.

So far, no problems.

I know that many people have had countless problems with the (late) 2013 Mac Pro and I'm sorry to hear of the difficulties. However for me, the (late) 2013 Mac Pro has been the most reliable Mac that I've ever owned. It's even more rock-solid than my 2019 MacBook Pro (13-inch), running macOS Big Sur 11.6.4.


richmlow

 

[ikir]

It is been rock solid for now, still waiting for original SSD to perform firmware update and install Monterey. After cleaning the all machines and changing thermal paste, it has been super stable. Also the upgraded CPU is quite fast.

 

[mikas]

Report as of today:
32 days of uptime. Trying manual sleep now, and it's been working for 4 days now (32-28, please see earlier post). The best ever results without a crash for me with this D300 tcMP. So I wonder if the total hardware maintenance help, or is it the low resolution monitor (TB 27").