Contribute
Register

How to build your own iMac Pro [Successful Build/Extended Guide]

Status
Not open for further replies.
Hi @borysbass, @kgp, or anyone who has working Thunderbolt Hotswap.

I have my TB3 device showing up and appearing to be fine with @kgp's method - but hotplug still does not work.
My only TB device is a UAD Apollo 8 going through the startech TB3 -> TB2 adapter. Could that be my issue?

I was wondering what devices you or anyone is testing their hotplug with?
 

Attachments

  • Erik’s iMac Pro.ioreg
    43.2 MB · Views: 190
  • SSDT-X299-iMacPro - EG.aml
    6.9 KB · Views: 93
  • Screen Shot 2018-03-22 at 7.26.12 PM.png
    Screen Shot 2018-03-22 at 7.26.12 PM.png
    226.9 KB · Views: 149
Please remove the Kext for testing purpose and test again.

Hi, sorry for the confusion: I did not mean the NoNvidia.kext. I meant it looks like the IntelMausi is looking for the onboard ethernet adapter. It seems this adapter cannot be found by the IntelMausi kext, because it got disabled by the ssdt.aml. Or it got confused before, when the CLI said: "ACPPI error: No Local Variables Are Localized For Method ..." Could it be that already here the resource looks table got messed up, which caused the IntelMausi error?
 
Hi, sorry for the confusion: I did not mean the NoNvidia.kext. I meant it looks like the IntelMausi is looking for the onboard ethernet adapter. It seems this adapter cannot be found by the IntelMausi kext, because it got disabled by the ssdt.aml. Or it got confused before, when the CLI said: "ACPPI error: No Local Variables Are Localized For Method ..." Could it be that already here the resource looks table got messed up, which caused the IntelMausi error?

you'll find that if you reboot again, it will likely hang in another spot. Which may throw your analysis on the cause off.
The DSDT removes pretty much all methods for the BR1A device, but doesn't disable it. I'm guessing that adding a STA method for it returning false would improve things.
 
you'll find that if you reboot again, it will likely hang in another spot. Which may throw your analysis on the cause off.
The DSDT removes pretty much all methods for the BR1A device, but doesn't disable it. I'm guessing that adding a STA method for it returning false would improve things.

Hmm, actually it stopped with or without Nvidia installed in slot 3 at exactly this spot twice.

Thinking about the whole approach: Could it be that with eGPU support in 10.13.4 we might be able to use the Nvidia card in slot 1 and the AMD card in Slot 3 with the following roles: Nvidia card via web drivers is the main GPU, also connected to a monitor via HDMI or DP. AMD takes ONLY the role of an accelerator card for FCPX to crunch the bits.

Thoughts?
 
I've used a nvidia eGPU with 10.13.2 with my mac pro ! (which has dual AMD D500)
10.13.4 improve things but there's nothing new there.
So no, I don't think it will make a difference
 
I've used a nvidia eGPU with 10.13.2 with my mac pro ! (which has dual AMD D500)
10.13.4 improve things but there's nothing new there.
So no, I don't think it will make a difference

Hm, interesting approach - so you could power Windows with the Nvidia eGPU and just disconnect it while running OSX?
 
Hm, interesting approach - so you could power Windows with the Nvidia eGPU and just disconnect it while running OSX?
No, unfortunately not.
eGPU isn't supported by Apple on mac with bootcamp.
It always gives an error about having not enough resources due to the lack of memory space (limited to 30 bits). While you can get around it by playing with the DSDT on some macs, I never managed to succeed with the mac pro 2013.
So in windows, I could only use the AMD

Works just fine on mac however... Hot plug was supported even...
 
No, unfortunately not.
eGPU isn't supported by Apple on mac with bootcamp.
It always gives an error about having not enough resources due to the lack of memory space (limited to 30 bits). While you can get around it by playing with the DSDT on some macs, I never managed to succeed with the mac pro 2013.
So in windows, I could only use the AMD

Works just fine on mac however... Hot plug was supported even...

Ah, never mind, I am on Hackintosh, so I can run native Windows ;-)
 
Hey Guys,

So, after even uninstalling my USB audio drivers from my audio interface and activating the onboard audio, I can still manage to lock up the system trying to activate core-audio in logic every time. Logic X runs fine without core audio activated and the graphics issues have vanished, as with I think most anything. but playing a video or trying to activate audio is impossible. the graphics card works fine, sleep works fine, all the USB ports are active, plugging devices into any of them works fine. So I am not sure what else to look out for. have I missed out on anything? The only thing I see in my system profiler PCI section is my Nvidia 1050Ti which is in slot 6 even after implementing the DSDT patches. Used the SSDT-X299-iMacPro.aml also in the patched folder but nothing else shows up in the PCI section of the system profiler. The Nvidia card still shows up in slot 6 even when it is plugged into slot one.

If somebody has some clues or ideas to what I might be doing wrong or any solutions id be very grateful.

Also, my system hung typing this message again! let me try to slap part of the crash report here as this seems to be the only one that keeps happening now.

Thanks.



Update: So I've managed to get audio running on the system and logic with some rather peculiar behavior with native logic plugins. and the bug whilst typing on this forum still freezes the computer. Made some progress.

Moved the soundcard to the rare USB2 port and unplugged my Fresco Logic Usb3 PCIe card.

The investigation continues...

Update: Eureka! Looks like I've managed to get a super stable system! Logic Works without crashing now! External USB soundcard performs well with no core audio issues! No graphics issues! going to see if I can recreate the problem to be certain what the cause is.


Update 2: So I finally figured out after everything that the Ram XMP mode was basically causing the system to hang randomly, turning that off made the system super stable, however having purchased 3000Mhz ram, running it at 2133Mhz was unacceptable.

Under close observation in the Ai Tweaker, I noticed a significant change in the AVX and AVX 512 frequencies when the XMP was disabled. Thanks to the super amazing @kgp and his super accurate guide, I was able to Apply the correct offsets to the AVX and AVX- 512 to reflect the correct readouts and still keep the ram at 3000Mhz.

I also decided to use the sync all cores anyways as it did make a big difference with DSM2 hyperthreading logic test and as suggested by kgp even without the XMP mode for the ram. However, interestingly my test ran better with a low I/O buffer size of 64 than 1024. It threw up overload messages at the latter. 64 is also the recommended I/O buffer size for the Symphony i/o. Everything looked great on the Intel Power Gadget Readouts also.

I haven't implemented the CPU core voltage adjustment as there is a new Bios for The X299 -A just released with improved microcode. will give this a try now and confirm results.

Sucess! Definite improvement for me with a patched 1301 on Geekbench. the system runs normally with all other settings in place.

There is a new option under /Advanced/CPU Configuration/ called MSR Lock control. which can now unlock MSR 3Ah, MSR 0E2h and CSR 80h.

will try this out now with a clean 1301 bios.

 
Last edited:
Status
Not open for further replies.
Back
Top