Contribute
Register

Mac OSX 10.12 with X99 Broadwell-E family and Haswell-E family

Status
Not open for further replies.
Tried to upgrade to 10.12 Final last night and failed, so did a complete reinstall, had to boot with NullCPUPowerManagement and VoodooTSCSync and Haswell FakeCPUID and flag, but after that i was able to change FakeCPUID to Broadwell-E and remove the kexts and now have XCPM working again... weird

Any way just a little note for people upgrading or trying from a clean install, thanks nmano for OP
 
Tried to upgrade to 10.12 Final last night and failed, so did a complete reinstall, had to boot with NullCPUPowerManagement and VoodooTSCSync and Haswell FakeCPUID and flag, but after that i was able to change FakeCPUID to Broadwell-E and remove the kexts and now have XCPM working again... weird

Any way just a little note for people upgrading or trying from a clean install, thanks nmano for OP

Interesting. I may be upgrading to Broadwell-E in the near future. You are saying that FakeCPUID is required for Broadwell? If so, what are the ID values for the different models?

Thanks
 
Yes FakeCPUID is needed for Broadwell-e, at first i was stumped because "basically" everybody could delete FakeCPUID, i'm using:

0x040671

I just guessed it was the first chip of the broadwell-e

maybe:

0x040671 - 6800K
0x040672 - 6850K
0x040673 - 6900K
0x040674 - 6950X

who knows though just a guess

Also i'm only using rebootfix in my kernel to patch in clover
 
I swapped my UD4 mobo with the newest GA-X99-Gaming 5P. What a difference! the UD4 was a real POS. The Gaming 5P does 3000 DDR4 memory with no issues and the LAN is much faster than UD4. I have everything working except for sound, but don't plan on trying since I have a USB sound card.

Next up is a new Broadwell-E in the next couple months.
 

Attachments

  • GB9-29.jpg
    GB9-29.jpg
    99.4 KB · Views: 178
  • heaven.jpg
    heaven.jpg
    73.8 KB · Views: 192
  • disk.jpg
    disk.jpg
    56.1 KB · Views: 159
My system works well with the versions 10.12.1 Beta (16B2327e) and 10.12.1 PB 2.1 (16B2333a), without USB problems or other issues, but still no support from Nvídia for now. And no graphics acceleration 4k, I think I'll go back to the final version 10.12 again!
By the way, upgraded the RAM to 48GB 6x8GB, the OS only detect 40GB 5x8GB, but the Board UEFI Bios shows the correct amount in the correct place.
Funny is that the operating system recognizes that the board has 8 slots, but does not recognize that are occupied 6 instead of 5, or a total of Ram installed.:banghead:
 

Attachments

  • 12.png
    12.png
    265.4 KB · Views: 196
  • 13.png
    13.png
    142.8 KB · Views: 213
  • 14.png
    14.png
    133.2 KB · Views: 197
Last edited:
My system works well with the versions 10.12.1 Beta (16B2327e) and 10.12.1 PB 2.1 (16B2333a), without USB problems or other issues, but still no support from Nvídia for now. And no graphics acceleration 4k, I think I'll go back to the final version 10.12 again!
By the way, upgraded the RAM to 48GB 6x8GB, the OS only detect 40GB 5x8GB, but the Board UEFI Bios shows the correct amount in the correct place.
Funny is that the operating system recognizes that the board has 8 slots, but does not recognize that are occupied 6 instead of 5, or a total of Ram installed.:banghead:

I was unaware you can run memory is less than x4 increments for OS X. Are you sure that it's supported? Even if you could, that's less than ideal performance wise.
 
Fixed unsupported GPU Error After Effects with CUDA

1-Install Cuda
2-Edit to raytracer_supported_cards.txt
Just add your device on top

GeForce GTX 970
GeForce GTX 285
GeForce GTX 675MX
GeForce GTX 680
GeForce GTX 680MX
GeForce GT 650M
Quadro CX
Quadro FX 4800
Quadro 4000
Quadro K5000
/Applications/Adobe After Effects CC 2015/Adobe After Effects CC 2015.app/Contents/Resources/raytracer_supported_cards.txt

Fast Draft: Available
Texture Memory: 1638.00 MB
Ray-tracing: GPU
OpenGL
Vendor: NVIDIA Corporation
Device: NVIDIA GeForce GTX 970 OpenGL Engine
Version: 2.1 NVIDIA-10.13.65 367.15.10.05f01
Total Memory: 4.00 GB
Shader Model: -
CUDA
Driver Version: 8.0
Devices: 1 (GeForce GTX 970)
Current Usable Memory: 2.66 GB (at application launch)
Maximum Usable Memory: 4.00 GB
 

Attachments

  • Screen Shot 2016-10-01 at 8.50.10 AM.png
    Screen Shot 2016-10-01 at 8.50.10 AM.png
    68.1 KB · Views: 161
  • raytracer_supported_cards.txt
    149 bytes · Views: 316
  • Screen Shot 2016-10-01 at 9.21.33 AM.png
    Screen Shot 2016-10-01 at 9.21.33 AM.png
    103.6 KB · Views: 140
I was unaware you can run memory is less than x4 increments for OS X. Are you sure that it's supported? Even if you could, that's less than ideal performance wise.

Hi lokicat
I did as described in the board manual and was placed in the recommended slots, to not lost performance. Windows all ok, but in OS X does not. Either way I intend to increase to 64 GB and fill all the slots. I hope doing that, OS already recognize all. (or not);)
 
Hi lokicat
I did as described in the board manual and was placed in the recommended slots, to not lost performance. Windows all ok, but in OS X does not. Either way I intend to increase to 64 GB and fill all the slots. I hope doing that, OS already recognize all. (or not);)

When you populate in x2 increments, then your memory controller defaults to x2 bandwidth. X99 is capable of x4 bandwidth but requires every memory dimm to be identical and installed in x4 groups. I don't know if OS X supports both x2 and x4 like Windows. As for performance, it really depends on what you are going to to but for sure you are cutting your memory bandwidth in half by populating in x2. However, the cache of the Haswell-E processors is pretty large and in many cases, overall performance is not memory bandwidth constrained. Definitely report back if going all x4 resolves the issue.
 
One more note to share - in order for my system to run 3000 mem speed, the CPU BCLK needs to increase to 125. This will cause the Intel Power Gadget utility to show the wrong CPU speed since that is based on a 100 BCLK. It's a simple cosmetic issue and your CPU will be running at the right speed which you can confirm with running GB benchmark. I have an older version of Power Gadget on my Windows drive and it does the same thing.
 
Status
Not open for further replies.
Back
Top