Contribute
Register

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

Status
Not open for further replies.
yes, everything is OK however not much difference. I would more likely avoid any power saving settings and use the base frequency as the lowest one because in cinebench test the CPU jumps between 4300 and cca 2500MHz. THese jumps are responsible for the lower performance.

During Cinebench CPU benchmark, the CPU must remain at the defined max.load frequency! It it does not, improve your BIOS settings.
 
During Cinebench CPU benchmark, the CPU must remain at the defined max.load frequency! It it does not, improve your BIOS settings.

Here the picture from IntelGadget during the Cinebench. BIOS settings are those recommended from the page one for gigabyte MBs. I tried many different combinations but none of them keep the peak all the way up during the whole Cinebench test. I added also ssdt.aml which is in APCI patched folder. IT looks ok as well.
 

Attachments

  • IntelGadget.png
    IntelGadget.png
    53.8 KB · Views: 143
  • ssdt.aml
    3.4 KB · Views: 107
Here the puctire from IntelGadget during the Cinebench. BIOS settings are those recommended from the page one for gigabyte MBs.

Terrible XCPM performance and totally distorted cinebench frequencies typical for Gigabyte mobos.. However, one can do better as I have repeatedly seen along this thread. Please search for the respective posts and address your issue to the respective people, like e.g. @prunzi !
 
Terrible XCPM performance and totally distorted cinebench frequencies typical for Gigabyte mobos.. However, one can do better as I have repeatedly seen along this thread. Please search for the respective posts and address your issue to the respective people, like e.g. @prunzi !

thank you
 
Hello all!

Long time observer of this thread, but thought it was about time I posted myself :)

Firstly thanks, to KGP for the excellent guide, all also to all the peeps whos hard work have made this thread what it is!
You are legends.

This is my first hack. So in at the deep end.
Kit as follows:

7900x
Corsair 115i AIO Cooler (not overclocking and temps seem good may go EK eventually)
96GB Corsair Vengeance 3200 (Soon to be 128GB)
Dual GTX 1080ti's
1TB 960 Evo NVMe
x2 2TB 950 Evo SSD

To cap it all off I built my rig inside a modified G5 case:
IMG_20171123_141303.jpg

Absolutely ecstatic with the build, performance is ace.
Using for MotionDesign and 3D animation. Lovely to use Octane on OSX.
Benchmarks are as follows:
Screen Shot 2017-11-20 at 16.54.35.pngScreen Shot 2017-11-20 at 16.59.12.png

Only niggling issue is that I am occasionally experiencing some of the UI lag that other users have mentioned.
It does get better for a little while after sleep, but still slightly frustrating.
Anyone stumble onto a solution or have an idea whats causing the lag?

Again, thanks to all especially Mr. KGP.
Your all scholars and gents.

Cheers!
 
Last edited:
Thanks for the guide. installed HS on my rig (7900X, ASUS X299 TUF MK2, GT730) and has one issue.
USB External disks are shown as internal. How can I fix this?

Screen Shot 2017-11-24 at 1.57.14 AM.png
 
Thanks for the guide. installed HS on my rig (7900X, ASUS X299 TUF MK2, GT730) and has one issue.
USB External disks are shown as internal. How can I fix this?

View attachment 294401

No way to fix that for now.. you have to live with until it has been fixed by Apple or by some user... there are more important things in live though ... this is purely cosmetic error..
 
No way to fix that for now.. you have to live with until it has been fixed by Apple or by some user... there are more important things in live though ...
Thank you for fast reply. I appreciate.:wave:
 
Well my experience is limited at best, but here is what I think I have learned:

XCPM is activated for modern processors, and can not be disabled. XCPM can set the processor frequency regardless of any bios setting: you can disable turbo and EIST and even c states, and XCPM will work. From the min to the max frequencies. At this stage you get an error on verbose boot informing you that cpu pm might be incomplete.

If you have a plugin type:1 for the first cpu in your cpu ssdt, xcpm will also load the frequency vectors, and is supposed to run even better. With this, the error disappears. Oh BTW ssdtprgen frequencies doesn't matter. If you remove from them, those p states are used anyway. I think xcpm gets the min/max frequencies from somewhere else, the bios, probably.

With these settings, my system is works okay, but stays at 1.2 Ghz most of the time and my perceived OS speed is slow.

There is also the matter of HWP (a.k.a. SpeedShift), or hardware controlled P states. My understanding is that this technology should enable faster speed changes as the os does not set p states, the cpu does, and does it really fast.

While HWP should not work with iMac17,1 (it is not activated by iMac17,1's cpu freq vectors), setting SpeedShift on/off in the bios does have effect on cpu freqs.

- If it is on, cpu goes to 1.2 Ghz, turbos properly to the max with any benchmark, benchmark speeds are fine. But OS feels slow. (After the first sleep, OS responses seem to speed up, but nowhere near good enough)
- If it is off, cpu hovers around 2-3 Ghz, thus feels faster, but under benchmarks, frequencies are all over the place for no reason, and results are bad.

Sadly, I have not found a way to:
- disable all CPU PM to see how my system would feel at max speed
- check if HWP is working
- activate HWP if it is not

So if you want the best benchmark results (or good performance under heavy load), try enabling SpeedShift.

I have also been wondering this exact same thing. This why I asked a few posts ago about a way to disable XCMP. Adobe Lightroom typically isn't doing much but when I move to the next photo there seemed to be a larger than normal lag from photo to photo than my old 4770k build. Since the CPU is sitting at 1Ghz I was wondering if the load wasn't enough to ramp up the CPU (or if there were a slight delay in the ramp up). I too have tried every setting on BIOS/With/Without SSDT and had similar results as you... although I am still missing my custom water cooling loop so I do run into thermal throttling on the high end of clock speeds which has made my testing a little more difficult.
 
Status
Not open for further replies.
Back
Top