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

[kgp]

Hi kgp,
Do you have trim enabled?
If so this is good news indeed and maybe time for me to try the APFS again.
Mine used to stall for a bit on some encryption message (nothing to encrypt) then bail.

Hi Pete,

yes, I even have TRIM enabled.

Edit:
If you still face APFS boot issues, please try a clean install of 10.13.3 (17D2047) and you should be able to confirm that everything works as expected.

Good luck,

KGP

 

[nnorris7]

Thanks for reminding me. I read the insanelymac post on this, and I forgot all about it. That explains why I need EmuVariableUefi with old OsxAptioDrvFix.

Here it is. Does it help anything?

View attachment 309180

BTW, while rebooting with AptioMemoryFix.efi to capture this, I just found out that it works if I do a cold boot, but it does not work from a restart.

Hi @paulotex,

You might want to check if you have the correct kernel patches enabled/disabled as suggested to me by @Thireus in post #4418. Double-checking solved this issue for me.

Good luck!

 

[nnorris7]

Do you have a slide value set? If so, try removing it. AptioMemoryFix should be able to find the correct slide value, in fact I wonder why it is not trying, those tries result a lot of lines saying it is checking a slide value. Did you see those when it worked from a cold boot?

Also, post a memmap (efi shell, fs0: cd EFI memmap > memmap.txt) and we can check if you have too much fragmentation or not.

EDIT: Sorry, didn't realize these weren't posed to me. I'll leave my reply just incase it helps.

I believe I had the "xcpm_core_scope_msrs © Pike R. Alpha" patch disabled but was trying to boot from a non-patched BIOS. Now that my BIOS is patched and I have that patch disabled and the "10.13 Installer/Updater essential" patch enabled, I'm able to get past this line. I don't have any slide value set, so I assume AptioMemoryFix is now calculating it correctly. I never saw lots of lines saying it was checking slide values when it wasn't working.

Let me know if posting a memmap would still be helpful.

Thanks.

 

[paulotex]

Do you have a slide value set? If so, try removing it. AptioMemoryFix should be able to find the correct slide value, in fact I wonder why it is not trying, those tries result a lot of lines saying it is checking a slide value. Did you see those when it worked from a cold boot?

Also, post a memmap (efi shell, fs0: then memmap > memmap.txt) and we can check if you have too much fragmentation or not.

No slide set. And I don't remember seeing those lines of trying. Strange... I'll check it again. And the lines of "trying" don't show up. I wonder if I am doing something wrong, or if my motherboard is strange enough that does not work with this new AptioMemoryFix.

About my memmap, LOL, I did it weeks ago. No HW changes and no BIOS changes, so it should be the same:
https://www.tonymacx86.com/threads/...d-extended-guide.229353/page-247#post-1651616

 

[Thireus]

Hi Thireus.

I am with KGP here, I suspect your speedup might be because of a new installed system.

I have a NVME with my normal system (from 10.13.0 until 10.13.3, from iMac18,3 to iMacPro1,1) and a SSD with a freshly installed 10.13.3 with iMacPro1,1 for testing.

They are both APFS, so I am not really comparing with HFS+, but they both boot really quick. The full system takes about 5s longer than the fresh system, but they are both around 10-15s 25-35s boot time.

Hold my beer, I'll go and video them booting...

SSD booting from a fresh 10.13.3: 25s
https://drive.google.com/open?id=18vMhYSNRSi1HXWwQcoLNxUti7IASHk1k

NVME booting from my full system: 35s
https://drive.google.com/open?id=1oZ2lLz6Hh_Nzim2WCE1xC7DMYkOmlKHX

The main different is a 10s wait after FIPSPOST. This appeared between 13.2 and 13.3, still with the iMac18,3 SMBIOS. But it was never there on my freshly installed 13.3.

Both partitions on APFS.

It could be because of a fresh system install, indeed.

The previous APFS system that I had was running 10.13.3 (17D2047) but I installed it by upgrading from 10.13.2 (17C88). The new system I have now on HFS+ is running 10.13.3 (17D2047) but I used a fresh install from 10.13.2 (17C2120), I then upgraded it to 10.13.3 (17D2047) and used the migration utility to migrate my data.

But from the NVME video, this is pretty much the same speed I had under APFS. I can try to make a video with HFS+ to compare.

 

[kgp]

Well, first of all, you should be able to boot without any kernel patches if you use the iMacPro builds.

Secondly, the "can not allocate" error is an error thrown by aptiofix before clover loading boot.efi and boot.efi loads the kernel. So no kernel patch can solve the "can not allocate" error.

The "can not allocate" error is a result of two things:
- Memory below 100000000 pages can be very fragmented by memory mapped pci devices (basically your bios decides some memory are needed for certain pci devices to work. it COULD group them better so a bigger chunk remains free, but it does not.
- Kernel address space layout randomisation is a process where boot.efi tries to find a location for the kernel, a random location, so no malicious program can know for sure where the kernel is.

So when the random process selects a good value, all is well. When it isn't, you get can not allocate error.

It is possible to have such a fragmented memory map, that the kernel does not fit anywhere. Then you need OsxAptioFixDrv. That driver works around by relocating memory.

Or one can try it with OsxAptioFix3Drv and the slide boot flag to maintain native NVRAM.

 

[paulotex]

But from the NVME video, this is pretty much the same speed I had under APFS. I can try to make a video with HFS+ to compare.

Oh, I see. I guess I have low standards in boot time . So my 30s seconds are too slow for you. But notice that I don't experience any instability with APFS (as long as I keep the apfs.efi driver updated regularly).

 

[kgp]

It could be because of a fresh system install, indeed.

The previous APFS system that I had was running 10.13.3 (17D2047) but I installed it by upgrading from 10.13.2 (17C88). The new system I have now on HFS+ is running 10.13.3 (17D2047) but I used a fresh install from 10.13.2 (17C2120), I then upgraded it to 10.13.3 (17D2047) and used the migration utility to migrate my data.

But from the NVME video, this is pretty much the same speed I had under APFS. I can try to make a video with HFS+ to compare.

So the entire discussion finds it's continuation, even after repeatedly and clearly stating that my 10.13.3 (17D2047) AFPS System boots within 8 seconds after performing a clean install?

 

[Thireus]

HFS+ - 16sec boot time (after ++++ and before background loaded)

https://thireus.com/HACKINTOSH/HFS+_Boot_2TB_NVME_SAMSUNG_PRO_10.13.3_(17D2047)_CLOVER_r4392.mp4

APFS - 1min10sec boot time (after ++++ and before background loaded)

https://thireus.com/HACKINTOSH/APFS_Boot_2TB_NVME_SAMSUNG_PRO_10.13.3_(17D2047)_CLOVER_r4392.mp4

You can notice the slow down when the messages "FIPSPOST_USER" are displayed, but not only, the whole boot process is slow.

Both tests were conducted on the same NVME:

- 960 PRO 2TB NVMe M.2
- Same MacOS 10.13.3 (17D2047)
- Same CLOVER bootloader r4392

The only differences are that I did a fresh install on HFS+ today and migrated all my data with Apple's migration assistant. I've been having these slow boots since the beginning on APFS, so I thought it was normal...

Now to complete the test and see if this is due to a fresh install or to APFS, I'd have to perform a clean install again this time selecting APFS instead of HFS.

So, either the slow boot is due to the fact I did a fresh install directly on builds that support iMacPro1,1. Or it is because of the filesystem. Or both.

Maybe someone will be able to test these scenarios and check what's slowing down the system.

I don't think this is an outdated discussion, I believe many are using APFS with NVME here and are probably not realising that their system is a bit slow.

 

[paulotex]

Okay so your memory map is fragmented as hell. Are you on the latest bios?

You need 12b80 pages and the biggest available region is 15702 pages at 3A8FE000. The slide value for 3A8FE000 is 468, based on the recommended calculation, but that seems high. But it might worth a try.

Still, AptioMemory fix should be able to work this out.

Yes, I know it very fragmented. It has been like this through all the BIOS I've tried (3 or 4). And yes, I have the latest BUIS, 170 ATM.

Thanks for taking the time to calculate a slide for me. I did try doing that sometime ago (http://www.insanelymac.com/forum/to...d-post-installation-steps/page-2#entry2565119)
but I misunderstood the start value and choose a wrong one.

Unfortunately yours also did not help, I get the same behaviour. I wonder what is the problem. bdmesg shows that the driver is loading. Here are the lines in bdmesg that concern driver loading:

0:111  0:000  === [ LoadDrivers ] =======================================
0:111  0:000  Loading UsbMouseDxe-64.efi  status=Success
0:113  0:001   - driver needs connecting
0:113  0:000  Loading AptioMemoryFix.efi  status=Success
0:115  0:001  Loading OsxFatBinaryDrv-64.efi  status=Success
0:116  0:001  Loading UsbKbDxe-64.efi  status=Success
0:118  0:001   - driver needs connecting
0:118  0:000  Loading PartitionDxe-64.efi  status=Success
0:119  0:001   - driver needs connecting
0:119  0:000  Loading AppleImageCodec-64.efi  status=Success
0:120  0:001  Loading AppleUITheme-64.efi  status=Success
0:122  0:001  Loading DataHubDxe-64.efi  status=Success
0:124  0:002  Loading AppleKeyAggregator-64.efi  status=Success
0:126  0:001  Loading VBoxHfs-64.efi  status=Success
0:127  0:001   - driver needs connecting
0:127  0:000  Loading AptioInputFix.efi  status=Success
0:129  0:001  Loading FSInject-64.efi  status=Success
0:130  0:001  Loading Fat-64.efi  status=Success
0:132  0:001   - driver needs connecting
0:132  0:000  Loading apfs.efi  status=Success
0:134  0:002   - driver needs connecting
0:134  0:000  Loading FirmwareVolume-64.efi  status=Success
0:135  0:001  Loading SMCHelper-64.efi  status=Success
0:137  0:001  6 drivers needs connecting ...
0:137  0:000  PlatformDriverOverrideProtocol->GetDriver overriden
0:137  0:000  Partition driver loaded:
0:137  0:000  APFS driver loaded
0:137  0:000  Searching for invalid DiskIo BY_DRIVER connects: not found, all ok

You can see the AptioMemoryFix right at the beginning.

Oh, well, I'll stay with my weird OsxAptioFixDrv + Test combination...