Contribute
Register

pastrychef's Asus ROG Strix Z370-G Gaming (WI-FI AC) build w/ i9-9900K + AMD 6600 XT

Thank you for your prompt reply pastrychef! I appreciate it a lot.

I installed RC scripts (Emuvar. was already there). No difference.

I have the ACPI patch if needed for this motherboard, thank you.

I have tested the NVME-theory by removing my second NVME drive, and leaving the primary (under the GPU) in, with macOS on it. No difference.

Because of the NVME-theory I removed all PCI-E cards, and only used the Vega64 and a SATA SSD for 10.14.5.
- Using the iMacPro1,1 SMBIOS (EFI attached) I find a score of about 16 seconds;
- Using the iMac19,1 SMBIOS (same EFI, just SMBIOS changed and iGPU enabled), it's about 14-16 seconds;

Thank you for the heads-up about the other topic.

Do you reckon your VGATab has a positive influence on your scores? Would you recommend, even though officially unnecessary for 10.14.5, for users to make their own profile, even just for power saving?

I attached my own EFI which gives these results.
 

Attachments

  • EFI.zip
    18.5 MB · Views: 222
Thank you for your prompt reply pastrychef! I appreciate it a lot.

I installed RC scripts (Emuvar. was already there). No difference.

I have the ACPI patch if needed for this motherboard, thank you.

I have tested the NVME-theory by removing my second NVME drive, and leaving the primary (under the GPU) in, with macOS on it. No difference.

Because of the NVME-theory I removed all PCI-E cards, and only used the Vega64 and a SATA SSD for 10.14.5.
- Using the iMacPro1,1 SMBIOS (EFI attached) I find a score of about 16 seconds;
- Using the iMac19,1 SMBIOS (same EFI, just SMBIOS changed and iGPU enabled), it's about 14-16 seconds;

Thank you for the heads-up about the other topic.

Do you reckon your VGATab has a positive influence on your scores? Would you recommend, even though officially unnecessary for 10.14.5, for users to make their own profile, even just for power saving?

I attached my own EFI which gives these results.

In your config.plist > CPU > Type, you have 0x1234. You can change it to 0x1005. It will help About this Mac identify your CPU correctly. It's just cosmetic, but it's better than 0x1234...

You have several different kexts for different NICs. You probably don't need them all, but having them there shouldn't cause any problems either.

What kind of Wi-Fi/Bluetooth card are you using?

Just as an experiment, can you try removing all your SSDTs except the UIAC one and run the BruceX test?

It's interesting that after pulling all your PCI-e cards you are now getting sub-20 second BruceX scores. This gives more credence to the theory that the issue is PCI-e lanes related.

It's probable that my VGTab tweaks have a bit of impact on compute tasks. This can be seen in the dramatic improvements in the LuxMark benchmark. It might be worth a shot to do a little tweaking of your own. At the very least, you should be able to help get your card running a little cooler.
 
Thanks for getting back on this, appreciate it! I made the CPU Type change as you noted. I use the second NIC kext for the second NIC on the Z390 Designare, I believe it's required to make both work, as either kext supports a different set of Intel NICs, much like using SmallTree for 10Gbit Intel NICs.

For WiFi/BT I'm using the BCM94360CS2 PCI-E x2, if I'm not mistaking. It's not a replacement for the onboard chip, as that has a special socket (CNVi), but it's also an add-in card with a USB2 cable to internal header for BT4. It's removed for testing purposes, just like the two M.2 drives for now.

I absolutely agree that there seems something quirky going on with the PCI-E interaction, but I don't imagine our Vega64 being reduced to 4x, which could potentially explain the drop in performance from 8x (typically when the system is loaded with M.2 and perhaps other PCI-E like my WiFi-card, and onboard TB3 enabled). In my understanding an OS should never influence the lane divisions of the motherboard, as this is controlled in the BIOS (also more sub-categories of PCI-E lane divisions like bifurcation of a single slot, etc.). It may for example be that macOS, in this specific Mojave version, has trouble enumerating PCI-E slots like it can have trouble iwth USB (controllers / ports / hubs). Then again, I could be wrong as I'm no true expert on this subject.


So back on the testing bench (10.14.5), my results are following:
iMac19,1, iGPU enabled, WEG: 12~15 seconds (from hitting the "Save" button to the point where the coil whine on the card stops, after a second or two Quicktime pops up, hence my ~15 seconds result; I don't see a progress bar).

iMacPro1,1, iGPU disabled, WEG: identical performance

The result also seems to vary per try on the same boot, with variance of about 2~3| !@ seconds. Watching iStats in iMacPro1,1 mode I see the Vega64 memory clocking typically at 500MHz, and sometimes shortly going up around 850MHz but staying below 1GHz for sure. Perhaps this is of influence, and how do you perceive the clock speeds when exporting BruceX in 12 seconds?

Using Sony's Food Fizzle 4K sample, I see a iGPU speed of up to 650MHz using iStats / Intel Power Gadget. When exporting BruceX: 0MHz (in iMac19,1 mode)


I made a VGTab that undervolts and retains roughly stock clock speeds, but a user here claims that macOS ignores any of these clock/voltage settings: https://www.tonymacx86.com/threads/...acos-without-flashing-the-vbios.268965/page-7. Regardless, I removed the kext to keep the benching as clean as possible.


I believe someone reported good BruceX speeds in 10.14.3? I just went ahead and first installed Catalina (beta 2, updated yesterday):
iMacPro1,1, iGPU disabled, WEG: identical performance (12~14, without variance)
iMac19,1, iGPU enabled, WEG: identical performance

Now let's add some PCI-E cards to test the theory. I check the BIOS everytime I reboot to add a part, and test BruceX.
- With just one NVME added, under the GPU: x16, x4 | 12~14sec
- Two NVME's: x16, x4, x4 | 12~14sec
- Two NVME's + WiFi: 16x, 4x, 4x, 1x | 12~14sec

So PCI-E devices are assigned lanes correcty in the BIOS, no worries about that.

After reverting back to 10.14.5 and 20 seconds per try, we can probably conclude that 10.14.5 has some issues. I don't know why they arise for some, and not for others, and it seems to be related to multiple PCI-E devices and how macOS handles them.

My solution: Stick to Catalina @ 19,1 and update from there. With updated kexts one doesn't need anything special compared to Mojave to run Catalina on Z390 and X299, it seems.

PS: The system can boot with iMacPro1,1 and iGPU enabled. There's no practical advantage, but the green flat line shows up.
 
After using the latest EFI folder (and changing the system definition to iMac Pro) I've noticed that the Mail.app stopped working - it lauches, but it doesn't draw/display any windows at all. Any idea how to fix that?
 
I finally got Mojave installed. Now I have a problem were the WIFI gets increasingly slow and finally get so slow that it can not connect to any websites etc. Any suggestions to how to solve this issue? It help when I reboot.
 
After using the latest EFI folder (and changing the system definition to iMac Pro) I've noticed that the Mail.app stopped working - it lauches, but it doesn't draw/display any windows at all. Any idea how to fix that?

I don't use Mail.app but I just launched it and everything looked normal. Do you have IGPU disabled in BIOS?
 
I finally got Mojave installed. Now I have a problem were the WIFI gets increasingly slow and finally get so slow that it can not connect to any websites etc. Any suggestions to how to solve this issue? It help when I reboot.

What kind of Wi-Fi card are you using?
 
@veirum I have exactly the same issue, and I couldn't fix it, had to use wired connection to get 'stable' internet to my hackintosh
 
@pastrychef I've disabled iGPU as per your instructions around changing profile to iMac Pro, but I've also tested enabling it and it doesn't seem to make any difference.
 
@pastrychef I've disabled iGPU as per your instructions around changing profile to iMac Pro, but I've also tested enabling it and it doesn't seem to make any difference.

Have you tried rebuilding the mailbox?


@veirum I have exactly the same issue, and I couldn't fix it, had to use wired connection to get 'stable' internet to my hackintosh

What kind of Wi-Fi card are you using?
 
Back
Top