I have determined that the correct rtcfx_exclude range (for use with RtcMemoryFixup.kext) for my EliteDesk 800 G3 Mini is not a contiguous range (e.g. not B0-B4). The correct range for my G3 Mini is
rtcfx_exclude=B0-B3,B7
I determined this by trial and error when attempting to wake from sleep. I haven't yet gotten wake from sleep to work on the G3 Mini (Kabylake, i7-7700T, HD630), but with this rtcfx_exclude range, attempts to wake from sleep do not corrupt RTC.
I'm posting this in this thread, because the RTC issue on the EliteDesk 800 G4 Mini is hard to reproduce. Maybe the range for the G3 Mini will work (or at least be a good starting point) for the G4 Mini.
EDIT: When framebuffer-conX-type is set to Single-link DVI as described below, Hackintool crashes when selecting Patch > Connectors. It appears that Hackintool is not compatible with Single-link DVI setting. Since Dual-link DVI appears to work fine even when actual connector is Single-link DVI, it is best to leave framebuffer-conX-type setting at Dual-link DVI if you use Hackintool.
--------------------------------------------------
I just learned something new after reading this Dortania page: the DVI connector type 0x0004 that I've been using is for dual-link DVI. My DVI connectors are single-link (display resolution 1680x1050), so my framebuffer-conX-type should be 0x0200 (reverse byte order 0x00020000). I am now running with this revised connector type and am not seeing any difference in behavior. I suspect that this doesn't matter for me, because my display resolution is below 1920x1200 as discussed here. For those with DVI displays having resolution higher than 1920x1200, you would need dual-link DVI.
While upgrading to Big Sur 11.1 with OC 0.6.3 (here) I have confirmed that the rtcfx_exclude range discussed here resolves RTC corruption issues with the HP EliteDesk 800 G4 Mini.
Note that my baseline is still CLOVER r5122 with Catalina 10.15.7. I continue to run Big Sur / OC only in my test environment with a separate SSD.
I attached a revised CLOVER r5122 EFI (EliteDesk 800 G4 Mini - CLOVER r5122-v6.zip) archive to Post #1. This archive for CLOVER r5122 / Catalina includes the following updates:
Updated AppleALC, IntelMausi and WhateverGreen kexts
Changed boot-arg darkwake=3 to darkwake=2 (back to my original - I had changed to 3 based on a suggestion, but it didn't make a difference)
Added rtcfx_exclude=B0-B3,B7 (for RTCMemoryFixup.kext) to resolve RTC issues with macOS installer (still testing this exclude range)
Disabled TRIM kernel patch (I only have NVMe SSDs. You may need to enable if you have SATA SSDs)
This archive no longer includes an LE folder (all kexts are in EFI/CLOVER/kexts/Other in the archive). If you choose to install kexts in /Library/Extensions (I still do for Catalina), you will need to select /L/E kexts from the E/C/k/O folder. Also, if you use my CLOVER config.plist, you will need to replace all instances of XX-MASKED-XX with your own values.
Thanks to @viorel78 for showing me how to examine the boot log for ACPI errors, I found that CLOVER r5122 is injecting a bad SBUS Fix (causes ACPI parse error on SBUS.BUS0._DSM) when CLOVER's "Fix SBUS" is enabled. The solution for CLOVER r5122 is to use the attached SSDT which we are using for OpenCore and to disable CLOVER's "Fix SBUS" and disable CLOVER's "Add MCHC." I upgraded to CLOVER r5127, but the problem remains (not fixed in CLOVER r5127). This fix will be incorporated into a future update of the CLOVER EFI attached to Post #1. I will not be fixing this, since I am switching to OpenCore.
EDIT2: I have reproduced this interesting CPU performance behavior with OpenCore here.
EDIT: The Geekbench 5 benchmarks below with CLOVER r5127 were collected with SSDT>Generate>PluginType enabled in Clover's config.plist and without SSDT>PluginType=1. While performance is slighly better (as measured with Geekbench 5), most Energy Saver options (in Energy Saver Preferences) are not available unless SSDT>PluginType=1 is also set in Clover's config.plist. With SSDT>PluginType=1, performance with r5127 is the same as with r5122. Also, while testing with r5127 and only SSDT>Generate>PluginType (without SSDT>PluginType=1), I experienced a corrupted SSD that required me to completely erase and recover from TimeMachine backup. I can't be certain that this catastrophic failure was caused by incorrect power management settings, but I would not advise using Clover r5127+ without setting both SSDT>Generate>PluginType and SSDT>PluginType=1. Note that this requirement to add SSDT>PluginType=1 is new after Clover r5122.
-----------------------------------------------------------
Posting my Geekbench 5 scores captured before and after CLOVER r5127 upgrade. I can't say for certain what contributes to the increased multi-core performance, but I don't believe that I've ever had a GB5 multi-core score over 5550. This latest is 5611.
I have switched back to CLOVER r5122 and have confirmed that my Geekbench5 benchmarks have returned to the r5122 levels that I reported here. My system performance is definitely better with r5127. Also, my minimum CPU multiplier with r5122 is 13x and it is 10x with r5127 (max multiplier is still 46x). EDIT: This strange CPU performance behavior is because CLOVER r5127 handles plugin-type differently than r5122. r5127 now requires SSDT>PluginType=1 in addition to SSDT>Generate>PluginType.
One issue with r5127 is that Energy Saver settings are "lost" - with Clover r5127, only "Prevent computer from sleeping..." and "Put harddisks to sleep..." are available options. With r5122, the options are as pictured. I think this is a Clover r5127 issue, but have yet to convince the Clover Team.
EDIT: This CLOVER PluginType bug is fixed in a commit after the official release of CLOVER r5128.
In the other forum's "Clover General Discussion," @iCanaro reported that there have been issues with Clover r5127 Power Management. Advice is to avoid Clover's SSDT->Generate->Plugintype and to use SSDT-PLUG (just like we do for OpenCore). Soon, we won't be able to tell the difference between OpenCore and Clover
EDIT: Due to a change in Clover r5127+, the Clover config.plist requires "SSDT > Plugin Type = 1" in addition to "SSDT > Generate > PluginType." With this Clover config.plist change or with SSDT-PLUG.aml (same as used with OC), the Energy Saver issues are resolved. Unfortunately, the increased multi-core performance and lower CPU multiplier are gone when this "fix" is implemented. Maybe someone will find a way to retain the fix and the performance.
My current baseline for this rig is still Catalina 10.15.7.03 with CLOVER r5122. After much effort to maintain two candidate baselines for both CLOVER and OC, I have decided to drop CLOVER in favor of OC as I migrate to Big Sur. I'll migrate to Big Sur when it reaches incremental version 4 and plan to have an OC bootloader solution at that time. Many thanks to the CLOVER Team for their great work and to everyone in this forum for the outstanding CLOVER support. It's been an amazing journey!
To switch from CLOVER to OC, do the following:
Make a full system backup before you convert
Open Hackintool and collect the following info (collecting this from your CLOVER config.plist is possible, too, if you can avoid the confusion of RtVariables.MLB and SMBIOS.BoardSerialNumber):
Serial Number (this will be OC's PlatformInfo.Generic.SystemSerialNumber)
System ID (this will be OC's PlatformInfo.Generic.SystemUUID)
ROM (this will be OC's PlatformInfo.Generic.ROM)
Board Serial Number (this will be OC's PlatformInfo.Generic.MLB)
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
