Gigabyte X299X - Catalina Support

[byteminer]

Sounds good - what loop kit have you got? I'm still running on my CoolerMaster 360mm AIO. I've spent nearly two weeks testing and tuning the overclock, and talking to the overclocking experts over at overclock.net. There's a 90+ page thread there on overclocking the 10980XE in which I've learnt a lot.

Based on that I've now decided to take the plunge and go full custom loop. Never done it before, but I thought given I spent so much on the system, and am being limited by my AIO in terms of how much I can OC, I'd go the full distance. I'm hoping to get a large external radiator, the MO-RA3 420 (9 x 140mm fans!). CPU block I've not 100% decided yet; I'd really like the Optimus Signature V2 but it's very expensive and only available in the US. I just learned about a new company called TechN who are based in Europe and have a 2066-mount block that could be good.

Funny you would mention that, because I'm rocking a fully custom loop in a 909EK (the one with a distro plate), coincidentally with a TechN waterblock!
It works extremely well, I've had it pulling almost 500 watts at times and it still cools pretty well. Got two thick 480 radiators (Black Ice Nemesis 480GTX) from HardwareLabs and they work extremely well. Unless I run it for hours at a time on full load I can basically run with the fans off or on minimum.

As far as the EFI goes, thank you for providing it I'll check it out when I get back from work.

 

[TheBloke]

Funny you would mention that, because I'm rocking a fully custom loop in a 909EK (the one with a distro plate), coincidentally with a TechN waterblock!
It works extremely well, I've had it pulling almost 500 watts at times and it still cools pretty well. Got two thick 480 radiators (Black Ice Nemesis 480GTX) from HardwareLabs and they work extremely well. Unless I run it for hours at a time on full load I can basically run with the fans off or on minimum.

Ah very nice! My current peak power usage is around 370W, eg during RealBench stress tests in Windows. I've topped 400 with some settings but then I run into throttling.

So you've got this water block TechN CPU Waterblock Intel LGA 2066 / 2011? I only learned about them yesterday, and gather they're pretty new. Can't find any reviews of the Intel block but an AM4 block beat out the EK Magnitude and HeatKiller IV Pro on a German website's review.

If you have that block and are getting good results, maybe I'll go for that. What frequencies are you able to reach? Are you going all-core, or some cores higher than others? When I was testing in Windows I ended up with 6 cores on 4.8Ghz and the rest on 4.6Ghz, but right now I'm running macOS with a simpler all-core 4.6. I used the "X-Cores Active" setting to achieve that, ie set the BIOS such that if 1, 2, 3, 4, 5 or 6 cores were active they'd be at 4.8Ghz, but if it was 9 or more they'd all run at 4.6.

I'm hoping once I go full custom I can get at least all-core 4.8, and 4.9 isn't out of the question. I don't know how good my silicon is. At first I thought I'd done badly in the silicon lottery because I need at least 1.27V to get stable all-core 4.8 (which throttled all the time on my cooling); any lower voltage was a quick BSOD or freeze in Windows stress testing. But then I was told that lower temperatures can enable lower voltages, so now I don't know. Maybe once I have better cooling those voltages will go down. I've been told that 1.2 volts is 'average' for all-core 4.8.

As far as the EFI goes, thank you for providing it I'll check it out when I get back from work.

Hope it gets you a bit further! I forgot to mention, I'm testing exclusively with macOS 10.15.7. I've not tried Big Sur yet and won't until it's out of beta, maybe not even until 11.0.1 or 11.0.2 reaches gold. I've been running 10.14.6 for the last 1+ years so 10.15.7 is already new for me

 

[byteminer]

If you have that block and are getting good results, maybe I'll go for that. What frequencies are you able to reach?

The block so far has been great. Of course I don't have anything to compare it to, but heat transfer so far seems exceptionally good to me. I don't have any hard numbers at the moment since I'm still trying to get my OpenCore boot to be stable, but I'll post some results after that is done.

In my limited testing thus far I was able to reach around 4.8 relatively easily on acceptable voltages, unless Prime95 comes into the equation, which really puts a dampener on my OC. With Prime95 I'm currently sitting around 4.5, 4.6, but that's mostly due to not spending enough time tweaking the OC yet. Still trying to balance VCCIN with the core voltage as well.

 

[TheBloke]

Great to know re the block, thanks.

In my limited testing thus far I was able to reach around 4.8 relatively easily on acceptable voltages, unless Prime95 comes into the equation, which really puts a dampener on my OC. With Prime95 I'm currently sitting around 4.5, 4.6, but that's mostly due to not spending enough time tweaking the OC yet. Still trying to balance VCCIN with the core voltage as well.

Yeah if it's the latest Prime95 then that's all AVX and that's murder for the CPU as I understand it. There's actually warnings in that Overclock.net thread against running long Prime95 tests at high voltages because it's so demanding, more so than real life AVX workloads. As a result I've been testing AVX with AIDA64 - it has checkboxes to enable or disable AVX and AVX512 computation.

Are you using the AVX offsets in the BIOS to set the lower frequencies for Prime95? If not, then I'd recommend trying them - that's what they're for, to enable running a lower frequency for AVX while still full speed for non-AVX. Right now I need -2 AVX and -6 AVX512. Ie I'm running 4.6Ghz for non-AVX, 4.4Ghz for AVX2 and 4.0Ghz for AVX512. I hope to need lower offsets once I have better cooling.

AVX512 in particular was a major pain - I found that even with high offsets, I had to also bump up my vcore else an AVX512 test would always fail. So I'm running a higher voltage than I would if AVX512 didn't exist, just so it doesn't lock up when AVX512 is used. Likewise I had to boost VCCSA, VCCIO and Mesh voltages if I wanted to sustain an Mesh/Uncore overclock while doing AVX stuff.

Right now I'm running all-core 4.6 with XMP 3600 and Uncore 30x. I've had 32x working but I've backed that off for now until I've got better cooling. I've currently got Mesh at 1.3V, VCCSA and VCCIO at 1.25V, VCCIN at 1.92V and LLC at Medium. I don't really need such high voltages for 4.6 and Uncore 30x but I whacked them up for benchmarking, to be sure I wasn't hitting any throttling.

With this setup my best macOS GeekBench 5.2.5 score is 1229 single, 18026 multi. Best Windows score is 1203 single-core, 18013 multi. At stock my scores are 1199/15698 in macOS, so that's approx +15% on multi-core.

Cinebench shows a bigger improvement, with Windows scores up from 8658 at stock to 10507, an increase of 21%. My recorded macOS scores show a similar % increase, but the absolute scores are consistently 4-5% lower than in Windows, which I've not yet figured out. I suppose it's not impossible that Cinebench just performs better on Windows for some reason.

So I've got some useful performance improvements compared to stock, but I hope to do better once I've got a custom loop.

 

[TheBloke]

OpenCore EFI updated to 0.6.3, and latest Acidanthera kexts

I just noticed that OC 0.6.3 was released yesterday, along with new versions of the Acidanthera kexts.

EDIT 2020-11-06

For anyone just finding this post, please be aware that while this EFI should get you booted and running macOS reasonably well, it has at least two issues:

> CPU power management not fully working - there is functioning frequency adjustment, but X86PlatformPlugin is not attached so it may not be working perfectly
> Sleep not working at all - this issue has been traced to the SBUS SSDT. Disabling this SSDT will enable sleep, but wake is not yet working.

These issues are being worked on and an improved EFI will be uploaded in a later post.

END EDIT

I've applied these updates to my system, and they seem to work fine. So far I have not changed anything in my config.plist. I will later have a look in more detail at the few config changes 0.6.3 brings to see if there's anything that might be useful to our systems.

Attached is my EFI updated to 0.6.3, with the following kext updates:

• Lilu 1.4.4
• WhateverGreen 1.4.4
• VirtualSMC 1.1.8
  • SMCProcessor
  • SMCSuperIO
• AppleALC 1.5.4
• AirportBrcmFixup 2.1.1

I've not so far noticed any difference from the upgrade to 0.6.3 or the kexts. The boot-from-BIOS issue I described still exists, as does the issue where my VCCSA resets to Auto after shutting down in macOS.

As before, PlatformInfo serials are changed to CHANGEME.

One thing I forgot to mention @byteminer : in both the EFIs I have shared, I have removed agdpmod=pikera from the boot options as I have a Vega 64 GPU on which it is not required. However I think it is required on your RX580? So you might want to put that back in before you boot, else graphics might be affected.

The only difference I noticed from removing agdpmod is that I now get an Apple logo progress bar in the final stage of booting, prior to the appearance of the login screen. With agdpmod=pikera, I instead got a black screen at that point.

Final thing to mention: both this EFI and my earlier 0.6.2 EFI are using the OpenCore RELEASE version. Kexts are also all RELEASE. So you won't get any debug logging. I plan to try a debug version later with logging enabled to see if I can learn anything about my boot-from-BIOS failure, but haven't got around to that yet.

EDIT: I've updated the EFI with the following minor changes:

> Added OpenCanopy and the necessary resources for it to work
> Removed BT4LEContinuityFixup.kext and enabled the ExtendBTFeatureFlags Quirk which removes the need for the kext

 

[TheBloke]

I just updated the 0.6.3 EFI ZIP in my previous post to add OpenCanopy, and remove a deprecated kext (BT4LEContinuityFixup). Details added as an edit to the previous post.

 

[byteminer]

Alright, I had a little bit of time to mess around with this some more and I'm happy to report that your updated EFI works marvellously!

I made a few tweaks, namely:
- Switch SMBIOS to iMacPro1,1 because that's closer to the actual hardware we're running and I already had the data for that generated
- Update NVMEFix to the latest version (1.0.5)
- Add SmallTree driver kext from the OP to get working Ethernet
- Add itlwm to make the AX200 WiFi work (moderately well at least)

I also figured out that the OpenCore Configurator app -something that has worked well for me on my Z390 build- screwed up my config every time I used it. As such I went back to configuring everything by hand and that has drastically reduced the amount of boot failures. I diffed the config.plist files that the app generated, and it added a lot of previously not-specified elements with questionable defaullt values, some of which likely ruined the config.

Last but not least I also did a tiny bit more overclocking, this time running macOS. I was able to successfully breach 10.700 points in Cinebench R20 at 4.8GHz. Temperatures are still under control, somewhere in the mid-70's while the chip pulls about 400 Watts. I can probably optimize that a bit more, but it'll do for now. I'll probably stop at 4.8/4.9 no matter what though, since I might need to run it at full load for a few days straight and I really don't need yet another space heater.

 

[TheBloke]

Alright, I had a little bit of time to mess around with this some more and I'm happy to report that your updated EFI works marvellously!

I made a few tweaks, namely:
- Switch SMBIOS to iMacPro1,1 because that's closer to the actual hardware we're running and I already had the data for that generated
- Update NVMEFix to the latest version (1.0.5)
- Add SmallTree driver kext from the OP to get working Ethernet
- Add itlwm to make the AX200 WiFi work (moderately well at least)

That's great, glad to hear it's working!

I've got the SmallTree drivers in /L/E rather than as part of the EFI. I guess it works either way.

I'd not heard of itlwm, that's great to hear that there's a way to get the standard WiFi working. I don't plan to use WiFi as I'm always cabled, but I know of at least one app that I couldn't use because it expected to find a WiFi connection even when cabled on Ethernet (Duet, an app that allows using an iPad as another monitor).

I also figured out that the OpenCore Configurator app -something that has worked well for me on my Z390 build- screwed up my config every time I used it. As such I went back to configuring everything by hand and that has drastically reduced the amount of boot failures. I diffed the config.plist files that the app generated, and it added a lot of previously not-specified elements with questionable defaullt values, some of which likely ruined the config.

Yeah I used to use Clover Configurator, and I like the idea of having a UI configurator. But I know the developers of both Clover and OC hate the Configurator tools and say they will mess up the config. So I've been configuring OpenCore manually, primarily in Visual Studio Code with an XML validator plugin installed, but using ProperTree when I need to edit binary data.

If you're not aware, there's a great OpenCore 'sanity checker' online at opencore.slowgeek.com. You specify your platform and OpenCore version, then upload your config.plist. It checks it both against OpenCore syntax and against the recommendations for the selected platform.

I ran it when I was first setting up OC 0.6.2 and it helped me fill in some config values I'd missed. The site was just updated for 0.6.3 and I just re-ran it and I'm missing one new property in Output ("ForceResolution"). Also I have a few settings that it thinks aren't recommended for Skylake-X, such as DevirtualiseMmio = No which it thinks should be Yes, and EnableWriteUnprotector = Yes which it thinks should be no. I can't recall why I have those settings as they are, but obviously things are largely working as-is.

In the next few days I'm going to read the OpenCore documentation from cover to cover and try to understand a bit more about these settings and try to diagnose my remaining issues.

Last but not least I also did a tiny bit more overclocking, this time running macOS. I was able to successfully breach 10.700 points in Cinebench R20 at 4.8GHz. Temperatures are still under control, somewhere in the mid-70's while the chip pulls about 400 Watts. I can probably optimize that a bit more, but it'll do for now. I'll probably stop at 4.8/4.9 no matter what though, since I might need to run it at full load for a few days straight and I really don't need yet another space heater.

Great score! What are your overclock settings? Are you adjusting VCCSA, and if so are you finding it gets reset to Auto after a macOS shutdown? Specifically, after selecting shutdown in macOS (but not restart), the System Agent voltage will be running at its Auto setting, ie around 0.9V. The BIOS setting will still be set at whatever you change it to, however on the right side of the BIOS screen it will show VCCSA at around 0.89 - 0.9V, instead of whatever it should be.

Also, when you have a moment could you test to see if you have the same issue as me: can you try booting OpenCore after having accessed the BIOS? So go into the BIOS and hit enter on the Boot Override entry for OpenCore. And/or go into the BIOS then Exit Without Saving, then let OpenCore boot automatically.

In both scenarios I will get a shutdown at the end of OpenCore initialisation, and when it comes back up I will find the system is running at default CPU frequency and voltages, and RAM is at 2666Mhz, ignoring XMP.

If you don't have these issues, then I might try iMacPro 1,1 as it sounds like that's the only major thing you've changed.

 

[TheBloke]

Quick question guys: So is anyone here running a dual boot system with MacOs and Windows 10? Everytime I start my computer i'm still greeted by the dreaded error message:

I posted about that on Monday, here and here. Short answer is that I don't get that BIOS failure message and I am able to boot both macOS and Windows successfully. I don't yet have a working OpenCore-based dual-boot, which I believe is down to the fact that the SSDTs need to be edited to only run for macOS, as OpenCore will inject SSDTs regardless of OS (unlike Clover).

I currently have two NVMe drives installed. The first has OpenCore EFI bootloader and macOS partition, the second has Windows EFI bootloader and Windows 10 partition.

I boot macOS from OpenCore, but I can't yet boot Windows from OpenCore: Windows starts booting OK, but then quickly blue-screens with an ACPI error - very likely due to the SSDTs being applied to it when they should not be. To fix this will require editing each of those SSDTs, which I am looking at now.

To boot Windows successfully I currently need to select its bootloader entry in the BIOS Boot Override menu, which works fine.

Therefore I do have a way to boot both OS, just not yet from a single menu. I hope to get that working soon.

"Boot failure detected. The system has experienced a boot failure possibly due to incorrect configuration. Previous setings in BIOS may not be compatible with current hardware state."

So every single time at boot, I need to"enter bios" then "exit without saving"....gets REALLY annoying after a while. Have we established if the x299x is capable of dual booting into opencore and Windows10 seamlessly?

If you look at the post I linked first above you'll see me talking to beltzak about that, and quoting a post from dolgarrenan regarding issues having two EFI partitions accessible causing a boot failure like you described.

However, I think it's possible it's been fixed in recent OpenCore versions because I've not had that issue. As mentioned I can't yet boot Windows direct from OpenCore , but I can boot successfully with two different EFI partitions available and that's what dolgarrenan said wasn't working back then.

I looked at your EFI partition you posted a while back when I was researching my own. I noticed you seemed to be missing a lot of the SSDTs from the OP, and in some cases using generic versions instead (eg using SSDT-PLUG-DRTNIA instead of dolgarrenan's SSDT-X299X-DESIGNARE10G-PR00 and SSDT-AWAC instead of his SSDT-X299X-DESIGNARE10G-RTC0).

Was there a reason you went away from dolgarrenan's SSDTs? They seem to be working well for me and others. They're not necessarily connected to this issue, just wondering why you chose those alternative versions. SSDT-PLUG-DRTNIA for example is a generic version which, according to the Dortania guide, may boot slower than a system-specific version like dolgarrenan has made.

However, one advantage of the generic SSDTs is they probably do already support booting Windows from OpenCore, as they'll have the necessary If statement that makes them run only on macOS.

My recommendation would be to try the OpenCore 0.6.3 EFI I posted yesterday and see how you get on with that as a base. Copy your PlatformInfo details into it and then try booting with it to see if it makes those boot failures go away. Also, like byteminer said, don't use OpenConfigurator.

It won't yet work to boot Windows from OpenCore, but it should be possible to boot Windows from the BIOS Boot Override menu, and I'm not having any of the issues you describe with constant boot failures, and it sounds like it's resolved that also for byteminer.

I'd definitely try clearing your NVRAM before first booting with my (or anyone else's) EFI. If it still fails, it can't hurt to do a CMOS reset and re-configure the BIOS, in case anything has got setup wrong there. I can post a BIOS config file if that helps.

Today I'm going to try editing dolgarrenan's SSDTs to allow booting Windows from OpenCore directly, and I'll update once I have news on that.

 

[TheBloke]

I've started the process of investigating booting Windows from OpenCore. What I learned is that SSDTs need to be edited so they only apply to macOS, because OpenCore will apply the ACPI folder to any and all operating systems it boots.

My first test was to boot Windows 10 from OpenCore using a config which disables all SSDTs. I expected this to boot OK, and then I'd re-enable the SSDTs one by one to confirm if they all needed to be edited.

But I immediately hit a problem: Windows still gave me an ACPI_BIOS_ERROR blue screen even with all SSDTs disabled.

So then I tried a test config in which I also removed all of the ACPI->Patches in dolgarrenan's config.plist, eg where _OSI is renamed to XOSI and PC00 to PCI0. These will also be applied to any and all operating systems.

And that worked, I got Windows booted from OpenCore. But of course this config would then fail completely with macOS.

I think that's why dolgarrenan was using n-d-k's fork of OpenCore, because it had special changes that meant it didn't apply any ACPI stuff to non-macOS operating systems. As far as I am aware, these features are not in OpenCore.

So now I'm investigating whether it's possible to remove the ACPI patches and replace them with SSDTs, which would enable adding the necessary If statement to make them run only for macOS.

I'd also like to understand what these renames are for exactly and whether they're all actually necessary.


Bottom line: it's definitely already possible to dual-boot macOS and Windows if you put up with only being able to load Windows via the BIOS Boot Override menu.

But I think it won't be possible to boot Windows from OpenCore without editing the SSDTs and the ACPI->Patches section. Or there's also the possibility of the BootCamp method, ie booting Windows as if you were booting a real Mac. There's details about that in the OpenCore guides but I've not investigated it myself yet.

If anyone has already done this work and managed a working OpenCore dual-boot using mainline OpenCore, I'd love to see your EFI.

EDIT : talked too soon, multiboot isn't working properly with windows... working to edit the ACPI so opencore passes them straight to windows.

Did you make any progress with this @Walterfilms?