MP All Models Tired of low Geekbench scores? Use RadeonBoost!

[CMMChris]

@iGPU Why do you add an AGPM profile for iMac19,1? iMac19,1 already comes with an AGPM profile targeting GFX0 and IGPU ACPI names instead of device IDs like the iMacPro1,1 and MacProX,X profiles. If you have properly renamed your ACPI devices no AGPM patch is necessary.

Also, why do you use an SSDT for the PowerPlay Table? You can inject it via Device Properties through Clover or include it right into the kext. This s much easier to implement and to change the PowerPlay Table if needed.

Regarding Whatevergreen, I don't have any issues using it with my VII. It isn't necessary though if your Clover configuration is properly set-up. I personally run my Hackintosh without Whatevergreen as well for the sake of keeping things clean. For the sake of working DRM and video encoding / decoding through the GPU I would recommend to use Whatevergreen though if you use iMac19,1 instead of iMacPro1,1. With boot-arg shiki=80 you should get working Sidecar and DRM then. If you don't use Sidecar you can just use iMacPro1,1 SMBIOS without Whatevergreen for working DRM. Just Sidecar won't work then because the T2 handles the video stream encoding on the real iMac Pro.

 

[jncasey]

Thanks for sharing this @CMMChris!

I saw the same huge bump in benchmarks on my machine (Radeon VII / 10.15.4 / OpenCore), using v1.3 of the kext. The only problem I had was when I disabled WhateverGreen, my second display stopped working. But now I have both RadeonBoost and WhateverGreen enabled, and I have the same improved performance and dual displays again, so I'm happy.

In your initial post you said WEG wasn't needed with Radeon VII. But is it still needed for dual displays to work, or is there some quirk of my system that made that happen?

 

[CMMChris]

If you don't use WEG you need to implement some of its patches manually. That includes renaming of the ACPI graphics device to GFX0 and applying the AppleGraphicsDevicePolicy board-id patch. The latter one makes sure none of your cards outputs get disabled

 

[jncasey]

If you don't use WEG you need to implement some of its patches manually. That includes renaming of the ACPI graphics device to GFX0 and applying the AppleGraphicsDevicePolicy board-id patch. The latter one makes sure none of your cards outputs get disabled

Got it. But there's no disadvantage in continuing to use WEG, right?

 

[Ausdauersportler]

Thanks for creating and sharing!

Currently I use WX4170 aka RX560 in my iMac12,2 with OpenCore injecting the iMacPro1,1 board id.

@CMMChris Would it be possible to add support for this card?

I changed the ID to 67e8 and got the kext working with my Opencore, but it had no effect on the GeekBench5 metal score...it is anyway fluctuating.

With this card and my vBIOS I get between 16-18000 before the system went to sleep and 9-12000 after sleeping, same drop is visible in UNIGINE valley scores...(OpenGL).

Thanks in advance

P.S.:
In our current setting we experience also a slight form or Malaria.

After the first sleep of the system the video performance of the WX4170 and WX7100 cards drop down to approximately 50%.

One could measure this using GeekBench and UNIGINE Valley... Could this kext address this problem, too?

 

[09872738]

Thanks @CMMChris!

Great success here as well:

58416 --> 78869 (Metal)
46492 --> 66711 (OpenCL)

i7 -8700k Hackintosh /w Radeon VII. Outstanding work!

 

[miztahsparklez]

Is CS:GO using Metal now? Or still OpenGL?

I think only Metal games may able to benefit from this kext.

Open GL. Maybe that is why BL3 had such a great improvement. Almost double!

@tsialex You know yourself that a 4.1 can also be flashed on 5.1 and Catalina can also be used with a 4.1/5.1. So why remove it?

@miztahsparklez @IndioX And just by the way OC injection works without any issues.

No matter Kext injection or Device Properties.

Yup, i was injecting improperly, i figured out the execution path issue and now know more about OC than i need lol. I also upgraded to the latest OC, lilu, and whatevergreen.



I do think there's something wonky for sleep. My computer has been crashing when it goes to sleep. Not sure what it is. I also can't install the latest OS security patch for Mojave, but I think this is a separate issue.

 

[Ludacrisvp]

Open GL. Maybe that is why BL3 had such a great improvement. Almost double!


Yup, i was injecting improperly, i figured out the execution path issue and now know more about OC than i need lol. I also upgraded to the latest OC, lilu, and whatevergreen.



I do think there's something wonky for sleep. My computer has been crashing when it goes to sleep. Not sure what it is. I also can't install the latest OS security patch for Mojave, but I think this is a separate issue.

I too am having some sleep oddities but I think yours might be related to the thunderbolt card unless you also see issues without that card installed. (My xserve wont sleep with the thunderbolt card in it with modded firmware and DSDT injection)

 

[DanSilov]

Is there anything additional to listed instructions I need to do to enable this kext on Vega 64? Or is it supported at all? I've added it to OpenCore alongside Lilu and WhateverGreen, used the corrected config.plist snippet, but it still doesn't load (don't see VBIOS in Graphics/Displays). HWaccel still works fine. (macOS 10.4.6)

 

[CMMChris]

How about reading the first post?

 

[DanSilov]

How about reading the first post?

Oh, sorry, missed the part about removed support. Is there a way to download 1.1 and test it anyway? I have Vega64, which never experienced fan issues, so I’m not sure if the whole thing applies to me or not, would be glad to play around and test how it behaves.

 

[CMMChris]

All Vega 64 cards had the fan issues prior to 10.14.5. RadeonBoost brings back the fan bug since it apparently is located in the SMU Firmware and was never fixed by Apple. Instead they just stopped uploading the SMU Firmware for PC cards, crippling them in some ways.

The only way to "fix" the fan issues is to inject an improperly configured PowerPlay Table where Accoustic Limit RPM is set to 0. This will make it seem, like fan control works, but it actually doesn't. It will react extremely slow to temperature changes which can cause overheating issues. Also, when the card was under load, fan RPM will increase for a couple of minutes before it starts to spin down slowly.

So please don't attempt to use RadeonBoost with that fix. Instead, everyone (not just Vega 56/64 users, also VII, Navi10/14 owners) should open bug reports with Apple so they finally fix their ****ed up OS. I personally consider their fix as a scam and just another proof for the fact that there is something seriously wrong in their software department or they just don't care about the Mac at all.

 

[09872738]

... or they just don't care about the Mac at all.

👆🏻 this.

 

[FTLN]

Hi All,

First time poster but long time reader

I just set up a hack using opencore and all is working nicely, smbios is set to MacPro7,1.

After installing radeonboost.kext via open core and using the default powerplay table from RadeonVII-PPT-Generator.xlsx I am getting over 100k in Geekbench5 in Metal benchmark on my RadeonVII where previously is was around 40k.

OpenCL Geekbench5 benchmark doesn't seem to have changed much to be honest.

One thing I have noticed though is that fans don't seem to spin up, I'm a little worried I may end up frying the RadeonVII if I am doing a long render using GPU acceleration.

Any ideas how I go about making the fan profile slightly more aggressive ?

My config.plist is available here: https://github.com/ftln94220/Ryzen-Mac-Pro---OpenCore-EFI-for-X570-Phantom-Gaming-ITX-TB3

 

[startergo]

All Vega 64 cards had the fan issues prior to 10.14.5. RadeonBoost brings back the fan bug since it apparently is located in the SMU Firmware and was never fixed by Apple. Instead they just stopped uploading the SMU Firmware for PC cards, crippling them in some ways.

The only way to "fix" the fan issues is to inject an improperly configured PowerPlay Table where Accoustic Limit RPM is set to 0. This will make it seem, like fan control works, but it actually doesn't. It will react extremely slow to temperature changes which can cause overheating issues. Also, when the card was under load, fan RPM will increase for a couple of minutes before it starts to spin down slowly.

So please don't attempt to use RadeonBoost with that fix. Instead, everyone (not just Vega 56/64 users, also VII, Navi10/14 owners) should open bug reports with Apple so they finally fix their ****ed up OS. I personally consider their fix as a scam and just another proof for the fact that there is something seriously wrong in their software department or they just don't care about the Mac at all.
View attachment 909112

The problem with the report is this:
1.This is an unsigned kext so you have to disable SIP to use it
2. I doubt any MP7,1 users will use this. It will be all used by cMP users (vintage=no support)
So who can submit a report that they can really listen to? Certainly not hachintosh users.

 

[CMMChris]

@startergo What you write there makes no sense. This issue has nothing to do with my Kext, Hackintoshes or vintage Macs. Apple have screwed up their drivers and they know it. Any Mac user using third party GPUs in a Mac Pro or eGPU set-up is affected by this. If they notice that people start to figure it out they might change their mind and finally fix things up. Maybe it would be a good idea to get the press on this as well.

Fixing things up is rather easy for Radeon VII and all Navi cards. All they need to do is to implement two additional parameters in the RadeonFramebuffer configuration for those cards. Regarding the Vega 56/65/Frontier cards they will have to do develop a fixed SMU firmware or fix how the driver interfaces with it depending on where exactly the issue with the fan control is.

And while they are at it they should also finally fix the multitude of different bugs affecting the Navi drivers since day one (glitching, stability issues, performance issues). All of them have been well documented and escalated to Apple since end of 2019 without them doing anything. Also the Navi graphics glitches that have been ported to the Radeon VII and Pro Vega II with 10.15.4 should finally be fixed. It's just hilarious how they are doing nothing despite bug reports rolling in.

 

[h9826790]

The problem with the report is this:
1.This is an unsigned kext so you have to disable SIP to use it
2. I doubt any MP7,1 users will use this. It will be all used by cMP users (vintage=no support)
So who can submit a report that they can really listen to? Certainly not hachintosh users.

If I read the info correctly, that "fan bug" was fixed in 10.14.5 (Mojave). But at the same time, Vega 56/64/FE performance dropped.

Since Mojave is still a supported macOS, and 5,1 support Mojave, we can still submit reports about "performance drop for Vega 56/64/FE observed since 10.14.5"

For 7,1 users, they can't run Mojave, so, nothing to compare. But for those who own both 5,1 and 7,1, can submit bug reports about "the same Vega 56/64/FE perform better in 10.14.4 (and before) on 5,1 than on a 7,1 that is running Catalina"

We can't quite submit a "fan bug" because Apple fixed that. It's the performance issue now.

RadeonBoost is a work around for performance fix, which should be done by Apple. We can focus on performance in the bug reports, that still valid.

 

[startergo]

If I read the info correctly, that "fan bug" was fixed in 10.14.5 (Mojave). But at the same time, Vega 56/64/FE performance dropped.

Since Mojave is still a supported macOS, and 5,1 support Mojave, we can still submit reports about "performance drop for Vega 56/64/FE observed since 10.14.5"

For 7,1 users, they can't run Mojave, so, nothing to compare. But for those who own both 5,1 and 7,1, can submit bug reports about "the same Vega 56/64/FE perform better in 10.14.4 (and before) on 5,1 than on a 7,1 that is running Catalina"

We can't quite submit a "fan bug" because Apple fixed that. It's the performance issue now.

RadeonBoost is a work around for performance fix, which should be done by Apple. We can focus on performance in the bug reports, that still valid.

And they will ask:
steps to recreate this bug and observations?

 

[CMMChris]

We can't quite submit a "fan bug" because Apple fixed that. It's the performance issue now.

No they did not fix it. All they did was a dirty hack that now affects nearly all PC graphics cards used on a Mac.
[automerge]1587820328[/automerge]

And they will ask:
steps to recreate this bug and observations?

Steps to recreate this bug? Run Geekbench! LMAO

 

[cdf]

So who can submit a report that they can really listen to? Certainly not hachintosh users.

Indeed.

Steps to recreate this bug? Run Geekbench! LMAO

Right. And I really admire the passion. But to be taken seriously, the report should not mention hacks.

 

[CMMChris]

Really?! Passionate hackintosh users probably have the most in-depth knowledge of the macOS operating system. Apple knows that. They also never had problems with hackintosh users submitting bug reports in the past. I personally had an Apple engineer on mine via remote access to investigate an issue back in early 2019. Surprisingly, he even knew how to handle Clover. Not joking.

 

[startergo]

Surprisingly, he even knew how to handle Clover. Not joking.

Not surprising . Anyway for the sake of the argument lets pretend I am the devil's advocate (which I am not ).
So I will say:
Yes we had that high performance, but... we had that annoying fan racing bug. So we fought over an year trying to keep the higher results and eliminate the racing bug, but we could not achieve that. So we decided to degrade the performance and eliminate the racing bug. If you can fix this issue can you create a signed kext and we can see how it works and will fix it in the drivers. (they will never say that out loud but that is what they think ).
By the way even unsigned kext will work for us (Apple)

 

[DanSilov]

All Vega 64 cards had the fan issues prior to 10.14.5. RadeonBoost brings back the fan bug since it apparently is located in the SMU Firmware and was never fixed by Apple. Instead they just stopped uploading the SMU Firmware for PC cards, crippling them in some ways.

Thanks, I’ve more or less understood the issue now, will see if I can file a bug report.

At the same time I would still like the option to use RadeonBoost when I need more performance and then disable it later when I need a quiet operation. I hope there’s a way to download 1.1, as I perfectly understand that adding the support back is not a viable option for everyone

 

[MuNLoK]

Thank you for sharing CMMChris!!

How do you think about this performance with the RX 5700 EFI (MacVidCards)? shouldn't the RX 5700 give a little more?

Mac Pro 4.1>5.1
Radeon RX 5700 8GB Apple EFI (MacVidCards)
Catalina 10.15.4 (@dosdude1)
No OpenCore

 

[h9826790]

And they will ask:
steps to recreate this bug and observations?

Steps:

1) Install macOS 10.14.4 on Mac Pro 5,1

2) Install a Vega64 on the Mac Pro 5,1 (this config should be supported according to Apple's official info), remove all other GPU

3) Run Geekbench 5 Metal benchmark

4) update the OS to the latest 10.14.6

5) Run the same Geekbench 5 Metal benchmark again

6) Compare the score from step 3 and step 5

Expectation: The scores are very close to each others.

Observation: The score in 10.14.6 is noticeably below in 10.14.4 for the same graphic card (Vega64), on the same computer (Mac Pro 5,1)
[automerge]1587828507[/automerge]

No they did not fix it. All they did was a dirty hack that now affects nearly all PC graphics cards used on a Mac.

To be fair, they fixed the fan bug (no more fans issue). But they use a wrong way to fix it, which create another bug that even worse (the big performance hit bug).

If we submit a bug report that's about "fan speed", most likely they will simply reply "cannot reproduce".

If we submit bug report about "fan speed issue with RadeonBoost installed", then they will reply "contact the developer of RadeonBoost".

Therefore, I will treat it as "fan bug was fixed", but a "performance bug" exist now. In this way, they can reproduced the issue without any 3rd party software (apart from benchmarking, because no GPU benchmark exist in macOS natively). And they can't say "the bug was fixed back in 10.14.5".