MP 1,1-5,1 MAC PRO 5.1 RADEON VII NO VIDEO OUTPUT!!

[tootansimooon]

May be you can use something like Kext Utility to install the kext for you.


If that kext work, and you 100% rely on that kext. Then the answer is YES.

I always keep SIP disabled since it was introduced. No problem so far.

And if you want to re-enable it. Then you better learn how to use OpenCore to inject the parameters you want.

Hi @h9826790

Alas, I disabled sip, successfully installed the kext you kindly sent.

Still no dice. Everything is as it was earlier...

Damn shame, I was very much looking forward to using this card.

Any other advice very welcome.

Cheers!

Simon

 

[h9826790]

I was very much looking forward to using this card.

If you really want to make this card works, and must be in Mojave, then I suggest do the following.

1) Flash your cMP with EnableGop, so that this card can display boot screen.

2) Install OpenCore

3) Create a SSDT to spoof your card's Device ID into Radeon VII's Device ID.

Apart from using SSDT, we may also try some simply parameters injection first. And inject some extra parameters may help the Vega II run better anyway. But by considering your Radeon Pro VII is almost identical to the Radeon VII. And Radeon VII can work flawless in Mojave. Spoof Device ID should work.

 

[tootansimooon]

If you really want to make this card works, and must be in Mojave, then I suggest do the following.

1) Flash your cMP with EnableGop, so that this card can display boot screen.

2) Install OpenCore

3) Create a SSDT to spoof your card's Device ID into Radeon VII's Device ID.

Apart from using SSDT, we may also try some simply parameters injection first. And inject some extra parameters may help the Vega II run better anyway. But by considering your Radeon Pro VII is almost identical to the Radeon VII. And Radeon VII can work flawless in Mojave. Spoof Device ID should work.

Hi @h9826790

Thanks so much for all your help. It was great to test this card's compatibility out, hopefully others will be able to learn from this that the Radeon PRO VII is not natively compatible with mojave. I was hoping to have a non-pixlas mod, plug-and-play version of the VII.

I think the wisest thing to do (for me) is to buy a card which is compatible, as above. A 6800? a 5700?

Realise I'm changing the topic here, and that it's been well documented elsewhere, but what is the current recommendation for the most powerful card which fulfils the above requirements (non-pixlas mod, plug-and-play, native to mojave)? Entirely for Davinci Resolve with heavy colour grading.

I will upgrade to OPENCORE soon enough, but at the moment I'm in the middle of several big editing projects and don't want to rock the ship. I'd like to spend time to learn about it properly before actioning it when I have more time to dig into it.

Thanks again!! Lemme know what you think about an alternative.

S

 

[h9826790]

Hi @h9826790

Thanks so much for all your help. It was great to test this card's compatibility out, hopefully others will be able to learn from this that the Radeon PRO VII is not natively compatible with mojave. I was hoping to have a non-pixlas mod, plug-and-play version of the VII.

I think the wisest thing to do (for me) is to buy a card which is compatible, as above. A 6800? a 5700?

Realise I'm changing the topic here, and that it's been well documented elsewhere, but what is the current recommendation for the most powerful card which fulfils the above requirements (non-pixlas mod, plug-and-play, native to mojave)? Entirely for Davinci Resolve with heavy colour grading.

I will upgrade to OPENCORE soon enough, but at the moment I'm in the middle of several big editing projects and don't want to rock the ship. I'd like to spend time to learn about it properly before actioning it when I have more time to dig into it.

Thanks again!! Lemme know what you think about an alternative.

S

No worries, totally understand that you just want a working machine, but not works like a developer to make the Radeon Pro VII works in Mojave.

Anyway, for Mojave, if you also put power draw into account, and don't even consider use kext to undervolt the graphic card, then the best option should be Vega56.

Anything above Radeon VII need newer macOS.

Radeon VII itself need Pixlas mod or downvolt. Otherwise, under very high stress, may hard shutdown the cMP.

So, the next option will be Vega56 and Vega64. Vega64 of course is stronger, but same problem as Radeon VII, may trigger the shutdown protection. That will be a nightmare in the middle of a DV project. Therefore, Vega56 should be the best bet.

 

[paul_bace]

Don't know how to read the info that you've asked for so just posting it here. I'm on Monterey if that means anything.

UPDATE: So the Radeon Pro VII is working great. Other than not working at full capacity for most renders/exports. I've somehow gotten a little more juice after a fresh install, more on the processor side of things. And barely on the GPU side of things. The Luxmark results are great. Full on both. But exporting in both FCPX and Davinci are meager (30% max) even with ProRes, but definitely seeing a better result for CPU's.

Other than that, the cable fixed my original problem of image cut-outs/black screen. I'm guessing if anything the speed boost of fresh install might be because I'm running on a PCIe boot drive now. Which is great to have back. I'm ok with results at the moment given the improvement PCie boot is giving me. That said, if there's anyway to actually get the max use of GPU I'm all ears.

UPDATE 2: So reading over the other posts you made @h9826790 I used OCLP to inject GOP, and seem to be getting a little more performance. I couldn't actually get your package to work properly with my system. Even after fresh install on a mojave drive. So currently going with the tried and true OCLP but with the GOP injection option selected.

 

[paul_bace]

Exporting in Davinci

Pro Res - 90% CPU :: 30% GPU
H.265 - 60% CPU :: 1% GPU
H.264 - 15% CPU :: 20 GPU

Exporting FCPX
Pro Res - 15% CPU :: 30% GPU
HVEC - 30 % CPU :: 15% GPU
H.264 - 30% CPU :: 30% GPU

EDIT: Just had a thought, I'm not using triple channel memory right now. Using 4x 16gb and 4x 8gb Ram, would that effect anything?

 

[tootansimooon]

No worries, totally understand that you just want a working machine, but not works like a developer to make the Radeon Pro VII works in Mojave.

Anyway, for Mojave, if you also put power draw into account, and don't even consider use kext to undervolt the graphic card, then the best option should be Vega56.

Anything above Radeon VII need newer macOS.

Radeon VII itself need Pixlas mod or downvolt. Otherwise, under very high stress, may hard shutdown the cMP.

So, the next option will be Vega56 and Vega64. Vega64 of course is stronger, but same problem as Radeon VII, may trigger the shutdown protection. That will be a nightmare in the middle of a DV project. Therefore, Vega56 should be the best bet.

Hi

I can get a Vega56 damn cheap. So I'm thinking of getting one.

However, other threads suggest that even that requires Pixla's mod... but I see no first-hand experience evidence to suggest that. I'd be pushing the card to the limit in Resolve no doubt, so want to make sure. Anyone have first hand experience?

@h9826790 - thanks again. You saved me countless hours and stress with your advice and generosity of time. 🏆

@paul_bace - How much GPU is being used when in timeline in Resolve, with Noise Reduction and film grain (etc) nodes?

S

 

[paul_bace]

Another interesting fact. Kind of annoying knowing this.

 

[tsialex]

Another interesting fact. Kind of annoying knowing this.

Apple deprecated OpenCL for years already, since WWDC2018/Mojave/iOS12 Apple moved to METAL, zero sense in comparing OpenCL scores anymore.

 

[paul_bace]

Hi

I can get a Vega56 damn cheap. So I'm thinking of getting one.

However, other threads suggest that even that requires Pixla's mod... but I see no first-hand experience evidence to suggest that. I'd be pushing the card to the limit in Resolve no doubt, so want to make sure. Anyone have first hand experience?

@h9826790 - thanks again. You saved me countless hours and stress with your advice and generosity of time. 🏆

@paul_bace - How much GPU is being used when in timeline in Resolve, with Noise Reduction and film grain (etc) nodes?

S

Ah, there we go. Lots of GPU is being used. Like 80% average with barely any CPU support. I guess I just figured the GPU would be used a lot more when exporting in various codecs. Especially with FCPX (although I am fully making the switch to Davinci so yeah).

 

[h9826790]

I can get a Vega56 damn cheap. So I'm thinking of getting one.

However, other threads suggest that even that requires Pixla's mod... but I see no first-hand experience evidence to suggest that. I'd be pushing the card to the limit in Resolve no doubt, so want to make sure. Anyone have first hand experience?

That based on few variables, but mainly

1) Which Vega 56 - Some Vega56 are heavily factory overclocked, which can draw a lot more power. For your info, the boost clock of a reference Vega56 is 1471MHz. If you see a Vega56 has something close to 1600MHz, you better avoid it. That extra 100MHz will bring you some more performance, but draw a lot more power, not cost effective. And in your case, may cause trouble.

2) The card's actual power draw pattern - e.g. a dual 8pin Vega 56 may heavily rely on 8pin A, but very low usage on 8pin B. And only draw more power from 8pin B when you overclock it.

So, if you connect 8pin A to mini 6pin A, and 8pin B to mini 6pin B. Mini 6pin B will be under use, but mini 6pin A have to deliver way more than 75W and hit the SMC's hard shutdown protection.

Anyway, for non heavily overclocked Vega56, as long as you balanced the power draw between mini 6pin A and B, it shouldn't need Pixlas mod.

As long as you use the correct cables combination to connect the card. The power draw will be balanced automatically. No other user input required.

 

[tootansimooon]

Hi

Good advice as per usual. I wish there was a list of GPU's - with brand variants as well as the base GPU model - showing clock speed/power draw/compatibility etc.

@h9826790 do you think the same applies if I were to use the following combination:
DUAL mini 6pin -> Single 8pin + Single 8pin FEMALE -> DUAL 6+2 pin

i.e. as per your recommendation but with an 8 pin rather than 6 pin bridge? I have this cable lying around, and I'm going to buy a Vega 56 now.

@paul_bace yep, I find with Resolve, using BRAW footage with lots of nodes - just looking at the timeline playback and scrubbing -with my current RX580 the GPU is at 100% with the CPU (a single X5960) at around 20-40...

I'm gonna try a Vega 56. I assume it'll be running at 100% too. But perhaps with more responsiveness. If it doesn't work - it's probably time to get a new setup entirely.

Cheers!

PS, assuming the card does work, is there a way of optimising it for Resolve?

 

[h9826790]

@h9826790 do you think the same applies if I were to use the following combination:
DUAL mini 6pin -> Single 8pin + Single 8pin FEMALE -> DUAL 6+2 pin

i.e. as per your recommendation but with an 8 pin rather than 6 pin bridge? I have this cable lying around, and I'm going to buy a Vega 56 now.

Absolutely no problem to use 8pin bridge. I use 6pin bridge simply because that's less unnecessary wires, and that cable was cheaper.

But in general, it's much easier to buy the 8pin cables for this purpose.

 

[tootansimooon]

Absolutely no problem to use 8pin bridge. I use 6pin bridge simply because that's less unnecessary wires, and that cable was cheaper.

But in general, it's much easier to buy the 8pin cables for this purpose.

Brilliant

Installed the Vega 56. So far, so good. Resolve isn't lighting fast, but I'd say the edit page, using all .BRAW 4k - with grading on - now plays at half quality faaairrly well. The RX580 couldn't play at quarter quality. No shut down's even when GPU is at 100% (which is all the time when playing back with grading on)

Export time has decreased too:
- RX580 8min32sec
- Vega 56 5min32sec

Mainly reporting this for posterity.

@h9826790 - is there a way of enabling AMD hardware acceleration without opencore? Do you think there would be any benefit in resolve?

Cheers!!

 

[h9826790]

@h9826790 - is there a way of enabling AMD hardware acceleration without opencore?

Yes, except HEVC hardware encoding.

Do you think there would be any benefit in resolve?

If any of your source videos are H264 / HEVC, then yes, the timeline playback will benefit from it.

Or if you export to H264, then the export time most likely will be shorten as well (depends on how hard to render the video, but in general, the encoding part is the slowest).

But if you only deal with ProRes, then most likely your workflow won't benefit from it.

 

[Kantgittrite]

For boot screen, Radeon VII need DirectGopRendering, which the RX590 may not, you better check if you enabled this option in the OpenCore config.

Since Catalina cannot be run on the cMP without patching. If you pull out the OC drive but you can still boot to that with your RX590. This is only possible because you patched Catalina.

For patched macOS, the GPU driver somehow may be "locked" to the GPU that ran the macOS installer. This can also explain why the GPU driver doesn't load for the Radeon VII, and you can't see anything.

In your case, I will say the first thing to try is just enable DirectGopRendering.

Where in the heck is this "DirectGOPRendering" tab, button, or check mark in OpenCore? I've finally was able to install Monterey with the help of OpenCore boot screen and all. I would LOVE to use my Radeon VII instead of Radeon7950 for video editing. I know the VII will work, I'm just lost on how with where to look and go.

In the OpenCore Settings: Advanced tab on bottom it says AMD GOP Injection. I check it shut down install VII don't work....uncheck shut it down don't work.

Did I miss something in OpenCore? Did I use the wrong OpenCore? I know it took awhile to find the right one for me which is OpenCore Legacy Patcher 1.5.0

Sorry for the rant and thanks in advance

 

[shijan]

This is the RadeonBoost kext created by CMMChrist, but I modified it a bit to fit Radeon Pro VII. Hopefully the extra device properties injected can make Mojave ident your GPU as a Metal device. If you don't mind, you may try this kext. It's a codeless kext, should be no harm. And you can open the info.plist to see what's injected.

If you are OpenCore user, you may simply inject the property via OC config. Or inject this kext via OC should also work.

If you installed this kext to /L/E, then something goes wrong, then a NVRAM reset to re-enable SIP should automatically disable this kext.

There is a mistake in code and so plist file can't be opened. Seems you forgot to add <integer>0</integer> after <key>PP_DisableULV</key>

Here is correct version:

            <dict>
                <key>model</key>
                <string>AMD Radeon Pro VII</string>
                <key>@0,name</key>
                <string>ATY,Donguil</string>
                <key>@1,name</key>
                <string>ATY,Donguil</string>
                <key>@2,name</key>
                <string>ATY,Donguil</string>
                <key>@3,name</key>
                <string>ATY,Donguil</string>
                <key>@4,name</key>
                <string>ATY,Donguil</string>
                <key>@5,name</key>
                <string>ATY,Donguil</string>
                <key>ATY,Copyright</key>
                <string>Copyright AMD Inc. All Rights Reserved. 2005-2020</string>
                <key>ATY,FamilyName</key>
                <string>Radeon Pro</string>
                <key>ATY,DeviceName</key>
                <string>Vega II</string>
                <key>ATY,EFIVersion</key>
                <string>01.01.186</string>
                <key>ATY,Rom#</key>
                <string>113-D160BW-444</string>
                <key>ATY,EFIVersionB</key>
                <string>113-D163A1XT-045</string>
                <key>ATY,EFIVersionROMB</key>
                <string>113-D163A1XT-045</string>
                <key>PP_DisableULV</key>
                <integer>0</integer>
            </dict>