Jump to content

AMD Polaris IDs on Sierra / High Sierra

Ciro82

By Ciro82,
in ATi

 Share
871 posts in this topic

Recommended Posts

Ciro82
Ciro82
Posted
  • Author
Posted

I don't think that the card is faulty, beside the annoying fans it works well, and the boot screen looks normal on non UHD displays.

 

The bios splash screen full of artifacts seems pretty bad to me...

Link to comment
Share on other sites
surfinchina
surfinchina
Posted
Posted

The weired thing is, that DELL monitor has the same settings (3440x1440x50hz) as when i use GUARIBA as framebuffer - but no matter what i try to do (select other resolution or timing), screen stays back with message "lost signal". I hope it is clear what i was meaning. Will provide some screenshot to provide better understanding of what i was talking about.

I have a dell U34 and it only works on the HDMI. On DP it did what yours did and sleeps. I'm a bit newbie and maybe it's something I've done wrong in clover, but the Dell U27 (2560x1440) works fine on the DP.

I have an R9 Nano and run the fake id 73001002 but it's not accelerated.

I'm a bit stuck there because I have an X99 mobo and no other graphics are possible.

Someone said it works on nmp as eGPU but I have no idea what that means.

 

Anyway, try HDMI port with the 34" monitor. 

Link to comment
Share on other sites
surfinchina
surfinchina
Posted
Posted

post-1564334-0-20995700-1476382792_thumb.jpg

What does you having a X99 mobo have to with what graphics card you use?

 

The Fury and nano won't accelerate without starting from integrated graphics, it seems. The X99 doesn't have this. By other graphics I meant integrated.

Also I've made my PC so that it only fits a single tiny watercooled card

Link to comment
Share on other sites
MattsCreative
MattsCreative
Posted
Posted

x99 isn't recommended for os x by any sane hack users and until apple finishes the rx480 or clover issues are fixed you need the igpu to use this card

20160812_144825.jpg

 

The Fury and nano won't accelerate without starting from integrated graphics, it seems. The X99 doesn't have this. By other graphics I meant integrated.

Also I've made my PC so that it only fits a single tiny watercooled card

Link to comment
Share on other sites
Fljagd
Fljagd
Posted
Posted

Heres a crazy idea. What happens if we disable IGPU from BIOS. But what if we create a fake IGPU in DSDT, would that work? Any dsdt expert wanna try?

the idea is interesting

But iGPU is an instruction included in the CPU not only in the DSDT

Link to comment
Share on other sites
metacollin
metacollin
Posted
Posted

x99 isn't recommended for os x by any sane hack users

 

This is utter nonsense, please ignore it.  The X99 and C612 chipsets work great and are by far the easiest chipsets I've hackintoshed, and don't require a single specialized DSDT patch to be fully working.  As I am typing this from a Dual Xeon V4 C612 hackintosh running Sierra that I use in a production environment, I can empirically demonstrate that wardoc doesn't know what he's talking about.  This is the most stable and smooth computer to run OS X I've ever used, and I used a 2008 Mac Pro (like, a real one) for 8 years before this current hackintosh rig.  I am sane, and I would recommend it over any other chipset.  It just works.  

 

attachicon.gif20160812_144825.jpg

 

The Fury and nano won't accelerate without starting from integrated graphics, it seems. The X99 doesn't have this. By other graphics I meant integrated.

Also I've made my PC so that it only fits a single tiny watercooled card

 

the idea is interesting

But iGPU is an instruction included in the CPU not only in the DSDT

 

Everyone, this has nothing to do with an IGPU.  At all.  This has nothing to do with Apple's RX 480 support.  This is entirely an issue with clover.  

 

I have an RX 480, and I am also a registered macOS developer member ($100 for the 'honor' =P) and am running Sierra 10.12.1 Beta 3.  As stated earlier, I have a dual Xeon Rig, with no IGPU to speak of, or even the possibility of one.  Fortunately, that isn't at all necessary to get the RX 480 working.  This is a clover issue, something it is doing to ATI cards but only if they're the 'boot' GPU (primary) is somehow incorrect and causing the X4100 kext to panic during the initialization process of the RX 480.  

 

It's easy to tell this is what is happening.  There is a subtle but important difference between the various ways one might get a black screen.  If it was a problem with the kext, then it would simply fall to load and fallback to the unaccelerated frame buffer.  Or, similarly, if it was something along the lines of the MacPro6,1 GFX0 unload issue, that results in a true black screen.  i.e., there is a video signal coming out of the card driving the display, but telling it to show nothing.  Or more specifically, to show 000000h for every pixel (black).  Instead of a display showing black, this RX 480 issue is presenting as an actual lack of video signal entirely, but only after the X4100 kext tries to load.  After this, all system I/O halts, which would only occur if there was a panic.  Depending on your setup and the state macOS is in when a panic occurs, this can result in an indefinite stall or an automatic reboot.  I have seen my machine automatically reboot with a dead screen (note dead screen vs black screen.  Black screen is still a working GPU, dead screen is no video signal at all) as well as stall in that state.  Regardless, a panic is the only explanation for all the symptoms. 

 

This panic does not occur on MacPro5,1s for example that have had their owners install an RX 480.  There is no issue for Apple to fix.

 

Finally, coming back to the IGPU thing.  All that matters is that Clover doesn't touch the RX 480 in someway that it only does to the 'boot' (primary) GPU before macOS can get at it.  What this means is that as long as Clover is told that a GPU other than the RX 480 is the 'boot display', then it won't do whatever it is that its doing that causes Polaris cards to crash upon initialization of the card by macOS's X4100 kext.  

 

As far as I know, most BIOSes will make any GPU the primary GPU if there is only one.  So all you need is 2 GPUs, and to force the one that ISN'T the RX 480 to be the primary/boot gpu, or in other words, the gpu the post screen and clover menu/EFI video is sent to.  

 

The IGPU is a convenient, already-present second GPU for some people.  For those without, they simply need a second GPU at all.  It doesn't matter what.  In my case, I have tried this piece of {censored}  and as long as it is primary, the RX 480 works great.  It also works with what I have installed right now as primary, a GTX 960.  It also works with an old Radeon HD 6870.   It doesn't matter how, as long as the effect is the same.  That effect is preventing the RX 480 being used as primary.  Using an IGPU is one of many possible ways of achieving this, but there is nothing specific to an IGPU involved, only that it is a GPU.  A gpu that is not a Polaris GPU, but is primary.

 

 

Now, I've been digging through the Clover and EDK code, and frankly, I am at a total loss so far.  I think it is safe to say that it is something specific and only done to the boot GPU by Clover, as there is no problem with OS X using the GPU otherwise (on a real mac or hackintosh) that leaves some register in a state that the X4100 kext does not expect and would never occur normally, so it panics.  And Apple won't fix it because there is no problem to fix - except with Clover.  

 

Just to be clear, I am in no way 'blaming' Clover or slice or any other of the developers.  That they include code for GPUs before they are available and then it works at all, even with workarounds, is an amazing feat and requires a great deal of skill and intimate knowledge of both hardware in general as well as how macOS uses it specifically.  But, I am just saying, that is where the solution will be found.  Considering no one has even bothered to create a ticket on Clover's source forge page though, it is not going to be soon.  I'll open a ticket after this post, though considering how old this thread is, I don't understand why it hasn't already been done.  

 

 

Oh, and doing anything with DSDT is a dead end.  Clover is what loads the ACPI tables then hands them off to macOS, it also has to talk to the hardware, set registers, query device IDs, map memoryio base locations, etc. before ever loading any ACPI tables, so changes to DSDT etc. will not have any effect.  Beyond that, this is a case of Clover doing something it ought not to on Polaris GPUs, or perhaps NOT doing something it needs to after its done something else that it does only to the boot gpu on polaris CPUs.  We know this because there is no problem if it is not the boot gpu.  It already works.  What doesn't work is initialization by the driver if clover has already been using it.  

 

Oh, right, proofs, working even with the old crappy GeForce card (comma is my bad a typo that I haven't bothered to fix lol. I am using an SSDT hot patch to inject the ID for an RX 460 so the kext matches unmodified, as well as disable the other slot devices that could match instead...but I digress. It doesn't matter if it is done this way, or with the native ID and patched plist, and looking at the clover code, there is no meaingful difference in how it treats polaris 10 vs 11 anyway):

 

upshot_IJTi3YF4.png 

 

Also, the compute performance is a little disappointing, but certainly much faster than the GTX 960.  The performance under windows is almost 130,000, but this is to be expected.  Apple's OpenCL implementation has always been a {censored} show and probably isn't going to get any better soon or ever.  I also have mine underclocked, I'll try another full power bench later (I was making sure my thermals were adequate).  I don't expect to hit more than 100,000 however under OS X.  

  • Like 6
Link to comment
Share on other sites
Ciro82
Ciro82
Posted
  • Author
Posted

(comma is my bad a typo that I haven't bothered to fix lol. I am using an SSDT hot patch to inject the ID for an RX 460 so the kext matches unmodified

 

upshot_IJTi3YF4.png

 

 

 

I have the same "comma error" using DSDT injection, it's due to buffer sizes ;)

Link to comment
Share on other sites
Matthew82
Matthew82
Posted
Posted

"Now, I've been digging through the Clover and EDK code, and frankly, I am at a total loss so far.  I think it is safe to say that it is something specific and only done to the boot GPU by Clover, as there is no problem with OS X using the GPU otherwise (on a real mac or hackintosh) that leaves some register in a state that the X4100 kext does not expect and would never occur normally, so it panics.  And Apple won't fix it because there is no problem to fix - except with Clover."  

 

 

I thing that problem is in AMD8000,9000 and 9500Controller. Im running X99 system with R9290X and as primary I have 280X wich using the same X4000.kext. 

The "black screen" problem reporting by people in forum is about GPUs using above AMD7000Controller.kext. 

Link to comment
Share on other sites
Pavo
Pavo
Posted
Posted

 

"Now, I've been digging through the Clover and EDK code, and frankly, I am at a total loss so far.  I think it is safe to say that it is something specific and only done to the boot GPU by Clover, as there is no problem with OS X using the GPU otherwise (on a real mac or hackintosh) that leaves some register in a state that the X4100 kext does not expect and would never occur normally, so it panics.  And Apple won't fix it because there is no problem to fix - except with Clover."  

 

 

I thing that problem is in AMD8000,9000 and 9500Controller. Im running X99 system with R9290X and as primary I have 280X wich using the same X4000.kext. 

The "black screen" problem reporting by people in forum is about GPUs using above AMD7000Controller.kext. 

 

Been saying this a long time, this is a Clover issue because I can run R9 290X and RX 480 and any GTX 9xx series GPU in my real mac no issues, hence no iGPU

Link to comment
Share on other sites
Fljagd
Fljagd
Posted
Posted

This is utter nonsense, please ignore it.  The X99 and C612 chipsets work great and are by far the easiest chipsets I've hackintoshed, and don't require a single specialized DSDT patch to be fully working.  As I am typing this from a Dual Xeon V4 C612 hackintosh running Sierra that I use in a production environment, I can empirically demonstrate that wardoc doesn't know what he's talking about.  This is the most stable and smooth computer to run OS X I've ever used, and I used a 2008 Mac Pro (like, a real one) for 8 years before this current hackintosh rig.  I am sane, and I would recommend it over any other chipset.  It just works.  

 

I can only agree

And thank you for all other information

Link to comment
Share on other sites
stehor
stehor
Posted
Posted

im using ozmosis still needs igpu on rx cards and im using a rx480 4gb

Link to comment
Share on other sites
Henry2010
Henry2010
Posted
Posted

anyone wanna try this out? :)

 

Upgraded my mac pro tower to 10.12.1 and found these kext files updated. After a bit digging, there are references of PCI2PCIBridge and AppleDisplay.

 

AMDRadeon4100.kext has support for Polaris 10 and Polaris 11 GPUs so this is most likely an indicator of new Apple Display with eGPU integration comes Thursday.

 

AMDRadeonX4100.kext, AppleBacklightExpert.kext, Audio AUUC.kext, IOGraphicsFamily.kext, IONDRVSupport.kext, IOPCIFamily.kext, System.kext were signed off at 4:19am last Wednesday - right about when the invites for the 27th event went out.

 

source: http://forums.macrumors.com/threads/new-apple-display-with-egpu-integration-in-10-12-1.2008996/
Link to comment
Share on other sites
 Share
Go to topic listing
×
×
  • Create New...