macOS Sierra PB: Need testers for new AMD Radeon drivers!

[superic]

Giving up on Vega already? You might want to wait until both companies have cards on the local stores' shelves before making up your mind. This is incredible news that Mac users will have choices of, well --- lower cost, workstation level graphics cards, but keep in mind what was not listed in the announcements. How well the cards are going to work with Apple's Metal platform. Right now, unless there has been a drastic overnight change, supporting CUDA in your software means paying a licensing fee. So there is a lot of OS-X oriented software that is optimized to run on the Open Source approaches, OpenGL and OpenCL. Just plopping in a faster Nvidia card and running that same software does not automatically mean your results will also be faster.

But if you are going to write letters to AMD, please ask them to support audio over DP and HDMI.

Haha. Maybe we all should write letter. I guess for now I can wait a little

 

[rhys216]

So you have managed to get your RX480 running without iGPU by editing three bytes in the card's ROM??? As I understand, you have placed an invalid device id in it, right?

Yes that's what I did. I might try changing the ID to EF in the bios to see if I don't need to inject the device id in clover any longer.

 

[Sitherick]

Yes that's what I did. I might try changing the ID to EF in the bios to see if I don't need to inject the device id in clover any longer.

Probably you'll loose the ability to boot without iGPU then - EF67 isn't an invalid id... This invalid id prevents to load something (EFI driver/GOP) before Clover loads.

BTW, the 3.rd DF67 you replaced seams not to be an id, it can be also be found on RX460 BIOS...

 

[LarsBorris]

Yes, the R9 290X still needs some kind of "help" to init properly. Sadly no easy fix yet.

Did you try to completely disable your iGPU? You shouldn't need it anyway. Is your monitor connected to the FirePro?

Just tried it, it works but when i launch a GPU heavy application the system hangs, so not very stable sadly.
Thanks for the assist.


Edit:
This is the LuxMark score:

 

[Docxavier]

You probably will not like my answer. I am in 10.12.5 PB (s well as 10.12.4) and some sleep functions are working but hibernate mode is not working. I lost hibernate mode somewhere around 10.12.2 or so. Display sleep, all of the way to monitor shutdown mode and restart is working well. I have tried various options to recover audio after sleeping, but it seemed that longer sleep times caused these fixes to also not work, and hence the system needed to be rebooted. So for 'long enough' sleep times, I just shut the system down and restart later. For short term sleep, I open a terminal window, type in the command caffeinate and let it run. This will let the screen saver come on as you have it set up, then sleep the display, then allow it to shutdown at the times you have set up in the Energy Saver Preferences. The mouse and keyboard will awaken the system with acceleration, audio,. etc. no problems, but this command is not for long term use as it does not sleep the cpu.

edit: type the usual control-c combination to exit the command.

Thanks jerryy2345¡¡¡ You are right, I don't expect that answer but, as always, I'm learning new things from people who is trying to help others. I'll try caffeinated, I never heard about this command.
About my experience with sleep and wake, everything is going right 80% of the times, and for the 20% my system wakes without acceleration, but audio works always after sleep, short or long term, using codeccommander.kext for my build.
Can not have it all, hackintosh life ...

 

[jerryy2345]

Thanks jerryy2345¡¡¡ You are right, I don't expect that answer but, as always, I'm learning new things from people who is trying to help others. I'll try caffeinated, I never heard about this command.
About my experience with sleep and wake, everything is going right 80% of the times, and for the 20% my system wakes without acceleration, but audio works always after sleep, short or long term, using codeccommander.kext for my build.
Can not have it all, hackintosh life ...

I need to go back and check CodecCommander again, with the recent Clover/macOS changes in audio, it may give me better results -- that is the thing about updating, you have to check so many things that you 'knew did or did not work in the last version'. Though I really do not miss hibernate mode now that I have not been using it for a while, ..., I was never quite settled in my mind about its effect on SSD lifespan even though the ones I use are supposed to be safe. Are you coming out off sleep mode or out of full hibernate mode when you lose acceleration. caffeinate <-- someone at Apple has a sense of humor.

 

[Docxavier]

I need to go back and check CodecCommander again, with the recent Clover/macOS changes in audio, it may give me better results -- that is the thing about updating, you have to check so many things that you 'knew did or did not work in the last version'. Though I really do not miss hibernate mode now that I have not been using it for a while, ..., I was never quite settled in my mind about its effect on SSD lifespan even though the ones I use are supposed to be safe. Are you coming out off sleep mode or out of full hibernate mode when you lose acceleration. caffeinate <-- someone at Apple has a sense of humor.

What I can tell you is I'm using hibernatemode 0, but I don't know if this is sleep or full hibernate ... Do you think this could impact on wake behaviour losing acceleration (sometimes ...)? Which mode are you using?
Talking about something else, I have used both boot config I see in this thread for our card, using Berbice+Mork/Jerry patch+ATI inject and without ... Without patches, I get black screen (monitor self sleep) at boot and pressing monitor ON button I can reach desktop. Geekbench 4 Open CL benchmark is around 72000 in that conditions.
Using patches, I get desktop but Geekbench says around 60000. Makes sense?

 

[jerryy2345]

What I can tell you is I'm using hibernatemode 0, but I don't know if this is sleep or full hibernate ... Do you think this could impact on wake behaviour losing acceleration (sometimes ...)? Which mode are you using?
Talking about something else, I have used both boot config I see in this thread for our card, using Berbice+Mork/Jerry patch+ATI inject and without ... Without patches, I get black screen (monitor self sleep) at boot and pressing monitor ON button I can reach desktop. Geekbench 4 Open CL benchmark is around 72000 in that conditions.
Using patches, I get desktop but Geekbench says around 60000. Makes sense?

1.) Sleep
2.) Benchmarks

1.) "hibernatemode 0" is regular 'ol classic sleep. If you open the Terminal app and type in : man pmset and hit return you will get a long explanation of the various options available for the kinds of sleep modes. Apple recommends using hibernate mode =0 or 3 or 25, but there are lots of other options available. These numbers are the decimal equivalent of a binary string of bits turned on or off. The internet has quite a few pages explaining the settings you can use, here is a brief explanation: http://blog.taylormcgann.com/2012/07/05/hibernate-from-terminal/

If you want to test these without pounding away at your SSD, you can change to location of your sleep image file to a hard drive, but remember that doing so will cause the wake up time to increase because the disk has to spin up out of sleep as well before it can be read. That file size is the same size as your RAM size, so make sure you have the space to spare. Since you asked, I personally am using hibernate mode = 0.

One thing to check is how well your power supply works for your processor family and chip set. It is not enough, just to match the power requirements for the system but the processor chip power states have be matched as well, because the various power levels the chip moves between have to be met by the power supply in a timely manner or things go weird in a hard to track down way. The bad news is that testing the power supply requires some serious bench testing equipment, so usually the best you can do if check the manufacturers' websites to see if they list the system compatibilities.

2) Benchmarks can vary wildly depending on the system load. Running them several times over differing use scenarios can give you a consistent view, but what you are listing is a big enough disparity that it is worth looking into. But I do not wish to read too much into your phrasing, because some folks have written they are doing something similar that seems to take the IGPU out of use, which does give such big differences. Could you elaborate on how you are dong the test as well as how you have your system set up regarding the IGPU?

 

[Corslayth]

Yes that's what I did. I might try changing the ID to EF in the bios to see if I don't need to inject the device id in clover any longer.

Hi Rhys, Did you tried boot on windows with the modified RX 480 ? Is possible?.

 

[sparco1987]

Hi Rhys, Did you tried boot on windows with the modified RX 480 ? Is possible?.

If your RX 480 have two bioses, you can select second and boot any OS. With modified can't .