-->Donations via PayPal [External Link Removed for Guests] :guitar
-->Installation/Post Installation Tutorial [Local Link Removed for Guests]
-->Useful Links [Local Link Removed for Guests]
-->Olarila Images - Updated images [Local Link Removed for Guests]
-->Guides and Tutorials [Local Link Removed for Guests]
-->Clover Folder for All Chipsets [Local Link Removed for Guests]
-->DSDT Patch Requests [Local Link Removed for Guests]

AMD GPU on 10.14 with no iGPU: many issues

TheBloke
Posts: 29
Joined: 30 Oct 2018, 16:14
CPU: 1: X5670 2: i5-3550
Motherboard: 1: GA-X58A-UD3R 2: P8H77-M
GPU: 1: Vega 64 OC 2: R9 280X

AMD GPU on 10.14 with no iGPU: many issues

Post by TheBloke » 18 Sep 2019, 15:21

[Local Link Removed for Guests] wrote: [Local Link Removed for Guests]
18 Sep 2019, 13:28
TheBloke,

when you say 10.14.6 are you talking what build g97.. the latest with both supplementary updates or the base 10.14.6.. there have been at lease 2 supplementary updates since 10.14.6
I am on 18G95 - I assumed that that was the latest, given I only applied the update yesterday, and no further updates are shown as available? (Besides macOS 10.15 beta)

EDIT: Yes, found confirmation that 18G95 is supplemental 2, the latest version.
[Local Link Removed for Guests] wrote: [Local Link Removed for Guests]
18 Sep 2019, 13:28
I have read multiple threads where with WEG, you can either have DRM or encode, but not both at the same time.. seems nobody has broken the code to figure that one out yet with whatevergreen
For me, I think both are working. Certainly h264 and h265 encoding are OK, I tested today via Handbrake. I do have the issue I mentioned earlier where h264 encoding in Premiere Pro caused a GPU freeze, which is not yet diagnosed or resolved. But in general, both HW encoding and decoding appear to work.

As for DRM, I've tested Netflix in all of Google Chrome, Firefox and Safari. In Safari I hear sound but I just see a red screen. But it works fine in Chrome and Firefox. I also tested Amazon Prime streaming in Chrome, which also worked OK. Both with Netflix and Prime, I checked ACtivity Monitor while playing a video, and CPU usage was low, which I believe indicates the decoding was being accelerated. Also in Chrome I see a process called "Google Chrome Helper(renderer)" and another "Google Chrome Helper (GPU)" using small amounts of CPU while playing videos.

I think that shows DRM is working fine? It's not something I really use much, as I have another computer for watching streaming/movies etc.

EDIT: After doing some more Googling, I'm not sure my Netflix test prove that DRM works. The fact that it doesn't in Safari seems like a common issue, which obviously I haven't solved. And the fact that it does work in Chrome/Firefox possibly doesn't prove anything, given that I've read that on these browsers, it defaults to 720p rather than 1080p. There's a browser extension for Chrome that is meant to remove this restriction, but it's no longer on Play store and I can't be bothered to try installing it manually from Github.

I have now also tried some iTunes movie trailers, which played fine at full screen, with minimal CPU usage. I don't know if this proves anything or not, given they're only trailers? I'm not going to buy anything to see if that works :)
ONE: X58: G.byte GA-X58A-UD3R, Xeon X5670 @ 4.00Ghz, 48GB Corsair DDR3 1600
GPU:  Asus 7970Ghz (R9 280X) 3GB
SMB: MacPro 5,1. Clover: R4945 Legacy

TWO: H77: Asus P8H77-M, i5-3550 3.5Ghz, 24GB DDR3 1600
GPU: Gigabyte RX Vega 64 8GB OC
SMB: iMacPro 1,1. Clover: R5096 UEFI.
duece
Posts: 19
Joined: 17 Sep 2019, 22:04
CPU: w3680
Motherboard: Dx58
GPU: vega

AMD GPU on 10.14 with no iGPU: many issues

Post by duece » 18 Sep 2019, 18:28

TheBloke,

yeah. sorry g95 latest. end of aug supplemental .. slip of the keyboard fingers..

I just did a fresh instal of 10.14.6 g95 with unbeast clover 4930 I think

pretty much no options selected in config.plist and no DSDT .. just plain vanilla and just fakesmc and internet kexts

the system will boot and get graphics on my dp>HDMI to 4k TV... no hardware accel obviously and sleep no go.. but surprisingly IOREG has most things loaded so clover has improved a ton over the years


the standard aptiofix for tonycrap on this installer is plain aptiomemoryfix-64 I think and I get native nvram (whether its fully complete and working who knows)

once I plug in the HDMI to my regular monitor I get no video on the DP and the hdmi monitor is solid green

trying out WEG just for giggles if I can get it to force load from the ESP kexts folder otherwise I am going to run.me and sent this in to Madl0n to take a look at

macpro5,1 smbios
TheBloke
Posts: 29
Joined: 30 Oct 2018, 16:14
CPU: 1: X5670 2: i5-3550
Motherboard: 1: GA-X58A-UD3R 2: P8H77-M
GPU: 1: Vega 64 OC 2: R9 280X

AMD GPU on 10.14 with no iGPU: many issues

Post by TheBloke » 18 Sep 2019, 18:56

[Local Link Removed for Guests] wrote: [Local Link Removed for Guests]
18 Sep 2019, 18:28
once I plug in the HDMI to my regular monitor I get no video on the DP and the hdmi monitor is solid green
Sounds similar to what I see, except I've not seen any solid green. If I boot with only DP1 connected, everything is fine, but if any more monitors are connected before a sleep is done, then all monitors are black, and sleep fails.

If your situation is like mine, then sleep & wake is going to be your only solution for the time being (pending any magic from mald0n), and I doubt WEG is the difference between being able to sleep or not. You're very likely going to need a proper DSDT to get the ability to sleep.

Certainly I've needed a DSDT for a long time. I could get away without one in 10.11 and 10.12, but the first time I tried to upgrade to Sierra my system went haywire - not only wouldn't it boot, but every time I tried to boot, it changed all my BIOS settings to random values :D Which given I had a finely tuned and complex OC, was quite a big annoyance :)

Admittedly my HW is even older than yours, and not even UEFI. But still; being unable to sleep is a very common situation of not having a proper DSDT, from what I've read.

I eventually came back about 6 months later, and went straight to High Sierra, and this time learnt about doing things from scratch. I got a proper DSDT, thanks to a lengthy thread on InsanelyMac regarding the Gigabyte X58 series of motherboards, and that fixed sleep/wake, power management/speedstep, onboard audio, and a whole bunch more things.

Not sure why you'd need to force load WEG? Just put it into EFI/CLOVER/kexts/Other? Along with Lilu of course.
ONE: X58: G.byte GA-X58A-UD3R, Xeon X5670 @ 4.00Ghz, 48GB Corsair DDR3 1600
GPU:  Asus 7970Ghz (R9 280X) 3GB
SMB: MacPro 5,1. Clover: R4945 Legacy

TWO: H77: Asus P8H77-M, i5-3550 3.5Ghz, 24GB DDR3 1600
GPU: Gigabyte RX Vega 64 8GB OC
SMB: iMacPro 1,1. Clover: R5096 UEFI.
User avatar
MaLd0n
Site Admin
Site Admin
Posts: 14545
Joined: 05 Dec 2010, 19:35
CPU: i9-9900K
Motherboard: GA Z390 M GAMING
GPU: RX 590 8 GB
Location: Rio de Janeiro, Brazil
Contact:

AMD GPU on 10.14 with no iGPU: many issues

Post by MaLd0n » 18 Sep 2019, 19:27

TheBloke, check if sleep work, use power button for wake
DSDT._PRW.zip
You do not have the required permissions to view the files attached to this post.
:guitar[External Link Removed for Guests]:guitar-[Local Link Removed for Guests]:guitar

Problems with Paypal Donate? Check [Local Link Removed for Guests]

-Donated? Send me one PM-
duece
Posts: 19
Joined: 17 Sep 2019, 22:04
CPU: w3680
Motherboard: Dx58
GPU: vega

AMD GPU on 10.14 with no iGPU: many issues

Post by duece » 18 Sep 2019, 20:06

TheBloke,

I think part of the problem with this DX58so.. which has to be older than yours because it was Intels reverence board for x58 is that when they addd UEFI via a firmware update, its really no a full UEFI implementation

there are no settings for it in bios, no CSM on/off, no UEFI boot menu maintenance. hence one of my issues right now where clover put something into the rom of the board for a UEFI volume that doesn't exist and I can't get rid of it.. so I can't just let the motherboard boot to disk, I have to F10 to the boot menu on EVERY boot and choose the drive.. the one and only installed, otherwise it goes nowhere

I still think that this issue with the ports in 10.14.6 is solvable but likely instead of putting GFX in DSDT, likely a VEGA specific SSDT with proper port definitions, signal, etc from the RADION compatibility guide on how to figure all that out.. requires getting VBIOS dumps etc..
TheBloke
Posts: 29
Joined: 30 Oct 2018, 16:14
CPU: 1: X5670 2: i5-3550
Motherboard: 1: GA-X58A-UD3R 2: P8H77-M
GPU: 1: Vega 64 OC 2: R9 280X

AMD GPU on 10.14 with no iGPU: many issues

Post by TheBloke » 18 Sep 2019, 20:11

MaLd0n, short answer: yes I can sleep & wake OK with new DSDT you provided.


Full results:
  1. First boot, one monitor connected (port DP1): Boot to login screen. Click sleep. Machine sleeps. Press power to wake, machine wakes OK.
    • Then I log in. I try sleep again. Machine sleeps. Press power to wake, machine powers on, but does not properly wake: no picture on monitor, machine cannot be pinged on network. I had to press reset button.
    • But I had this same problem one time earlier today as well, so this is probably not connected to new DSDT.
  2. Second boot (same one monitor), boot to login & sleep/wake OK.
    • Then sleep again, this time it wakes fine.
    As you said, I could only wake with power button, not keyboard.

    Note: I did one test booting with multiple monitors connected, and there was no change - black screen, could not sleep. Probably this is expected, I just mention to be complete.

    Thanks again for all the work you are doing for me, it is really appreciated.
ONE: X58: G.byte GA-X58A-UD3R, Xeon X5670 @ 4.00Ghz, 48GB Corsair DDR3 1600
GPU:  Asus 7970Ghz (R9 280X) 3GB
SMB: MacPro 5,1. Clover: R4945 Legacy

TWO: H77: Asus P8H77-M, i5-3550 3.5Ghz, 24GB DDR3 1600
GPU: Gigabyte RX Vega 64 8GB OC
SMB: iMacPro 1,1. Clover: R5096 UEFI.
duece
Posts: 19
Joined: 17 Sep 2019, 22:04
CPU: w3680
Motherboard: Dx58
GPU: vega

AMD GPU on 10.14 with no iGPU: many issues

Post by duece » 18 Sep 2019, 20:27

MaLd0n,

here you go Madl0n.. I have prided myself in getting by hackintoshing over the year, reading, trying but I have been banging my head on this for over a month... and getting nowhere.

I would prefer to try the MP5,1 route first, since I have many expensive software packages tied to my MP serial number. but if we have to go the iMac Pro route .. well I will figure something out

clover must have made great strides over the last couple years..

as I commented above.. this is a very vanilla install using multishit, and its generic 5,1 def. I only added ethernet kext and fakesmc and got 4k60 on DP, HDMI audio on that DP, and it looks like my cpu was being power and step managed. adding lilu,WEG,ALC got me line out on the alc889 without having a single line in the DSDT or patches in clover..

sleep obviously doesn't work.. and in both 5,1 and iMac pro , I only have 1 monitor, if I try to plug in or boot with the second monitor at HDMI, black screen, no signal , green screen ..etc

I have included my previous DSDT for you to look at .. I don't use that airport card right now so it doesn't need to be in the final product.. this dsdt give me full sleep, wake with usb, graphics on gtx660, HDMI audio, alc889, it just works great.... with almost no patches in clover enabled..

I would like to get this 6 port sata card working though, I don't have it plugged in for this right now.. but it attaches at PEX1@1c1 and funny it took the ARPT label in ioreg when I booted into el cap with it installed. it has kept me from booting 10.14 though on this motherboard.. I can boot it on a different machine running Mojave just fine. .can even boot osx from it.,. so I know its compatible.. perhaps with a fully functioning DSDT and good interrupt handling it will be plug and play

thanks again for your assistance!!!

Added in 4 minutes 54 seconds:
TheBloke,

sounds like this might be un-fixable with current drivers post 10.14.5

I think perhaps we might want to consider either rolling back to that point release or figure out which kexts are offending and try and bring those forward into 10.14.6 ... I doubt we will see apple release any new meaningful updates to Mojave if history repeats.. this is now EOL for major updates... security fixes only probably
You do not have the required permissions to view the files attached to this post.
User avatar
MaLd0n
Site Admin
Site Admin
Posts: 14545
Joined: 05 Dec 2010, 19:35
CPU: i9-9900K
Motherboard: GA Z390 M GAMING
GPU: RX 590 8 GB
Location: Rio de Janeiro, Brazil
Contact:

AMD GPU on 10.14 with no iGPU: many issues

Post by MaLd0n » 18 Sep 2019, 21:03

duece,
DSDT.duece1.zip
DSDT.duece2.zip
test two files and let me know
You do not have the required permissions to view the files attached to this post.
:guitar[External Link Removed for Guests]:guitar-[Local Link Removed for Guests]:guitar

Problems with Paypal Donate? Check [Local Link Removed for Guests]

-Donated? Send me one PM-
TheBloke
Posts: 29
Joined: 30 Oct 2018, 16:14
CPU: 1: X5670 2: i5-3550
Motherboard: 1: GA-X58A-UD3R 2: P8H77-M
GPU: 1: Vega 64 OC 2: R9 280X

AMD GPU on 10.14 with no iGPU: many issues

Post by TheBloke » 18 Sep 2019, 21:46

[Local Link Removed for Guests] wrote: [Local Link Removed for Guests]
18 Sep 2019, 20:32
sounds like this might be un-fixable with current drivers post 10.14.5
What makes you say that? Have you read something to that effect?

All I know is that I've been living with the same general issue since 10.13, first starting with my my previous GPU the R9 280X, which is now installed in my second Hack, the H77. It's worse on the Vega in that I can only boot with one monitor connected, else I can't sleep and have to reboot. But the general principle of needing to sleep in order to see a picture on all monitors has existed for me since I first installed my R9 280X over a year ago, in 10.13.x. It changed a bit in 10.14.0 in that I also started to get errors in Console, and the kernel using high CPU until a sleep was done. But the fundamental issue has remained unchanged - and unsolvable by me, despite tens of hours of trying things like FB patching, SSDTs, and every possible permutation of WEG/Clover options. Or rather, trying the options I understood - which is certainly not all of them.

So based on my own experiences, I'm not seeing .6 being any different. It certainly hasn't affected my symptoms thus far.

All that said, I'm certainly willing to test in .5 if there's some reason to think it would make a difference. I can restore a network backup to my spare SSD; my most recent one is from June which should be 10.14.5. Worst case the backup will be 10.14.1, in which case I can upgrade to .5 via the [External Link Removed for Guests].
ONE: X58: G.byte GA-X58A-UD3R, Xeon X5670 @ 4.00Ghz, 48GB Corsair DDR3 1600
GPU:  Asus 7970Ghz (R9 280X) 3GB
SMB: MacPro 5,1. Clover: R4945 Legacy

TWO: H77: Asus P8H77-M, i5-3550 3.5Ghz, 24GB DDR3 1600
GPU: Gigabyte RX Vega 64 8GB OC
SMB: iMacPro 1,1. Clover: R5096 UEFI.
duece
Posts: 19
Joined: 17 Sep 2019, 22:04
CPU: w3680
Motherboard: Dx58
GPU: vega

AMD GPU on 10.14 with no iGPU: many issues

Post by duece » 18 Sep 2019, 22:45

MaLd0n,

10.14.6

made no changes to the generic config.plist of the fresh install.. not my config.plist from my previous el cap

no sleep either one

your GFX0 implementation at PEG3 is a bit unusual ... perhaps not what you intended?

I don't think that the EGPU device is even being picked up being overshadowed or GFX0 being grabbed by whatever green.. as it was able to do with no DSDT edits at all...
You do not have the required permissions to view the files attached to this post.
User avatar
MaLd0n
Site Admin
Site Admin
Posts: 14545
Joined: 05 Dec 2010, 19:35
CPU: i9-9900K
Motherboard: GA Z390 M GAMING
GPU: RX 590 8 GB
Location: Rio de Janeiro, Brazil
Contact:

AMD GPU on 10.14 with no iGPU: many issues

Post by MaLd0n » 19 Sep 2019, 01:18

duece, the device don't exist and pegp is createt weg kext rename to gfx0 on the fly

the best is create all missing devices

Added in 51 seconds:
EDIT

here, without gfx0 in tree
DSDT.aml.zip
You do not have the required permissions to view the files attached to this post.
:guitar[External Link Removed for Guests]:guitar-[Local Link Removed for Guests]:guitar

Problems with Paypal Donate? Check [Local Link Removed for Guests]

-Donated? Send me one PM-
TheBloke
Posts: 29
Joined: 30 Oct 2018, 16:14
CPU: 1: X5670 2: i5-3550
Motherboard: 1: GA-X58A-UD3R 2: P8H77-M
GPU: 1: Vega 64 OC 2: R9 280X

AMD GPU on 10.14 with no iGPU: many issues

Post by TheBloke » 19 Sep 2019, 07:32

MaLd0n, FYI I noticed that with latest DSDT you provided, my processor is always at x12 multiplier, like this:

EDIT: Sorry, this is not related to new DSDT/config.plist. I have done more testing and same happens with old as well. It appears that before sleep, I only have x12 multiplier. After sleep & wake, I have full x22-x24, and this is same with both my original DSDT/config.plist, and the ones you have sent me.

I do not know for sure if this is new behaviour or not, as before I would always put my machine to sleep at login screen. So maybe this issue has always been there.
Last edited by TheBloke on 19 Sep 2019, 14:53, edited 3 times in total.
ONE: X58: G.byte GA-X58A-UD3R, Xeon X5670 @ 4.00Ghz, 48GB Corsair DDR3 1600
GPU:  Asus 7970Ghz (R9 280X) 3GB
SMB: MacPro 5,1. Clover: R4945 Legacy

TWO: H77: Asus P8H77-M, i5-3550 3.5Ghz, 24GB DDR3 1600
GPU: Gigabyte RX Vega 64 8GB OC
SMB: iMacPro 1,1. Clover: R5096 UEFI.
duece
Posts: 19
Joined: 17 Sep 2019, 22:04
CPU: w3680
Motherboard: Dx58
GPU: vega

AMD GPU on 10.14 with no iGPU: many issues

Post by duece » 19 Sep 2019, 12:15

MaLd0n,

MaLd0n,

I understand where you are going with it, but for example on PEG3 you had GFX0 and HDAU devices, then added PEGP in the same path also with GFX0 and HDAU and I think WEG latched on to the first instance. Graphics card was not showing up in you FABULOUS !! system profiler PCI listing man that is really nice!! So I was thinking that perhaps having multiple devices with the same name was throwing things off

I am not at the computer at the moment but will test out later. Looking at the DSDT though, did you do the edits and patches of the NEW raw DSDT that was in the run.me zip or add/remove patches from my already patched DSDT? Only reason I ask is that I see old Toleda method HDEF patches that are likely no longer required using LILU/ALC and that probably should have stayed AZAL? and you added / or kept APRT in at PEX, and I really don't have that installed now. I also can't remember if I deleted more devices but see PS2K/M and perhaps the big one, serial port .. and that one has a NOIRQ that is like 1,2,3,4,5,6,7,8,9,..... basically taking every irq..

Also, the my old patched DSDT was one bios revision down... I had to update bios for the new graphics card. from 5599 to 5600 IIRC.
the el cap install still seems to run fine with my old DSDT and clover config.

also, I think I had some lengthly edits in the USB devices... I might have to have them in there for sleep... again can't remember.

I notice at least one rename TMR to TIMR but left LCP instead of LCPB.. just wondering what kind of patching flags this DSDT is designed to run with ? should I be selecting Drop Unused for example? And do some renames on the fly?

also there is a warning for and external value \UMAP at the start of the DSDT.. is that a SSDT table or some other external I need to have in my ACPI folder?

muchas gracias senior ... really... sorry for the dumb questions.. trying to learn something as I go since you are far more skilled than I !!

Added in 3 minutes 24 seconds:
TheBloke,

Man I am jealous of you guys that run the gigabyte boards... I get very little sensor data from the intel boards because of my chipset... not mainstream enough for the community to include my values!!!

what does the clover boot.log say about your processor? it should recognize it.. and the x5670 was in real macs IIRC? perhaps you have selected something in clover configurator that is patching your processor and loosing native speed stepping ?
TheBloke
Posts: 29
Joined: 30 Oct 2018, 16:14
CPU: 1: X5670 2: i5-3550
Motherboard: 1: GA-X58A-UD3R 2: P8H77-M
GPU: 1: Vega 64 OC 2: R9 280X

AMD GPU on 10.14 with no iGPU: many issues

Post by TheBloke » 19 Sep 2019, 15:24

[Local Link Removed for Guests] wrote: [Local Link Removed for Guests]
19 Sep 2019, 12:18
Man I am jealous of you guys that run the gigabyte boards... I get very little sensor data from the intel boards because of my chipset... not mainstream enough for the community to include my values!!!
Ah yeah, the HWMonitor stuff is useful. I take it you've tried the various FakeSMC variations to confirm none offer more info? I re-did that comparison yesterday, comparing two:

[External Link Removed for Guests]
[External Link Removed for Guests]

The former is the one I have used for a long time, and is the one I'm still using now. It provides the info you saw earlier, with its own bundled HWMonitor.app. The latter I tried in case it provided more GPU data on the Vega. It doesn't have a bundled HWMonitor, so I used [External Link Removed for Guests], which looks a bit different to HWMonitor, and does have one useful feature - it has an option for reading GPU data from IOKit, meaning it can read all the utilisation/performance stats that GPUs like the Vega 64 now write there (and which is also accessible from command line using ioreg).

I ended up uninstalling vit9696's VirtualSMC and going back to RehabMan's, even though that's not been updated in a year or so. And because HWMonitor is the only one that shows the Frequency info, and HWMonitorSMC2 is the only one that shows the Vega 64 info, I now have both running :)

[External Link Removed for Guests]
what does the clover boot.log say about your processor? it should recognize it.. and the x5670 was in real macs IIRC? perhaps you have selected something in clover configurator that is patching your processor and loosing native speed stepping ?
I've edited my previous post - turns out this issue existed also with my original DSDT and config.plist as well. It appears that this is yet another issue that requires a sleep & wake: before sleep, my CPU won't go above x12. After sleep, it's normal x12-x22, with x24 on turbo boosting cores.

I have no idea how long this issue has existed for, because for the last 1+ years I have always put the system to sleep as soon as I got to the login screen, due to my ongoing GPU issues. I know that I haven't had the issue always, because back in High Sierra there was a time when I used an NVidia GPU, and didn't have to sleep. I know my frequencies were right then.

So it could be the issue has come with 10.14.0 or a later update. Or it could be a result of switching to iMacPro 1.1. No idea. Right now it's moot, because I have to sleep & wake shortly after boot anyway. If I ever get that solved, I'll investigate it further.
ONE: X58: G.byte GA-X58A-UD3R, Xeon X5670 @ 4.00Ghz, 48GB Corsair DDR3 1600
GPU:  Asus 7970Ghz (R9 280X) 3GB
SMB: MacPro 5,1. Clover: R4945 Legacy

TWO: H77: Asus P8H77-M, i5-3550 3.5Ghz, 24GB DDR3 1600
GPU: Gigabyte RX Vega 64 8GB OC
SMB: iMacPro 1,1. Clover: R5096 UEFI.
duece
Posts: 19
Joined: 17 Sep 2019, 22:04
CPU: w3680
Motherboard: Dx58
GPU: vega

AMD GPU on 10.14 with no iGPU: many issues

Post by duece » 21 Sep 2019, 12:32

TheBloke,

in the 13-14 years I have been building these things I have never found one that gives me more than basic info on my intel boards.. gluten for punishment .. I have a badaxe 1 775, a badaxe2 775, and DX58so and they were all more difficult to hack than their gigabyte brethren .. I have a EP45 board and that one had pretty good sensor support..

if and when I do this again and abandon the DX58 which I really don't want to do because the W3680 is good enough for me... it will be a much more mainstream board with better support .. I am getting too old spending a month working 6 hours a night trying to get this ironed out

MaLd0n,

I have not been able to get sleep working on the fresh install, but went back to my previous attempt which was a fresh install to .5 or .6, used migration assistant to pull in my el cap drive, and was upgraded to .6u2 .. went back to just my basic DSDT from my old install, my old config.plist that has almost nothing turned on.. and I am back to having sleep.

if I use the same procedures as the bloke.. I can boot the 4k60 TV at DP1 > HDMI adapter and pretty reliable, sleep wake, etc

if I try and boot just the HDMI monitor, black screen with the system running behind it logging GPU.restarts in the console panic/spindump section

if I boot with 4k60 and plug HDMI in , crash to black screen, system running.. can ssh in ect.. but sudo shutdown -h now just sits there.. wont turn off although drive activity ceases so I give it a minute and hard reset

if I boot 4k60 with the other HDMI monitor plugged in but OFF, not sleeping, OFF, and get to the login screen, sleep the system THEN turn on the HDMI monitor and wake, I have both monitors and can continue to sleep and wake them as a pair.. testing short term only right now.. sleept it all night last night will update later

not using darkwake in boot args which is strange since it only worked in the past with no or =0 .. using dart (even though VT-d disabled), SHIKI-96 SHIKI with imac boardid -rad24 (fix 30 bit color)

been experimenting with ATI deinit in clover but inconclusive right now..

the epic fail on the reinstall and the old install might be clover above 5000.. I have read more than 1 thread now where some folks are having issues with it. I went back to 4932 and using aptiomemoryfix and booting UEFI, native nvram not using EMU and RC scripts, and for the moment it looks more promising but the VEGA monitor init issues are plaguing .. latest LILU/WEG/ALC although with my DSDT I get the VEGA video and HDMI audio without WEG. but can't turn on hardware encode decode with boot args without it.. would need to hex edit AGDP to get hardware support if I drop WEG

if I use the iMacpro SMBIOS I loose speedstep and my multipliers are locked at 12 instead of stepping to 26-27 under load

I will send you some of the GPU.restart dumps later.. and will send you a .dsl of what maciasl dumps from my running system.. perhaps you can dif merge with your patches and figure out where the sleep error lies.. at first I thought it was the USB method for _DSM .. mine uses store and yours uses return .. tried editing yours to store, didn't make a difference .. could be just the clover version.. doing more testing today if I can find the time

thanks again for the help!!
TheBloke
Posts: 29
Joined: 30 Oct 2018, 16:14
CPU: 1: X5670 2: i5-3550
Motherboard: 1: GA-X58A-UD3R 2: P8H77-M
GPU: 1: Vega 64 OC 2: R9 280X

AMD GPU on 10.14 with no iGPU: many issues

Post by TheBloke » 21 Sep 2019, 13:32

I am using dart=0, and no darkwake line because mald0n's config didn't have it, and I never knew if it was doing anything anyway. I can sleep & wake fine, although sometimes it will sleep but then not wake up properly - the power light comes on, but displays never wake up, and the machine doesn't respond to ping. I imagine it's GPU related as it first started happening since I installed the Vega.

I tried RadeonDeinit a couple of times and didn't notice it making any difference. However Clover's InjectATI makes a very important difference for me. Specifically:
1. It changes my GPU name from "Radeon RX Vega 64" (WEG only) to "AMD Radeon RX Vega 64" (WEG + InjectATI)
2. It changes the System Info->Graphics/Displays Metal info from: "Metal: Supported, macOS GPUFamily 2 v1" (WEG only) to "Metal: Supported" (WEG + InjectATI)
3. My Geekbench 4 score increases from ~145k (WEG only) to ~200k (WEG + InjectATI)

I don't understand quite what InjectATI does, or why it makes this important difference, but clearly I need it for full GPU performance. It's possible literally all it is doing is changing the GPU name, in which case I could likely achieve the same thing with a WEG SSDT. But I haven't tested that.

My current situation:
1. Using iMacPro 1.1 SMBIOS + mald0n's config.plist + mald0n's DSDT.aml. Plus I have added Clover's InjectATI to config.plist, which I need for full GPU performance.
2. From boot, I do not have working CPU PM: CPUs are locked at x12 multiplier
3. After sleep & wake, power management works OK: CPU multipliers x12 - x24
4. I have working h264 + h265 HW encode/decode. However, trying to use h264 encode often results in GPU freezes. This may be related to OC settings, as I have also had occasional issues in Windows 10. But, it's much worse in macOS.
5. I cannot use MacPro 5.1 SMBIOS as my 4K@60 display (DisplayPort 1.2) will never connect at 4K@60. It either connects at 2084x1080, or 4K@30, but almost never at 4K@60.
6. With iMacPro 1.1 SMBIOS, my 4K display always connects at 4K@60. However, it often does not display a picture. After sleep&wake I often have to unplug/replug it multiple times until it shows a picture. And I need to disable Display Sleep, as if the display goes to sleep, it will again lose the picture on the 4K after the displays wake up.
7. In Console, I get occasional errors like these:

Code: [Local Link Removed for Guests]

2019-09-21 12:05:27.467141+0100  localhost kernel[0]: (IOAcceleratorFamily2) <IOAcceleratorFamily2`IOAccelEventMachine2::waitForStamp(int, unsigned int, unsigned int*)> virtual IOReturn IOAccelEventMachine2::waitForStamp(int32_t, stamp_t, stamp_t *): initial wait for 1 second expired. Continue wait for 4 seconds. stamp 1 (gpu_stamp=0)
2019-09-21 12:05:31.467235+0100  localhost kernel[0]: (IOAcceleratorFamily2) <IOAcceleratorFamily2`IOAccelEventMachine2::waitForStamp(int, unsigned int, unsigned int*)> virtual IOReturn IOAccelEventMachine2::waitForStamp(int32_t, stamp_t, stamp_t *): timeout waiting for AMDRadeonAccelerator stamp 1 (gpu_stamp=0)
2019-09-21 12:10:03.491975+0100  localhost kernel[0]: (IOAcceleratorFamily2) <IOAcceleratorFamily2`IOAccelFenceMachine::fence_timeout(IOTimerEventSource*)> void IOAccelFenceMachine::fence_timeout(IOTimerEventSource *): AMDRadeonAccelerator prodding blockFenceInterrupt
2019-09-21 12:10:03.601648+0100  localhost kernel[0]: (IOAcceleratorFamily2) <IOAcceleratorFamily2`IOAccelFenceMachine::fence_timeout(IOTimerEventSource*)> void IOAccelFenceMachine::fence_timeout(IOTimerEventSource *): AMDRadeonAccelerator prodding blockFenceInterrupt
2019-09-21 13:22:36.958127+0100  localhost kernel[0]: (IOAcceleratorFamily2) <IOAcceleratorFamily2`IOAccelDisplayPipe2::event_interrupt_gated()> AMDRadeonAccelerator [0]: transaction ID (165047) generated error (0xe00002d5). dirtyBits=0x1, options=0x1
2019-09-21 14:22:32.075500+0100  localhost kernel[0]: (IOAcceleratorFamily2) <IOAcceleratorFamily2`IOAccelDisplayPipe2::event_interrupt_gated()> AMDRadeonAccelerator [0]: transaction ID (184995) generated error (0xe00002d5). dirtyBits=0x1, options=0x1
All those errors are since my last boot, 2.5 hours ago. They are not terminal, but maybe result in slowdowns or something, I am not sure. However if I do extended h264 encoding, I will sometimes get a bunch of the blockFenceInterrupt errors, plus errors about GPURestart, and then all displays crash, and the system has to be rebooted.

MaLd0n, do you think there is any further improvement that can be made to my system? I know you have done a lot of work on it already. I'd be grateful if you could let me know if you think anything more can be done.

I am planning to buy a new PC sometime soon - my first 'new' PC in 9 years (new to me, though I will buy it used.) I am thinking maybe an HP Z840 workstation, as I need more PCIe lanes than modern CPUs have, eg for 10Gbe ethernet. But even if/when I get a new PC, I will still hope to use this one as a secondary system with a modern AMD GPU, eg RX 580.
ONE: X58: G.byte GA-X58A-UD3R, Xeon X5670 @ 4.00Ghz, 48GB Corsair DDR3 1600
GPU:  Asus 7970Ghz (R9 280X) 3GB
SMB: MacPro 5,1. Clover: R4945 Legacy

TWO: H77: Asus P8H77-M, i5-3550 3.5Ghz, 24GB DDR3 1600
GPU: Gigabyte RX Vega 64 8GB OC
SMB: iMacPro 1,1. Clover: R5096 UEFI.
duece
Posts: 19
Joined: 17 Sep 2019, 22:04
CPU: w3680
Motherboard: Dx58
GPU: vega

AMD GPU on 10.14 with no iGPU: many issues

Post by duece » 21 Sep 2019, 14:09

[Local Link Removed for Guests] wrote: [Local Link Removed for Guests]
21 Sep 2019, 13:32
eleratorFamily2) <IOAcceleratorFamily2`IOAccelDisplayPipe2::event_interrupt_gated()> AMDRadeonAccelerator [0]: transaction ID (184995) generated error (0xe00002d5). dirtyBits=0x1, options=0x1
I noticed I think in your files you sent up that you have other cards installed in this machine .. not just the vega..

I found while testing that my machine didn't like some cards... without a specific edit in DSDT .. most times installing supported card would break sleep.. but for example.. now that mojave completely breaks support for SIL3132 based sata cards.. I went out an bought a support card that uses Marvel and ASM controllers... I can't boot fine with it installed into el capitan with the gtx660 installed .. but the same machine, same DSDT and config.. wont boot mojave.. black screen... BUT the system is running .. BIGGER BUT.. the card is working and the 3 drive apple raid hanging on it is WORKING.. the files are there but had to ssh in to see that .. no video...

the VEGA is a very strange cat.. and SLICE was saying that there are known issues with later AMD cards and older chipsets.. likely no UEFI for you and a poor and old UEFI implementation for me.. because the same card installed in a $90 HP6300sff running mojave that took 15minutes to install mojave on .. .works just fine and doesn't block video from the installed gx710 .. I can even BOOT mojave off the $40 sata add in card.. it has 6 ports and its the only time I have ever seen an add in sata card boot OS X ... so the card is very compatible .. but on these old boards.. I think that most modern hardware is looking for UEFI support

I know that with the DX58.. when I added the vega I was running bios 5599 from a long time ago.. like say 2012.. no video.. at all not even post if I recall.. updated to 5600 which was intel last bios from say 2013.. so still very old .. but obviously they updated the UEFI option rom for graphics cards because now I have video at post...

in the old days apples hardware and COTS hardware were more similar and the OS X codebase was a bit more generically written.. today with the advent of allot of apple only tech like biometrics, T2, customized chipsets and integrated video and custom drivers and frameworks that arent used on the PC side.. hackintoshing has thinner margins for working hardware.. and 'working' is different .. lots of tradeoffs..

for example your woes with imacpro .. some of the hardware encoding that is done with that model lean HEAVILY on that T2 according to bearfeat, .. I think a lot of hardware offloading in macs is very difficult to get right in the hack because the frameworks send the work to chips we don't have .. on a pc the hardware encode would be VEGA only and the driver would send it all to the VCE stack.. in apple its drivers definitely don't work that way ..

something else to consider.. since out bioses are years behind.. is that intel has been constantly releasing updated patches for their CPUs last few years as part of the security patches ongoing from heart bleed, and others that have not even been publicly released.. well apple pushes out these firmware updates with the releases and say for macpro5,1 which hadn't had many new firmwares has had like 4 in the last year... so if apples kernel is optimized for these new microcodes and we are not getting these microcode patches in firmware . the instruction sets don't match up... BOOM

Added in 39 minutes 2 seconds:
TheBloke,

update

my machine not liking the INJECT ATI with deinit

when booted same way single monitor.. GPU fan full up, power bar on the card pulsates and I am using 200W at the wall instead of 120~
so clearly with macpro5,1 SMBIOS .. injecting is messing with GPU power management AND my luxmark scores are LOWER.. which makes no sense
duece
Posts: 19
Joined: 17 Sep 2019, 22:04
CPU: w3680
Motherboard: Dx58
GPU: vega

AMD GPU on 10.14 with no iGPU: many issues

Post by duece » 21 Sep 2019, 16:17

MaLd0n,

here are some of the gpu.restart dumps
You do not have the required permissions to view the files attached to this post.
User avatar
MaLd0n
Site Admin
Site Admin
Posts: 14545
Joined: 05 Dec 2010, 19:35
CPU: i9-9900K
Motherboard: GA Z390 M GAMING
GPU: RX 590 8 GB
Location: Rio de Janeiro, Brazil
Contact:

AMD GPU on 10.14 with no iGPU: many issues

Post by MaLd0n » 21 Sep 2019, 16:31

duece, for solve problems with hardware accel, systeem freezes and other problems after high sierra, the best is use IntelQuickSync
Image
Image
Image

after it system don't crash more, encoder/decoder is very very fast

check one real mac apple dump
Image :guitar

in this case just a iMacPro1,1 smbios for check if solve, i think
:guitar[External Link Removed for Guests]:guitar-[Local Link Removed for Guests]:guitar

Problems with Paypal Donate? Check [Local Link Removed for Guests]

-Donated? Send me one PM-
duece
Posts: 19
Joined: 17 Sep 2019, 22:04
CPU: w3680
Motherboard: Dx58
GPU: vega

AMD GPU on 10.14 with no iGPU: many issues

Post by duece » 21 Sep 2019, 16:47

MaLd0n,

i agree... intel quick sync would be an answer.. if I had a motherboard chipset or CPU that supported it

I am on a x58 platform (socket 1366) and a Westmere W3680 processor .. hence .. no intel quick sync due to no CPU embedded GPU .. hence all my questions as to why your DSDT has multiple GFX0 > PEGP > GFX0 etc conversions...

only have one dGPU at PEG3 ... I define it as GFXO to give WEG the place to latch on

have not tested your third DSDT yet but the previous 2 have no sleep and make no improvements over my already edited DSDT as far as VEGA booting multiple monitors

perhaps you ran the wrong autopach against it thinking I had sandy bridge or something
Post Reply

Return to “Mojave (10.14)”