Jump to content

Mavericks kernel testing on AMD (formerly Mountain Lion kernel testing on AMD)

theconnactic

By theconnactic,
in OS X 10.9 (Mavericks)

 Share
6,414 posts in this topic

Recommended Posts

spakk
spakk
Posted
Posted

now I have downloaded and started the windows version,

that should be shown a experienced developers, as there is a source code viewer and other tools, which unfortunately I have no idea, maybe it will help us further

Link to comment
Share on other sites
+ira
+ira
Posted
Posted
Please test latest kernel_10.8.3_rc1 !

Download here !

 

Luckily the weekend is coming so i will be able to test it properly. For now, tried to boot with the sse42 kernel and get some kp's but i think it's because of my ML installation, will do a clean one soon as possible. Anyway, thanks for the kernel Bronya!

Link to comment
Share on other sites
theconnactic
theconnactic
Posted
  • Author
Posted

Hi, Bronzovka!

 

Tested all of your three new Mountain Lion kernels. At first, all of them would return a kernel panic, regardless of the flags used, backtracing a "faulty CPU", but i found out it was the SleepEnabler.kext. So, from the start, the kernels break the only kext that allows AMD machines to sleep/wake properly since 10.7.x.

 

Disabling the SleepEnabler with pmVersion=0, the boot hangs at Kernel=LP64 with both the ssse3 and the FX kernels, regardless of the flag. With the sse4.2, the system boots to the desktop, but the new kernel breaks my graphics: no QE/CI and heavy glicthes in the Dock, the Desktop, Safari and other applications (pic attached). This new kernels also requires the patched Corecrypto from the first page of the thread, and as a suggestion you could include it with the files.

 

All in all, i cannot recommend this new kernel for all AMD users running 10.8.3, specially if they use a non-FX CPU: they should stay with AnV's and_kernel for a while. But i'm confident, Bronzovka, that you'll come up soon with a revised kernel addressing all these issues, making it as stable and reliable as your latest Lion kernels. For now, thank you very much for your hard work, and keep'em coming!

 

All the best!

 

Screen Shot 2013-05-10 at 1.06.05 AM.png

  • Like 4
Link to comment
Share on other sites
+ira
+ira
Posted
Posted

This is how i get it to work for now:

 

Kernel: mach_sse42

SMBios: MacPro3,1

DSDT: None

Bootloader: Latest Chameleon

Flags: mach_sse42 -f -v -x npci=0x2000 GraphicsEnabler=Yes

Kexts: NullPowerManagement.kext, AnV corecrypto.kext

Modules: FileNVRAM, Keylayout, Resolution

System: OS X 10.8.3

Graphics: Stock + Cuda

 

Will try to explain the whys of this setup:

 

1. Without NPM kext it gives a KP on boot.

 

2. With FileNVRAM get rid of the EFIRuntime but it changes to EFINVRAM backtrace.

 

3. With corecrypto.kext get rid of EFINVRAM kernel panic.

 

5. With old Chameleon versions get black screen no matter flags i used, updated to latest Chameleon and used the GE flag.

 

6. Reach the installer with no use of -x flag, but after installation this is mandatory otherwise it gets on the gray screen with pointer/black screen with white square loop.

 

Graphics: no QE and some glitches on rendering some images.

  • Like 1
Link to comment
Share on other sites
Bronya
Bronya
Posted
Posted (edited)

I now did and working 10.8.2 or 10.8.3 without glitches artifacts !!!

1) Video with artifacts - no_efi_5770.rom.mp4 - because it is no_efi_5770.rom ! It is i add in aty,bin_image standard 5770.rom.hex in EvOEnabler.kext !

2)Video without artifacts - 5770.efi.mp4.zip - because with efi_5770.rom ! It is i add in aty,bin_image official 5770.efi.rom.hex in EvOEnabler.kext !!!

 

You think it is problem only mach_kernel ??? Or problem drivers???!!!

Edited by Bronya
  • Like 3
Link to comment
Share on other sites
theconnactic
theconnactic
Posted
  • Author
Posted

You think it is problem only mach_kernel ??? Or problem drivers???!!!

 

Personally, i think telling location of problem itself is the greatest challenge, because the very same kernel which gives artifacts on AMD machines will work just fine on Intel, even on older (non ssse3) CPUs, while Bulldozers (which does have ssse3) will suffer the artifacts like any other AMD CPU.

 

It might be the mach_kernel. It might as well be in the drivers, both nVidia and ATI. It might be even somewhere else. We know, as for now, what isn't the cause: the opemu isn't the cause, otherwise Pentium-Ds would have artifacts with the kernel. CPUIDs also aren't the cause, otherwise the CPUID patcher tool would solve the issue for good.

 

All the best!

  • Like 2
Link to comment
Share on other sites
Shaneee
Shaneee
Posted
Posted

You think it is problem only mach_kernel ??? Or problem drivers???!!!

 

I did some testing earlier with the OpenCL framework. Versions ranging from 10.7.2 to 10.8.4 and most gave different results. Some had more artefacts and some rendered the system completely useless. Not sure if it helps you get to the problem but I hope it does.

  • Like 2
Link to comment
Share on other sites
Bronya
Bronya
Posted
Posted

AMDRadoenAccelerator.kext and IOAcceleratorFamily.kext - it is QE/CI !!!

Natit.kext , ATY_Init.kext, EvOEnabler.kext don't influences on it is kext AMDRadoenAccelerator.kext !!! But EvOenabler.kext use aty,bin_image !

Need patch AMDRadoenAccelerator.kext ! Who can it did ? :)

  • Like 1
Link to comment
Share on other sites
Sannti
Sannti
Posted
Posted

For me this kernel don't work. After read from disk computer restart before booting.

My SMBios is iMac 12,1. I install OS 10.8.3 with iAtkos ML2 + Combo Update 10.8.3. With amd_kernel computer start properly.

 

I tried SSSE3 and SSE4.2 version.

Link to comment
Share on other sites
clemj06
clemj06
Posted
Posted

Bronya, Tried The kernel for the Phenom, boot more 'faster than amd_kernel

but the result is always artifacts

 

still a great job

I am sure that you will find a solution

 

Hey carlo_67, I was wondering, what bootflags did you use? And did you have any extra kexts?

Link to comment
Share on other sites
spakk
spakk
Posted
Posted

Please test latest kernel_10.8.3_rc1 !

Download here !

 

Hi Bronzovka, I am unfortunately only now gotten to test your new kernel. I get some errors that I still have to adjust because my current setting is adjusted to Andys amd_kernel. There is, however, I noticed something positive, your kernel is running super fast!

 

Figure 1 shows the boot with arch = i386

 

 

Figure 2 with arch=x86_64

 

PS:I have tested with Mountain Lion 10.8.2

 

but under CPU features I do not see SSSE3, what does this mean?

IMG_6102.JPG

IMG_6101.JPG

Link to comment
Share on other sites
+ira
+ira
Posted
Posted

:)

 

l'étau se resserre , je confirme tes pensées , nos problèmes graphiques (ATI) sont directement liés aux pilotes graphiques "nouvelle génération AMD" notamment toutes les cartes graphiques qui sont listés dans "AMDRadeonAccelerator.kext" les series HD 4xxx et antérieur qui n'y sont pas , fonctionnent correctement , je viens d'essayer ML DP1 en mode 64 bit sur l'HD 5450 512 et elle fonctionne correctement sans aucun bug , chose que je n'avais jamais essayé .

Ce qui confirme bien que toutes les cartes graphiques de HD 5xxx à HD 7xxx rencontrent exactement les mêmes problèmes .

 

the vice the narrows, I confirm your thoughts, graphics problems (ATI) are directly related to graphics drivers "new generation AMD" including all graphics cards that are listed in "AMDRadeonAccelerator.kext" HD 4xxx series and earlier that n 'are not working properly, I just try ML DP1 in 64-bit mode on the HD 5450 512 and it works properly without any bug, something I had never tried.

 

This confirms that all the graphics card HD 5xxx to HD 7xxx face exactly the same problems.

 

My thoughts about nVidia:

 

Well, maybe i'm saying just {censored}, but, if i was a kernel developer i would definitely blame this AMD feature.

 

The Hybrid CrossFireX is a technology allowing the IGP, or Integrated Graphics Processor, and the discrete GPU, or Graphics Processing Unit, to form a CrossFire setup to enhance the system capability to render 3D scenes, while the Hybrid CrossFire X technology is present on the 790GX and 890G chipsets, with two supplied physical PCI-E x16 slots at x8 bandwidth,[12] can form a Hybrid CrossFire X setup with two video cards and the IGP, enhancing the 3D rendering capabilities.[13]

 

This can possibly explain why do we had achieved only full resolution but QE/CI on 10.8 using AnV kernel.

 

One notable issue is that when the IGP paired with a DirectX 10.1 video card, the whole hybrid CrossFire setup will support up to DirectX 10.0 only, and the UVD in IGP will be shut down. The new 785G chipset has solved this issue however and supports DirectX 10.1.

 

I don't know what "DirectX 10.1 video card" means but this is an exemple of incompatibility with other cards.

 

In short words, my guess is that some AMD Chipsets/Processors technology is "blocking" the graphics acceleration of nVidia cards in 64bit mode.

  • Like 1
Link to comment
Share on other sites
ADHDMedia
ADHDMedia
Posted
Posted

I now did and working 10.8.2 or 10.8.3 without glitches artifacts !!!

1) Video with artifacts - no_efi_5770.rom.mp4 - because it is no_efi_5770.rom ! It is i add in aty,bin_image standard 5770.rom.hex in EvOEnabler.kext !

2)Video without artifacts - 5770.efi.mp4.zip - because with efi_5770.rom ! It is i add in aty,bin_image official 5770.efi.rom.hex in EvOEnabler.kext !!!

 

You think it is problem only mach_kernel ??? Or problem drivers???!!!

 

Hey buddy - I have a Radeon HD 5750, same series as the 5770, would it be possible for me to achieve the same result in 10.8.2/10.8.3 on my machine using this method? Can I get details from you?! Thanks man. Great work as always.

Link to comment
Share on other sites
 Share
Go to topic listing
×
×
  • Create New...