130 replies to this topic

[akimoa]

since some real mac users flashed there ATI Cards with  lower clock settings  to get rid of glitches

why dont we  drop a rom with lower settings and try booting with that  

could that work  ? to reach the same effect ?

[Andy Vandijck]

akimoa, on 14 January 2013 - 08:52 PM, said:

since some real mac users flashed there ATI Cards with  lower clock settings  to get rid of glitches

why dont we  drop a rom with lower settings and try booting with that  

could that work  ? to reach the same effect ?

Yes, it could work.
Also it will give the opemu more time.

[Mad Marcsen]

hmm, they increased clock setting. my card also hat about 150mhz gpu/ 200mhz ram clock setting in power save mode. they suggested that you increase the settings to about 400/1000. it's more about deactivating the power-saving modes. for my card it didn't work, but there are a lot of mac people where that didn't work either. still has the feeling that this is a "mac" problem. but if someone would try to edit his graphics bios, i could give help.... did this yesterday all day long ... so i gained a little experience here. but if you could just do this with a rom......

but all the different ati-cards have other roms.... so one had to offer more roms for the different ati cards out there, or am i wrong?

[Deltac0]

So, underclocking GPU could work?
I could try with my old HD4650, if I can find it and the machine agrees to boot.

[happel]

Are we sure this isn't something opengl related? When I compare my 4870 (only glitches in chrome with flash as far as I know) and my 5450 (lots of glitches in safari etc).
The 4870 is way faster (more shader units, higher clockrate), so for me it's hard to believe this is speed related.

So I looked up some specifications, these below stood out as the only difference in capabilities (besides raw power of course).

		  4870	  5450
OpenGL version	  3.0	  3.2
Shader Model	  4.1	  5.0

[Bitcore]

happel, on 16 January 2013 - 11:21 AM, said:

Are we sure this isn't something opengl related? When I compare my 4870 (only glitches in chrome with flash as far as I know) and my 5450 (lots of glitches in safari etc).
The 4870 is way faster (more shader units, higher clockrate), so for me it's hard to believe this is speed related.

So I looked up some specifications, these below stood out as the only difference in capabilities (besides raw power of course).

4870 5450
OpenGL version 3.0 3.2
Shader Model 4.1 5.0

I think this could be the problem!
Is it true that HD5770 is woring OOB?
Because my little brother has graphic issues on that one to.
-About my HD6770 , after reinstalling with another installdisk i also have issues -

[akimoa]

yes that is true and your 6770 is Chipwise the same aswell AFAIK


wastez and myself tested with underclocked roms for 5770   without alot of changes in glitches /artifacts

so i guess that is a dead end .....

investigate more

Andy could u have a look inside the AMDRadeon kexts for any clue if you have a minute these days

[wastez]

happel, on 16 January 2013 - 11:21 AM, said:

Are we sure this isn't something opengl related? When I compare my 4870 (only glitches in chrome with flash as far as I know) and my 5450 (lots of glitches in safari etc).
The 4870 is way faster (more shader units, higher clockrate), so for me it's hard to believe this is speed related.

So I looked up some specifications, these below stood out as the only difference in capabilities (besides raw power of course).

4870 5450
OpenGL version 3.0 3.2
Shader Model 4.1 5.0

No it doesn´t have something to do with the speed. it´s related to qe/ci

[Bitcore]

Lol, i'm seeing my problem : i just had to reinstall the kexts for my card.
HD6770 full qe/ci

[wastez]

Bitcore, on 16 January 2013 - 03:30 PM, said:

Lol, i'm seeing my problem : i just had to reinstall the kexts for my card.
HD6770 full qe/ci

Without glitches?
Because it would be interessting if it works for 100% on a Bulldozer.
That could confirm andy´s theory about the ssse3 emulation and timeouts.

[Mad Marcsen]

akimoa, on 16 January 2013 - 02:41 PM, said:

yes that is true and your 6770 is Chipwise the same aswell AFAIK


wastez and myself tested with underclocked roms for 5770   without alot of changes in glitches /artifacts

so i guess that is a dead end .....

investigate more

Andy could u have a look inside the AMDRadeon kexts for any clue if you have a minute these days

did you try with "overclocked" roms, too? the thing is, that some people got rid of the glitches by changing the clock speed of certain powersaving modes of the video card so that the graphics card does not enter powermanagement mode. so normaly you just have to tweak one! clockspeed setting in the bios of the graphics card.

Bitcore, on 16 January 2013 - 03:30 PM, said:

Lol, i'm seeing my problem : i just had to reinstall the kexts for my card.
HD6770 full qe/ci

what kext's do you use for your 6770? and after reinstall you had problems with artifacts/glitches? and did you got glitches all over osx or just in certain tasks, eg. youtube or videplayback?

[wastez]

Akimoa and me played a little bit with roms today and we had a little success.

To disable power management seems to help.
The most glitches are gone, just some problems are still there. (Flash for example)

But i think this is not the the ultimate cause, it´s just a workaround.....

[Mad Marcsen]

wastez, on 17 January 2013 - 01:02 PM, said:

Akimoa and me played a little bit with roms today and we had a little success.

To disable power management seems to help.
The most glitches are gone, just some problems are still there. (Flash for example)

But i think this is not the the ultimate cause, it´s just a workaround.....

great. if disabling power management works for you should mean that this is not a problem with the amd kernel. because this workaround does work for real mac hardware, too. could you explain what you did? maybe it works for me, too.

in the meanwhile i tried the kext's from 10.6.8. copied ATI5000Controller, ATISupport, ATIFramebuffer and ATIRadeonx3000 to my /S/L/E. Everything is working except QE/CI. Card is also detected and all kext's are loaded. Maybe we can make them work?

Andy, do you think recompiling the kernel in 32bit could solve this?

[akimoa]

Quote

great. if disabling power management works for you should mean that this is not a problem with the amd kernel

Works .....is something i wouldnt call that atm .....its behaves diffrently sounds better

its far from useable .... i would say that would be a fix for Real Mac users but not for us since the core of the problem is still kernel related what i believe  .......

nope SL kexts no need to look at since the Acceleration is supplied/triggerd by the AMDRadeonAccelerator wich isnt present with the old kexts

try by yourself with the radeon bios editor and a chameleon version wich supports boot with rom  rev.1819 did it for me ->http://cl.ly/0x3E0s1c3x3f431R343w

be shure u name the rom the rite way otherwise it wont load and offcourse place it in extra folder together with the flag for ati roms....in the plist

[wastez]

Mad Marcsen, on 17 January 2013 - 01:46 PM, said:

great. if disabling power management works for you should mean that this is not a problem with the amd kernel. because this workaround does work for real mac hardware, too. could you explain what you did? maybe it works for me, too.

in the meanwhile i tried the kext's from 10.6.8. copied ATI5000Controller, ATISupport, ATIFramebuffer and ATIRadeonx3000 to my /S/L/E. Everything is working except QE/CI. Card is also detected and all kext's are loaded. Maybe we can make them work?

Andy, do you think recompiling the kernel in 32bit could solve this?

I´m also sure that the problem is still in the kernel, hope andys idea about ssse3emu times will be confirmed.
This doesn´t solve the main problem it just masks it.

Hope this insight will forward us to the main problem.

[Deltac0]

Mad Marcsen, on 17 January 2013 - 01:46 PM, said:

great. if disabling power management works for you should mean that this is not a problem with the amd kernel. because this workaround does work for real mac hardware, too. could you explain what you did? maybe it works for me, too.

in the meanwhile i tried the kext's from 10.6.8. copied ATI5000Controller, ATISupport, ATIFramebuffer and ATIRadeonx3000 to my /S/L/E. Everything is working except QE/CI. Card is also detected and all kext's are loaded. Maybe we can make them work?

Andy, do you think recompiling the kernel in 32bit could solve this?

How could we use the 32-bit kernel? ML kexts are 64-bit.

[Mad Marcsen]

so, seems like fx cpus have the same problems: http://www.insanelym...s/#entry1881675 and http://www.insanelym...20#entry1881693

question is, does the ssse3-emulator checks if ssse3 on the cpu is there and then disables itself or does ssse3 calls still go through the emu?

maybe it's really kernel related but has nothing to do with the emulator.

the problem is, for me, that all logs are fine. if at least there were some error messages about things going wrong, but nothing.

what patches are in the kernel now besides the in-kernel crypto and the emulator? are all of them needed or are there things that are don't needed any longer with the introduction of the emulator stuff? maybe we should start with a new kernel and applying patch after patch carefully with testing every single patch?

ah damn, it feels miserable if you can't code and are dependent on somebody who can.... kernel patching isn't that hard, did that quite often with linux. for compiling i only have xcode3, as i can't login to the appstore with ml.

another thing that perplexes me is the fact that the nvidia drivers have cpuid stuff. i mean why would apple do that for nvidia but not for ati?

[Bitcore]

wastez, on 17 January 2013 - 07:42 AM, said:

Without glitches?
Because it would be interessting if it works for 100% on a Bulldozer.
That could confirm andy´s theory about the ssse3 emulation and timeouts.

  Hi , yes without glitches.
But i'm not using andy's kernel, because it won't boot for me.

Bitcore

[wastez]

Bitcore, on 18 January 2013 - 11:47 AM, said:

Hi , yes without glitches.
But i'm not using andy's kernel, because it won't boot for me.

Bitcore

So which one are you using?

[Bitcore]

wastez, on 18 January 2013 - 12:09 PM, said:

So which one are you using?

Yeah, that's kind of a problem.
please check post 14 of this thread : http://www.insanelym...ng-amd-kernels/

Bitcore