250 replies to this topic

[bcc9]

piarullir, on 18 October 2012 - 04:45 PM, said:

OMG, v4 working for me! (9400 + 9200)

Great, so the 9400m part is working under 10.8 with full qe/ci (full resolution&acceleration), right?

piarullir, on 18 October 2012 - 04:45 PM, said:

@bcc9: tanks!! with linux i've used this solution: http://luizfar.wordp...-card-on-linux/

That's a new one...  But doesn't nvidia's linux driver (the closed source one) support the discrete card correctly?
In any case, yes, it looks like this is do-able via dsdt. It'd be a simple matter of calling  \_SB.PCI0.XVR0.Z01I.DGOF () from the _INI method.

Ok, here's a dsdt that includes all of the previous edits and also hopefully should keep the graphics coprocessor powered down.

Attached Files

•  dsdt-coprocessor.v5.zip   137.28K   35 downloads

[piarullir]

In Linux discrete card is not supported... it sucks only battery and produce heat...

bcc9, on 18 October 2012 - 05:24 PM, said:

Great, so the 9400m part is working under 10.8 with full qe/ci (full resolution&acceleration), right?

Yes, is working with full qe/ci... but there is a big problem... when i plug-in a second monitor (either VGA or HDMI) both monitors became full of artifacts and became unusable... whith no solution, i have to reboot! (plug-out second monitor doesn't help)
Any idea?

bcc9, on 18 October 2012 - 05:24 PM, said:

That's a new one...  But doesn't nvidia's linux driver (the closed source one) support the discrete card correctly?
In any case, yes, it looks like this is do-able via dsdt. It'd be a simple matter of calling  \_SB.PCI0.XVR0.Z01I.DGOF () from the _INI method.

Discrete card is not supported in linux, it sucks only battery... can we try this acpi call?

[bcc9]

piarullir, on 18 October 2012 - 05:56 PM, said:

not supported in linux, it sucks only battery... can we try this acpi call?

Yes see the new dsdt I just posted (stupid forum software merges posts together when they should not be)

[piarullir]

bcc9, on 18 October 2012 - 06:00 PM, said:

Yes see the new dsdt I just posted (stupid forum software merges posts together when they should not be)

I've tested just now,

system is working with full qe/ci, in system information "graphics informations" are completely disappeared... (for both 9440 and 9200)
and it tells me "Error retrieving informations" (but is still working)

i attach my ioreg -lw0

Attached Files

•  ioreg.txt   894.11K   3 downloads

[bcc9]

Hmm, check with lspci under osx to see whether the coprocessor seems disabled like under linux.

[piarullir]

bcc9, on 18 October 2012 - 06:33 PM, said:

Hmm, check with lspci under osx to see whether the coprocessor seems disabled like under linux.

i've just checked... no reference to

[/size][/color]
[color=#555555][size=3]02:00.0 VGA compatible controller: nVidia Corporation G98 [GeForce 9200M GS]

i can test this solution using the system and monitoring battery duration and heating

For the management of a second display do you have any idea?

[bcc9]

piarullir, on 18 October 2012 - 06:49 PM, said:

i've just checked... no reference to

[/size][/color]
[color=#555555][size=3]02:00.0 VGA compatible controller: nVidia Corporation G98 [GeForce 9200M GS]

i can test this solution using the system and monitoring battery duration and heating

For the management of a second display do you have any idea?

If the device isn't showing up at all with lspci then it seems to have been powered off correctly. Also not seeing pci memory resources for it  in your ioregistry so that looks good.
If you had a kill-a-watt, that'd probably be the easiest way to test for a power difference.
For plugging in via hdmi, I don't know, you're going to have to troubleshoot it some more I think.  I assume hdmi video was working for you just fine under 10.7?  Does it still work with 10.7 with the new dsdts?

[bfbruin]

v5 is still giving me the same result, but at least it's time for another ioreg output!

At this point, I feel like just opening up the chassis and smashing the non-9400M discrete graphics card with a hammer.

Attached Files

•  Terminal Saved Output v5.txt   788.02K   4 downloads

[bcc9]

bfbruin, on 18 October 2012 - 11:02 PM, said:

v5 is still giving me the same result, but at least it's time for another ioreg output!

At this point, I feel like just opening up the chassis and smashing the non-9400M discrete graphics card with a hammer.

Now your vendor-id is correctly remapped for your z01i device (and it looks powered down too).  So you have to go back and put back the nvidia kexts that you moved out of the way.  Then you can have working qe/ci with the 9400m part.

[bfbruin]

bcc9, on 18 October 2012 - 11:06 PM, said:

Now your vendor-id is correctly remapped for your z01i device (and it looks powered down too).  So you have to go back and put back the nvidia kexts that you moved out of the way.  Then you can have working qe/ci with the 9400m part.

Oh, praise be to you, bcc9. Good golly. Unnnnnnnnnghhh.

So to recap, I scrapped everything anyone else told me, installed dsdt-alt.aml v5 of your design, and ran the NVIDIA-supplied 10.8.2 drivers package (open this link, download, Right Click on pkg file, go to /Contents/Packages and run webDriverPayload_jLhsf.pkg) to restore GeForce.kext and ensure everything is of the latest quality and calibre.

And now, I have QE/CI. And it looks SO. DAMN. SMOOOTH. I mean, it's on its fourth year now, and still running strong. Thanks to you, I have the laptop equivalent of Helen Mirren.

bcc9, I cannot possibly thank you enough.

[piarullir]

bcc9, on 18 October 2012 - 09:25 PM, said:

For plugging in via hdmi, I don't know, you're going to have to troubleshoot it some more I think.  I assume hdmi video was working for you just fine under 10.7?  Does it still work with 10.7 with the new dsdts?

With older 10.7 installation dual monitor was ok (both vga and hdmi) but with older DSDT (in this moment i've deleted Lion installation...)

I've tested that booting in safe mode (-x at boot) make the video working properly and dual monitor works like a charm, but no qe/ci enabled... the problem is when the system loads the nvidia kext...

A little question... the DSDT edit i've suggested for disabling 9200m can replace your "z01i -> DSDT" edit?
in theory loading kext will skip the discrete card (the system ignore it)?

In this case we can try this combination for solve the problem? (if it make sense)

bfbruin, on 18 October 2012 - 11:48 PM, said:

And now, I have QE/CI. And it looks SO. DAMN. SMOOOTH. I mean, it's on its four year now, and still running strong.

bcc9, I cannot possibly thank you enough.

Good
can you try if dual monitor works?

Edit: the problem is generalized... that issue heppens every time the video card try to change or reset output resolution (for example when i try to change the resolution of my primary screen) thus seems to be no related to secondary output.

The good news is that when the issue happens you can solve it forcing a user logout (shift+cmd+Q then enter) and the system adjust your screens "automagically"

Any ideas?

I Forgot to tell that i've updated to 10.8.2

EDIT2: Problem Solved
Investigating on the net for the problem i've found that it's a nvidia driver problem... it was solved with the newest drivers (that was not released for our video card)

I've Downloaded this drivers http://www.nvidia.co...f02-driver.html
then i've extracted all contents with Pacifist and putted in /S/L/E

reboot and... solved! now works like a charm! dual monitor hot plug and changing resolution without problems

[bcc9]

bfbruin, on 18 October 2012 - 11:48 PM, said:

Oh, praise be to you, bcc9. Good golly. Unnnnnnnnnghhh.

Ha ha, thanks.  I'd think you guys wouldn't be so happy with a laptop whose graphics co-processor has never worked under osx.
Personally I've never needed nvidia's cuda drivers; the built-in osx ones have been OK for the 9400m.  I think you could have just put back the GeForce kext from 10.8.x instead.

piarullir, on 19 October 2012 - 12:41 AM, said:

A little question... the DSDT edit i've suggested for disabling 9200m can replace your "z01i -> DSDT" edit?
in theory loading kext will skip the discrete card (the system ignore it)?

So yes if the pci device for the z01i dsdt entry is not actually being detected as an active pci device then remapping the vendor-id shouldn't be necessary.
But I think we're getting ahead of ourselves here...  Does the dsdt mod to try and power off the coprocessor actually result in less power draw?  Won't anyone confirm/deny?  If it does truly power it down, then you could next try the dsdt without the vendor-id patch.  And also while you're at it the coprocessor power patch may also be simplified (I did some extra patching which may or may not be necessary).  
Otherwise, it's the coprocessor power patch that could be dropped.

If someone completes this testing then I'd know which part is necessary/sufficient to put into post #1.

[piarullir]

bcc9, on 19 October 2012 - 05:38 AM, said:

Personally I've never needed nvidia's cuda drivers; the built-in osx ones have been OK for the 9400m.  I think you could have just put back the GeForce kext from 10.8.x instead.

For me with 10.8.2 updating nvidia drivers was necessary for solving that annoying issue... (i've tried 10.8.x GeForce.kext with no luck)

bcc9, on 19 October 2012 - 05:38 AM, said:

If someone completes this testing then I'd know which part is necessary/sufficient to put into post #1.

There is a software for monitoring power consumption? otherwise i can make a test on the standby time for seeing difference (in the same conditions of course)
in linux the difference is noticeable (40-60 min of battery obtained)

PS: intuitively temperature seems to be substantially lower (but i haven't run stress test to confirm)

[ghatsby]

Thanks bcc9, v5 dsdt edit has got my 9400M G /G210 working. Only problem is that the system thinks there are no graphics cards in the profiler / system preferences. Also have no sleep function and I still have to use disabler to get any gui.....

Thanks in advance.

Attached Files

•  ioreg.txt   1.03MB   0 downloads

[bfbruin]

ghatsby, on 19 October 2012 - 03:18 PM, said:

Thanks bcc9, v5 dsdt edit has got my 9400M G /G210 working. Only problem is that the system thinks there are no graphics cards in the profiler / system preferences. Also have no sleep function and I still have to use disabler to get any gui.....

Thanks in advance.

Yeah, so mine has working sleep, and I do not need disabler.

Here are the steps that I used to finalize my setup:

• DSDT-alt v5
  
• Using Extra/Extensions AppleACPIPS2Nub, FakeSMC, VoodooBattery, VoodooPS2Controller
  
• org.chameleon.boot.plist is

Quote

<key>Kernel Flags</key>
<string>npci=0x2000</string>
<key>Timeout</key>
<string>5</string>
<key>PS2RestartFix</key>
<string>Yes</string>
<key>SystemId</key>
<string>18B471B3-1A99-4C89-8EB1-5E0517B6D426</string>
<key>GenerateCStates</key>
<string>Yes</string>

• Used latest Graphics Drivers from NVIDIA (see my previous post for details)
  
• Performed the two patches laid out in Post #1
  
• Running Chameleon v2.1svn r2061, and 10.8.2 with supplementary update
  
• Used KextWizard to assert correct permissions, make mkexts, etc.

Have you made any alterations to these steps?

[bcc9]

piarullir, on 19 October 2012 - 10:15 AM, said:

There is a software for monitoring power consumption? otherwise i can make a test on the standby time for seeing difference (in the same conditions of course)
in linux the difference is noticeable (40-60 min of battery obtained)

PS: intuitively temperature seems to be substantially lower (but i haven't run stress test to confirm)

The battery information gets you the details you need.  Or a kill-a-watt like I recommended a couple times now (and several times before now when we were testing the 9400m gstate power improvements).
You can check the battery information without wearing out your battery with a full cycle by checking  

system_profiler SPPowerDataType | grep 'Charge Remaining'

Time how long it takes an idle system to chew up the first 10% of the mAh capacity.  It should be measurably different if the co-processor is really using enough power to cause a heating issue.  Of course a kill-a-watt would be easier...

ghatsby, on 19 October 2012 - 03:18 PM, said:

Thanks bcc9, v5 dsdt edit has got my 9400M G /G210 working. Only problem is that the system thinks there are no graphics cards in the profiler / system preferences. Also have no sleep function and I still have to use disabler to get any gui.....

I don't know what disabler you're using or why.  No idea why you'd be having sleep problems either (I have no such problems).
You could just use v4 of my dsdt changes if v5 doesn't actually change the power consumption (nobody can say definitively yet?!)
Otherwise, perhaps someone could debug via dtrace what error the command:

system_profiler SPDisplaysDataType

is running into.

[bfbruin]

Using the first command, I get Charge Remaining (mAh): 5194 from my 9-cell.

Furthermore, with the second command, I get Segmentation fault: 11

[bcc9]

bfbruin, on 22 October 2012 - 06:20 PM, said:

Using the first command, I get Charge Remaining (mAh): 5194 from my 9-cell.

So if you checked the delta in mAh over a timed interval with dsdt v5 vs the same interval with dsdt v4 then you'd know if there is a difference in power usage.  This is assuming the two tests start at the same battery level and the system is in the same state (idle, screen on or off, etc.) across the two tests.

bfbruin, on 22 October 2012 - 06:20 PM, said:

Furthermore, with the second command, I get Segmentation fault: 11

Great, so a backtrace from gdb could probably point us to what the failure point is in obtaining the graphics info.

[piarullir]

Guys i have a lot of this HDA output on kernel messages:

Sound assertion in AppleHDAWidget at line 3095
Sound assertion in AppleHDAWidget at line 2175

repeated many times every minute...

it's normal?

[BWR]

Over the last couple years I have managed to keep my system running with my own tweaks here and there keeping it alive. For some reason though, with my 10.8.2 USB installer I am still unable to get graphics, using ANY of the new DSDTs including the latest one. For some reason when I boot into my 10.7.5 Install on the SSD ( graphics stopped working after the latest update in September) I can remote into the system using VNC but only after booting the system in safe mode. For some reason booting in verbose mode makes it kernel panic after the verbose output hits macx_swapon. Right before that line it prints NVDANV50HAL loaded and registered and then [AGPM Controller[ unknownPlatform followed again by the loaded and registered. The odd thing here is that when I am able to boot into it in safe mode with VNC, it doesn't seem to be detecting the display so I must be missing something. I can tell that the Installer for 10.8.2 has booted based on the blinking of my USB flash drive intermittently but the graphics never come up even after leaving it sit for a while (in the past the graphics would sometimes kick in automatically and the display would initialize . The problem to me seems to lie in display detection as the display doesn't even seem to power on at all now. Any input or diagnostics steps to troubleshoot would be much appreciated!
 1075_ioreg_nodisplay.txt   677.81K   3 downloads