Anyone mobile Kepler 6xxM working?

[TeoTheCool]

How did u activate the new ones, since they're unsupported with our cards?

[timp]

 

But in case of Keplers I really don't know where to dig - nvidia drivers, card configs, vga bios difference or something else.

 

 

Hay, What do you think about looking at how the displays are configured? I ask because I've been trying to get a K3000m in a Dell M6700 working. I've experienced much the same as you. However I've found that the internal LCD works, and with CE/CI, but the backlight is off. I can connect using desktop sharing and see in ioreg that the card is initialized and working. I can also change resolutions. So it's getting the internal LCD's EDID okay.

 

However when I connect an external display on any port it hangs the graphics system(I can still SSH in).

I also read that Kepler cards for the first time support 4 displays simultaneously. This implies some changes in the way the displays are connected and defined. Could it be that different things need to be injected for this new setup? One anomaly I noticed on the M6700, all the 4 displays have a 'port-number' property set to 9. Looking at the ioreg from an nvidia retina mac. It's port-numbers are 0, 1, 2, 3. If this is the 'port' in the same meaning as with Intel HD graphics. There is no way they should be the same.

[joshhh]

Yes, I think all our problems in display configs or NVCAP + some PC specific bios code.

 

It's good that you got QE/CI working even in VNC session. I didn't get it at all for now but try to reinstall OSX with latest 10.8.4 dev update E52 + new nvidia drivers.

 

On my system I got displays 0-1-2-3, but no action on connection any external display.

 

I've got some info in ATM window about graphics...

 

 

 

 

 

Will be great if you can save IOReg + System Profiler files via VNC + get all other dumps with DarwinDump 2.7.0

 

 

Hay, What do you think about looking at how the displays are configured? I ask because I've been trying to get a K3000m in a Dell M6700 working. I've experienced much the same as you. However I've found that the internal LCD works, and with CE/CI, but the backlight is off. I can connect using desktop sharing and see in ioreg that the card is initialized and working. I can also change resolutions. So it's getting the internal LCD's EDID okay.

 

However when I connect an external display on any port it hangs the graphics system(I can still SSH in).

I also read that Kepler cards for the first time support 4 displays simultaneously. This implies some changes in the way the displays are connected and defined. Could it be that different things need to be injected for this new setup? One anomaly I noticed on the M6700, all the 4 displays have a 'port-number' property set to 9. Looking at the ioreg from an nvidia retina mac. It's port-numbers are 0, 1, 2, 3. If this is the 'port' in the same meaning as with Intel HD graphics. There is no way they should be the same.

[timp]

Yes, I think all our problems in display configs or NVCAP + some PC specific bios code.

 

I've been digging around in the NVidia bios DCB tables. Seems quite clear that the contents are different between kepler mobile cards and others. Including, surprisingly the dump I have from a K4000 destop card. Could this be why Kepler desktop cards work but mobile cards don't? All the tables I have looked at are version 0x40.

 

As an example here is the DCB table from a Dell M6700 K3000m. Looks like some kind of multi line entries.. It very similar in the GTX680M bios I have too.

130F 0001 3400 0100
1F1F 0001 1000 0200
1F2F 0001 3000 0200
D65F 0108 2000 420F
925F 0108 2000 0200
C64F 0108 1000 420F
824F 0108 1000 0200
8F3F 0108 3000 0200
B68F 0304 1000 420F
7F8F 0304 1000 0200
A67F 0202 1000 420F
627F 0202 1000 0200
6F6F 0202 1000 0200
009F 0402 0000 0000
0F9F 0401 0000 0000
0F00 0000 0000 0000

 

And the code from a K4000(also similar to a FX3700m)

000F 0002 0000 0000
020F 0001 3000 0200
A61F 8102 1000 420F
621F 0102 1000 0200
B62F 8204 1000 420F
722F 0204 1000 0200
8F30 0300 3000 0200
0F00 0000 0000 0000
0F00 0000 0000 0000
0F00 .....

 

And the code from an iMac's 675mx

B601 8104 1000 220F
D612 8201 2000 220F
1212 0201 2000 0200
C624 8308 1000 220F
0224 0308 1000 0200
0E00 0000 0000 0000
0F00 0000 0000 0000
0F00...

 

I get the impression that the NVDAResman.kext is responsible for parsing the DCB table and getting the connectors set up right. I'm hoping to modify the DCB table in the K3000m dump and use Chameleon's video bios loader to load it at boot time. Thing is, what should go in the table? The code from Linux's Nouveau driver helps but it's not enough... See

http://lxr.free-elec...bdev/bios/dcb.h

http://lxr.free-elec...bdev/bios/dcb.c

 

 

Update..

Unfortunately using the UseNvidiaRom option with Chameleon(build 2238) makes the machine reboot or corrupts the display before even starting the kernel. So I can't try modding the DCB table..

[IDDPioneer]

@joshhh, @timp

 

Maybe we can patch the binary of the NVDAResman.kext to make it work with our mobile kepler, if it works OOB with Desktop Kepler, maybe patching some bytes we can make it to work but we need collaborators, organization, distribution of the work and mainly Compromising MINDS.

 

Interested people send mail here: iddpioneer_ymail.com - replace _ with @

[joshhh]

OK

Until last week I've got following possibilities of kepler stuck:

- GPU difference with desktop versions
- BIOS difference
- VBIOS difference
- DRIVERS

What I've figured out on last week

Desktop card + OSX = Working

Mobile MXM card + OSX = Working - on Alienware and Clevo laptops Mobile keplers forks fine and in system BIOS there is no sing of GPU vbios. So card on mobile GPU is independent as desktop solution

Mobile integrated card + OSX = Not Working



So we've got some differences in vbios and drivers. Our vbioses integrated as module of main laptop BIOS rom but in MacBooks it also integrated into main rom... and in macbooks it's working!


ALSO in attachment vbioses from mobile MXM GTX680 (GK104) and GTX660M (GK107) from Dell/Alienware and Clevo laptops

MXM-vbios-keplers.zip

 

@joshhh, @timp

Maybe we can patch the binary of the NVDAResman.kext to make it work with our mobile kepler, if it works OOB with Desktop Kepler, maybe patching some bytes we can make it to work but we need collaborators, organization, distribution of the work and mainly Compromising MINDS.

Interested people send mail here: iddpioneer_ymail.com - replace _ with @

[IDDPioneer]

Thanks for all the info, I am thinking if we can make it work by changin the dsdt or by modding the VBIOS do you have a bios of a real mac ?

[joshhh]

Yes! I've got all stuff from Macbook Pro Retina 15" in this post http://www.insanelymac.com/forum/topic/284767-anyone-mobile-kepler-6xxm-working/page__st__60#entry1915493

 

Thanks for all the info, I am thinking if we can make it work by changin the dsdt or by modding the VBIOS do you have a bios of a real mac ?

[kelobb]

Did You try loading any of this ROM from chameleon option, maybe our cards would be visible - with a bit of luck. Doubt it, but maybe it's worth to try.

[joshhh]

tried, didn't help

 

Did You try loading any of this ROM from chameleon option, maybe our cards would be visible - with a bit of luck. Doubt it, but maybe it's worth to try.

[joshhh]

Repeat, not all 6xx cards is Keplers

 

670M is Fermi, so don't post anything about it here

 

Try freezefix app by Dan542 works for my GTX 670m and 10.8.4 12E55.

ORIGINAL POST FROM SOMEone OSX WEBSITE:

 

Hello, people! My GTX 670m is working to 1 month without freeze. I'm using freezefix. - BY ME.

 

BY DAN542:

 

Just place it anywhere you like and run it, it'll be invisible (you may see a window with a white triangle blink, though). In case you want it to be opened automatically every time on start up, just add it in System Preferences, I believe it's somewhere under Users and Groups.

[joshhh]

Installed first beta of 10.9 and still no luck with Keplers, stuck after NVDAStartup official...

[joshhh]

Another attempt to find out what's wrong with mobile keplers

 

With additional debug info I get lines in log with NVDAHal NVDAgl registering - this is acceleration modules for QE/CI I get it working on retinabook. But when we remove NVDAResman we lost it.

 

So some checks in driver stop loading osx to GUI

 

Some shots from 10.8.5/10.9DP2 

 

 

 

 

 

 

[slayer2333]

Would you please let me know how you could get that much details in your verbose mode? I'm interrested in that.

[joshhh]

-v -f and debug=0x100 or debug=0x144 and maybe io=0xffffffff

 

Would you please let me know how you could get that much details in your verbose mode? I'm interrested in that.

[IDDPioneer]

Thanks joshhh by what I am seeing that "probe fail" is the problem, So I think we need to discover what that probe is, or hack the binary to pass by it as it passes by a desktop kepler.

[TeoTheCool]

I always wondered if netkas would know what to do in this case, since he seems like the best GPU "hacker" on the OSX scene imo. Maybe someone should point this recent find at him

[joshhh]

OK! I've decompiled NVDAResman with hopper disassembler

 

If someone can try to find probe function asm file is attached below 

 

NVDAResman.asm.zip

[IDDPioneer]

I´ve found 26 ocurrences of probe functions in the asm file.

 

And in the start of the line it may be the offsets of the binary file? if it is we should make a patch for the binary.

[joshhh]

Really don't know if all of 26 occurrences is related to this problem. But in case of bios whitelist patching most of offsets is relative and we don't need to worry about it... 

 

 

I´ve found 26 ocurrences of probe functions in the asm file.

 

And in the start of the line it may be the offsets of the binary file? if it is we should make a patch for the binary.

[Freshmaker™]

Hey guys, any new ideas on this one?

[joshhh]

Yeah! Only one idea - sell pc laptop and buy mac one

 

Hey guys, any new ideas on this one?

[IDDPioneer]

I am thinking in buying a new laptop With:

Intel Core i7 Haswell - 4700MQ

Nvidia GeForce GTX 765M GPU - 2GB Video RAM

This will be compatible ? or is better to take another with a GT 670M with fermi ?  

 

OBS: Remember that GT 765M  is GK106 NOT GK107. Maybe these chip codes can be the Aquiles foot of the hackintosh acceleration.

 

Did anyone knows that ?

[AREOS]

Thanks joshhh by what I am seeing that "probe fail" is the problem, So I think we need to discover what that probe is, or hack the binary to pass by it as it passes by a desktop kepler.

The "probe fail" is not related to your problem, it also happens in my fermi laptop and in a macbook with 320M and booth are working. does someone have an ioregistry explorer save from a laptop with a working Mobile Kepler card?

[joshhh]

Here it is, IOReg dump from my current Macbook Pro Retina 15 inch. After few month of kepler headaches with Lenovo W530 laptop I've sold it and bought Apple one.

 

Hope this info may help to people with Keplers in laptops

 

Andrew’s MacBook Pro.zip

 

The "probe fail" is not related to your problem, it also happens in my fermi laptop and in a macbook with 320M and booth are working. does someone have an ioregistry explorer save from a laptop with a working Mobile Kepler card?