OSX freeze after installing official nVidia Fermi drivers

[macgaga]

ok ... installing parallels with a DOS System ... and than do it .. I will give it a try

correct me but ... That guide is using other Tools and doing other Things

Thats what I menat a short guide for our Problem So it is a solution maybe Tonymac could make one + a Post one the Blog for the found solution ?!

 

[rtbas]

ok ... installing parallels with a DOS System ... and than do it .. I will give it a try

correct me but ... That guide is using other Tools and doing other Things

Thats what I menat a short guide for our Problem So it is a solution maybe Tonymac could make one + a Post one the Blog for the found solution ?!

The procedure goes like this:
1) Create a bootable DOS flash drive and put nvflash.exe on it.
2) Boot your computer from said flash drive and use nvflash.exe to extract your video card's bios.
3) Reboot into OS X. Open the Fermi BIOS editor in a Parallels virtual machine running Windows and use it to modify the extracted bios according to DrJohnnyFever's instructions. Save (or copy) the modified bios onto the bootable flash drive.
4) Restart and boot from the flash drive again. Use nvflash.exe to flash your card with the modified bios.

The details of how to create a DOS flash drive, how to use nvflash.exe and how to use the Fermi BIOS editor are covered in the guide which DrJohnnyFever kindly referenced:
http://forums.guru3d.com/showthread.php?s=10e47fd232e075b60cfe4a6c46841d0a&t=313989

(You should also be able to create the bootable flash drive in a Parallels virtual machine running Windows. When you insert the flash drive, simply connect it to Parallels instead of OS X and follow the directions in the guide mentioned above. I think this USB feature is only available in Parallels when you're using the 32-bit OS X kernel, but I'm not quite sure about that.)

 

[macgaga]

very kind of you !

This guide is nice.

Ok i will try but I have a 64bit OS X ... we´ll see


so whats the Status ? All modded Fermi runnig still ?

 

[rtbas]

My machine has been up for 29 hours, 53 minutes. This is the longest uptime I've seen since installing the gtx460. I don't want to jinx it though . It could be random chance that it's still working right now. I personally won't consider the problem solved until I see my machine running continuously for several days at least. I'll certainly keep everyone posted, whatever happens.

Even if it does end up working for my machine, I'm still not convinced that it's going to be a good general solution. From everything I've read, this problem seems to be very dependent on system configuration. Especially considering chancekang's comment, it seems likely that this fix may work for some hardware combinations but not for others.

 

[rtbas]

I just realized that my test may not be valid. In order to monitor uptime, I installed a dashboard widget called ShowOff Widget on my desktop. Dashboard relies on OpenGL, so who knows if that's (unintentionally) preventing the freezing/kps.

On the bright side, if it is the dashboard widget that's preventing the freezes, I've discovered a less intrusive fix than the screensaver desktop or looping video. I guess I have even more testing (aka waiting) to do.

 

[DrJohnnyFever]

Quarts Extreme is using OpenGL to draw the entire interface, so I wouldn't expect one widget to make the difference. That said, if it does work I'll try it.

I have 18 hours, 7 minutes up right now though. Still no crashes.

 

[rtbas]

I looked at the code for this particular widget and I noticed that it refreshes its display every two seconds. Then I noticed that if I use the "Show All Windows" feature of Expose as quickly as I can (i.e. hit F9 repeatedly), Expose has decreased performance while the widget is updating (a slight, but noticeable drop in framerate). These two things are what made me suspect Dashboard.

Of course I really hope that the bios modification is what's actually working, since it's a much better solution in my opinion .

It also occurred to me that DrJohnnyFever and I seem to be amongst the few people still running Core2's (at least in this thread). Besides that, we seem to have somewhat similar motherboards (depending on how similar you consider the P31 and P35 chipsets). Just an observation.

(And for the record, I'm now at 31 hrs, 04 min of uptime.)

 

[hawgguy]

I'm at around 21 hours (knock wood). I should have had a crash by now. Won't be truly convinced for another day or so.

 

[tonymacx86]

Thanks for the updates guys- at this point I've removed my GTS 450 and put it on the shelf. I'm hoping we have a fix that will allow me to put the 450 back in for good. Keep posting your results with this rom flash method- I don't really consider running a video or wallsaver a valid fix.

 

[scorcho]

voltage and clock speed modding helps, but there's much trial and error to be done to find the best mix of numbers without overheating the card. in my last attempt, i raised the lowest power state to 300/400 for core/memory, and evened out the voltage between the first two states. i kept the 3rd and 4th states the same.

while my system was much more stable - up to 28 hours straight at one point - it still crashed eventually. if i was more aggressive with raising the clocks/voltage such that there wasn't so much disparity between the various p-states, i think it would've worked better. oh, and a side benefit some of you might've noticed is that, by raising the idle clocks, OSX feels a good deal smoother in Spaces and Expose.

outside of updated NVIDIA drivers, i'm pretty sure this is the right track for stable-ish hackintoshing with the Fermis. some have speculated that Safari has caused more crashes with Fermis, which makes sense considering how much that browser leverages GPU rendering (the 3D-ish new tab landing page, google maps zoom/pan, etc) and thus cause very sudden p-state switching. if you're on Windows and use Firefox 4, which is completely GPU-accelerated, you'll see very quick p-state jumping as well. the difference, most likely, are more stable drivers on Windows that's forgiving to these spikes.