I know this has been asked before , but will Nvidia be coming out with drivers for the GTX470 / 480 anytime ? I had checked with Apple and this is what they told me to find out.
FYI - I use the card mainly for CUDA for Fluid computations…
It isn’t a driver you need, it is card firmware. The current PC VGA BIOS all the current Fermi product ship with doesn’t and can’t work on the Intel Mac hardware platform, which requires EFI firmware. Such firmware doesn’t exist and that is the biggest barrier to Fermi on Mac.
True - I will be forced to go with a GTX285 if there are no plans for a 4xx Mac release. I cant say I am thrilled to spend close to $450 for a GTX285 when I can get a 480 for 40 bucks more or a 470 for lesser !
So using some of the tips recently posted on forums.macrumors.com and insanelymac.com I have got a GTX 480 now driving my ACD 23 under OS X as well as bootcamp, on a 2008 Mac Pro. There is now a GF100 kext in my /System/Library/Extensions folder and the card loads OK with a post-boot injector the same way I used to load a PC 285, with a Mac 285 as the boot card. I am supplying power externally.
However, deviceQuery does not show the 480 (it does on the Windows side) despite me having loaded CUDA 3.1.
Could someone from Nvidia comment? Does the Mac CUDA kit allow Fermi access right now? Or is there something wrong elsewhere, e.g. with the OS drivers - the OpenGL benchmarks are also very slow.
BTW I disagree with avidday - the critical issue is not the EFI, but is the drivers. EFI firmware support is only needed if you want to use the card as the Mac’s booting card - the PC cards can usually loaded later with a post-boot injector, and I have done this fine with PC models of GTX 260, 285, and now 480, using 8800GT and later Mac 285 as boot cards.
Well the hackintosh folk often get things up slightly sooner due to ease of doing EFI workarounds but in this case I think we need clarity from Nvidia on OS drivers and CUDA lib.
How about it Nvidia? You already let the cat out of the bag with the (first version of) the last driver release for OS X, so why not release beta updates of the GF100 kext officially and let us work with it and the associated CUDA libs to get the gremlins out?