Mac Tesla C20X0 drivers? Why only win and linux

My excitement at seeing the Tesla C2050/70 announcement was dampened rather quickly by seeing that the support intended was only for certain flavours of Windows and Linux.

Can we expect to see (i) drivers for Snow Leopard (ii) a version with the extended EFI ROM support to allow Mac operation?

I could live without the ROM as a post-boot injection always seems to appear on the scene, but having Linux support but not Mac seems bizarre. We can always power the things externally if we have to, as many of us are already doing with GTX 295s and/or multiple 285s.

I hope Nvidia and Apple can work to sort this out before the hardware appears, at least in terms of drivers.

I can feel you pain, but Is it really that bizarre to support the dominant HPC operating system and not support the one that has almost no presence in the HPC space, given these new Tesla parts are squarely aimed at clusters?

I think your problem (missing Mac Tesla 2xxx) is more an card bios(EFI needed) problem, much less an driver problem !!
Hackintosh users (like me :) ) have less probs to get cards running that Apple official not support.
But at least they must be same group (gpu type) as the real Mac cards. Mostly all cards of same major gpu type work with the drivers, even the card belongs to an other subtype of that gpu.
Like NV 9600 GT (i had) run well with OS X, Apple drivers / Cuda / OpenCL, even not used in real macs.
I now switched to NV 8800GTX (EVGA) and very happy - also 8800GTX was never in Apple Store / nor made for Apple Macs.
Apple orig. is an 8800GT Mac (=80% of my 8800GTX), which is a bit different to 8800GTX, not only in MHz clocked.

I think it depends on whether one is using a Mac Pro vs a Hackintosh. I have a small Mac Pro HPC cluster (6 Pros and 2 Xserves) so really need this OS. My experience getting GTX 2xx cards to work in a Pro were that if you just need CUDA, for example, and do not therefore mind having the card live driving a monitor from boot or otherwise, the drivers were the key issue, and the Pro I am working on has a Mac 285 and a PC 285, with the boot coming from the first card. To get a card booting booting you do need the EFI for sure. On a Hackintosh I know a different set of tricks can be used but have not done that myself. So this is my belief that it is the driver that would get me up in a basic form.


Interesting that combination of 285 Mac (EVGA) and 285 PC card.
But does CUDA / OpenCL see that 285 PC card also at least as an compute device ?

Once injected post boot the PC card works fine under CUDA and OpenCL and also drives a monitor just like the Mac version. The only small downside is a smaller link speed (2.5 vs 5) but this has no noticeable effect unless you inspect the fine detail of CPU to GPU transfers. I only use that card for CUDA/OpenCL in fact, but others who have tried a 2G card like mine have reported a failure to recognize the memory in some ordinary OS X apps, which see 2G as -2G. This is a separate matter though.