- Moderator
- #581
Jeff Kilgroe
Well-known member
The issue on OSX is how the display driver interface coordinates with the PCIe subsystem. It's entirely a software issue, but how easy this would be to fix is hard to say. I'm thinking that it wouldn't be a big deal to make the necessary OS changes, however it my necessitate that all the video drivers be updated or re-written to work properly. I hope Apple gets it sorted, if not for the near term, I think having this ability will only make more sense in the future as both Thunderbolt and GPU abilities continue to expand.
There have been a couple attempts at hacks and work-arounds for this, some with limited amounts of success. The most successful attempt yet was the "Thunderbolterizer" hack which allowed for the Quadro 4000 GPU to be used via Thunderbolt. It wasn't entirely stable and I think the work-around's own issues nullified any potential performance benefit, but hey, it shows there is some hope.
Another possible solution could come from GPU vendors in the form of CUDA or OpenCL driver interfaces that circumvent the GDI on OSX. FWIW, the Xeon Phi cards will work via Thunderbolt on OSX and Linux (so I am told), but they are not video cards and do not use the display driver layer for their software.
There have been a couple attempts at hacks and work-arounds for this, some with limited amounts of success. The most successful attempt yet was the "Thunderbolterizer" hack which allowed for the Quadro 4000 GPU to be used via Thunderbolt. It wasn't entirely stable and I think the work-around's own issues nullified any potential performance benefit, but hey, it shows there is some hope.
Another possible solution could come from GPU vendors in the form of CUDA or OpenCL driver interfaces that circumvent the GDI on OSX. FWIW, the Xeon Phi cards will work via Thunderbolt on OSX and Linux (so I am told), but they are not video cards and do not use the display driver layer for their software.