Also still missing is support for Fermi overclocking (overclocking the GeForce 400/500 series). Last August is when I mentioned that it was missing and NVIDIA confirmed they had it disabled in their Linux driver (but not under Windows) for all Fermi hardware. When testing out the NVIDIA GeForce GT 520 last week, I noticed the support was still missing when trying to enable CoolBits. I asked NVIDIA's Andy Ritger for a status update concerning Fermi overclocking on Linux, but he hasn't yet responded to that message from five days ago.
Disabling Fermi overclocking support under Linux doesn't make too much sense from the surface. Unlike Optimus, which requires some operating system-specific work and being able to interact with the xorg-server in the right way, the overclocking support should be relatively straight-forward. The GPU overclocking code should be mostly (or entirely) shared across OS platforms and it's working fine under Windows. NVIDIA still supports overclocking pre-Fermi hardware via their proprietary driver too. There's also the open-source reverse-engineered NVClock utility, but that's largely defunct and obviously doesn't have any Fermi support.
Beyond Fermi lacking overclocking support, multi-GPU SLI support was also initially disabled for the GeForce 400/500 hardware. I haven't heard whether or not this support has been enabled yet and I don't have the available hardware to test, but SLI is also supported on pre-Fermi GPUs.
What else is missing from the NVIDIA Linux driver? What about the other open and closed-source Linux graphics drivers? Share in the forums your experiences.