View Single Post
Old 09-11-09, 05:19 PM   #39
FreeBSD cheering section
uOpt's Avatar
Join Date: Nov 2004
Location: Boston, MA, USA
Posts: 609
Default Re: Will everyone PLEASE stop trashing the Nvidia Linux devs?

I absolutely do not trash the developers, as in the people who give us the drivers.

However, from where I stand, the following is true:
  • The quality was much better than it is now. As an approximation it has gone down when NVidia, with a fixed set of developers, had to take on more and more different cards and more and more APIs and API features.
    • There are numerous corruption problems with 180.x in particular, among other reported by me and a huge thread not involving me that see no solution. There was no reaction on part of the developers, although it is a clear, 100% reproducible regression from 17x.x that can be tested back and forth with a driver reinstall.
    • Users are getting into increasing trouble with the problem of more and more cards being pushed into legacy. In particular it is difficult to run a new Linux kernel and anything that's AGP or PCI (not PCIe), although not impossible. For me it kinda runs but the backtraces in syslog get on my nerves.
    • Standard Xorg features such as dual-seat are not supported.
    • No attempt is being made to have a card requiring NVidia's binary drivers and a card requiring Xorg's drivers to co-exist with a LD_LIBRARY_PATH userlevel switch.
  • NVidia refuses to solve the problem by giving documentation to do even a low-performance driver that does have basic 3D acceleration. This would be useful for people like me who do Google Earth, some blender and the like but do not game.
  • NVidia has not revised their stance on the former after ATI did start documenting many of their reasonably current cards in a way that Xorg now improves the drivers. In fact, NVidia has not reacted to the ATI move at all.
  • NVidia insists on keeping the OpenSource driver free of 3D and not supporting many essential non-3D features.

I am very well aware that NVidia doesn't have the manpower, and that is why I'm not bashing the men.

However, it is clear that NVidia is not quite reasonable if they think that without an increase in manpower they can deal with so many cards and so many APIs and API features.

Subsequently, NVidia is blind to even the possibility of letting OpenSource developers help with any of this.

Does the stuff work with Xen, anyway?

My Unix benchmark results
uOpt is offline   Reply With Quote