[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Nvidia drivers

On Mon, 2 Apr 2001, Juha Saarinen wrote:

> I get hard lockups at random with my current set-up, but haven't been able
> to pinpoint what's causing it -- the Nvidia kernel driver, the kernel
> itself, X or whatever.

 Has it occurred to you that if there was more documentation and
source available, someone outside NV might have a chance of tracking
down the problems?

> Dave Konerding covered this issue very well -- if a hardware vendor
> releases the specs to all of its products, it might as well shut up shop
> that day because the clone makers will rip them off in an instant.

 This is like saying that car builders will be ripped off if they
tell people how do drive them.

> Apparently, neither ATi nor Matrox have given full access to Open Source
> developers, which means that their drivers aren't as good as the Windows
> ones.

 ATI have given enough to several DRI developers to enable them to
write and debug drivers.  From off-hand remarks on the dri-devel list
it appears they are even good enough to document know hardware issues
so they can be worked around in drivers.

> Something with even less support than nVidia's products?

 Come again?

 Bugs in the G400 driver have been fixed by folks with documentation.
It's quite unlikely anyone with those docs could recreate the G400 --
but they have been able to write a driver that is now extremely stable
and fast enough for me to enjoy Quake 3 and Unreal Tournament on a box
that is not exactly blazing in performance.

> > Again NVIdia Linux driver 90% speed of the Windows driver (that was the old
> > one).
> > Matrox Linux driver was something like 60% speed of the Windows driver.
> > Matrox is treating us as second class
> > customers.

 The Windows drivers were setting certain clock frequencies on the
card higher than the Linux ones.  There are options in the XFree86
CVS tree or DRI CVS tree to do the same.

> I think it would be helpful if ATi and Matrox were to release Linux
> drivers for their products. After all, they know them the best
> (presumably) and thus should be able to write decent drivers for them. So
> what if they're not Open Source...

 The ATi drivers have more folks looking at them.  They are being
worked on by folks with many years of experience programming graphics
hardware and software.  People who have helped to build the X server
and the Direct Rendering Infrastructure.

 The Matrox driver is working very well.  The DRI developers have
cured all the show-stopping issues I have had with it.  Apparently
the NV drivers are still triggering lock-ups for some people.  I
have watched a seasoned sysadmin here struggle to get the NV drivers
installed and working on a machine ... and read release notes which
clearly demonstrate some of the problems caused by trying to put
out binary drivers for Linux, such as the differences between the
different distributions.

 On the other hand, for my G400 at home, I can compile either the
current DRI CVS tree or XFree86 trunk, install, and start X up.  It
now works beautifully.  Quakeforge is playable at 1600x1200x32.

 I know from following the DRI development mailing list for the last
year that they were able to get to this stage because they had some
documentation.  I also know that changes in the Linux kernel were
tracked within days.  This would not happen at all quickly with a
binary-only driver distribution.

Bill Crawford, Unix Systems Developer, GTS Netcom
work: bill ops netcom net uk, home: billc netcomuk co uk

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]