r300/500 rant

[bridgman]

I think he was talking about dwindling support for the current OSS driver:

I suspect the problem here is not "dwindling support for the current OSS driver" as much as the fact that you need to look in new places to see the activity.

For 3xx-5xx 3D, the focus of activity is moving from the "classic" driver :

http://cgit.freedesktop.org/mesa/mes...ivers/dri/r300

to the "Gallium3D" driver :

http://cgit.freedesktop.org/mesa/mes...m/drivers/r300

... and for display/modesetting the activity is gradually moving from UMS :

http://cgit.freedesktop.org/xorg/dri...video-ati/log/

to KMS :

http://git.kernel.org/?p=linux/kerne...radeon-testing

... although a lot of changes are still being made in both places.

[deanjo]

Yep, although others (eg Ubuntu 8.04 LTS) should be good for another year or so.

The Gallium3D driver is making pretty good progress so hopefully one should arrive before t'other goes away.

Typical AMD/ATI response, "doesn't work, change distros". At least in Nv land legacy does not mean unsupported.

[bridgman]

Perhaps I'm being unfair to NVidia. A few people have posted here who have been unable to use their GF4s for a couple of years. I realize GF4 is back in the R300 timeframe but just using that as an example because there have been posts here about it. What would the NVidia response be (in 2010, not 2007) to those GF4 owners ?

[crumja]

bridgman is right. Nvidia's 71xx series of proprietary drivers have long since stopped support for the latest Xservers. This means old cards circa Geforce 2 don't work. If it weren't for nouveau, these cards would be nigh useless with the latest distros. Now the same slow/dropped support is happening with the 96xx series of drivers.

At least with ATI, OSS drivers will ensure that your card doesn't turn into a useless brick. Gallium3D and KMS support means that your card will take advantage of hardware acceleration on video, SVG, OpenCL, and other yet undefined future standards. I'm very satisfied with everything ATI has done.

[whizse]

On the subject of deprecation, there's some discussion about this on the mesa3d-dev list. Apparently there's still some interest in keeping the drivers for 3dfx and other oldies around:

http://www.mail-archive.com/mesa3d-d.../msg11751.html

[rohcQaH]

Typical AMD/ATI response, "doesn't work, change distros".

You can pick any distro you want, but then you'll need to manually downgrade a few packages. For some, just picking a different install CD is easier.

Of course the official response has to be "supported distros are still supported". *shrug*

[marek]

about to try gallium - wish me luck

Please let me know how it works for you (especially what works and what not).

they're making sure the R300-500 Gallium driver is working before they start on the R600+ driver.

The way I see it, it would be nice to make r300g (g=gallium) fast and stable first so that we later know what approach to take when designing r600g or possibly what code can be shared between the two.

-Marek

[DirtyHairy]

Since about one year, ATI has dropped support for my R500 based thinkpad T60 in fglrx. I am using gentoo (stable), so I eventually had to move to radeon. While this is able to serve my needs in most respects (I am no gamer), there is one _very_ serious flaw: power managment. Measured in watts, this has gone from 14 (when using powerplay) to 20-25 (using DynamicClocks) with the switch. I am a scientist relying on my notebook for work, and as I am traveling a fair amount of time, the reduction in battery live by about 30% is a serious issue.

I appreciate the work which is going into the opensource driver, but ATI has left me without proper GPU power managment for one year now, and I don't see stable support for serious power management for me in store anytime soon. The consequence for me is: my next machine most likely won't contain ATI hardware anymore.

[pingufunkybeat]

I don't see stable support for serious power management for me in store anytime soon.

I thought that dynamic power saving was coming in kernel 2.6.34?

It probably won't match fglrx just yet, but it should be a considerable improvement over current stuff.

[bridgman]

DirtyHairy; you mentioned DynamicClocks, I think that was replaced a while ago with three new options :

ClockGating (new name for DynamicClocks)
ForceLowPowerMode (reduces engine clock and restrices PCIE lanes)
DynamicPM (further reduces engine clock when display blanked via DPMS)

Are you using all three of these options ? It's possible the driver is rejecting the DynamicClocks option - check your log.