-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

Jonathan Dye wrote: | On Sun, Apr 18, 2004 at 11:17:36AM +0000, Tristan Scott wrote: | |>Bugger. its kernel 2.6.4 |>~ uname -a |>Linux temp.home.tristanscott.co.uk 2.6.4-rc1 #11 Sun Apr 18 00:46:10 GMT |>2004 i686 AMD Athlon(tm) XP 2000+ AuthenticAMD GNU/Linux | | | Have you loadeded/compiled the modules for the agp chipset as well as | the generic agpgart module? I made this mistake when I moved to 2.6, I | needed to load agpgart and via_agp to get my 3D acceleration to work | properly.

yep. built in or as modules. neither works. see other response

| | JD |

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

adam@thebowery.co.uk wrote: | On Sun, Apr 18, 2004 at 10:54:05AM +0000, Tristan Scott wrote: | <snip> | Hmmmm, non-free software, I shouldn't really reply to this but then I also use | the Nvidia non-free driver on my desktop (I unfortunately have an addiction to | FPS games). The nv driver does double-buffering but not glx (therefore no ut2004 et al)

| Does XFree86 actually work? as in you have a GUI but no AGP? I ask as this Yup, double-buffering seems to be broken too. 3d work respectably, such as ut2004 in 800x600, every effect off, ~30 fps. consistent with a pci card methinks. | isn't clear in the original email. If XFree86 won't start then it could be | something to do with the glx bits needing to be installed in a different manner they seem to work. /usr/lib/tls does not exist, and nvidia 5336 via gentoo's portage made no mention of it. | to work with kernel 2.6, I had a problem with that on my Debian machine in the | past. | | Anyhow I would recommend that you checkout the Nvidia support forum which is | linked to from Nvidias site (and Nvidia people do post there) and can be found | at http://www.nvnews.net/vbulletin/forumdisplay.php?s=&forumid=14 you will | probably find a solution there, or some people who have the same problems. Will do... in the morning *looks at clock* later this morning I mean. | | Adam

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

Tristan Scott wrote: | Havent found anyone else on nvnews or gentoo forums with my problem.. or | even a particularly similar problem. I ought to have mentioned I'd | looked at forums, but thought nvnews might be new one... well, it was | early morning. | | Changed kernel to 2.6.6 and rebuilt config from scratch. | agpgart built in - no go.

Compiled it out, tried nvAGP. nope. Compiled it back in, changed XF86Config from nvAGP to AGPGART, rebuilt with the non-gentoo-recommended 5336 drivers using nvidia installer. AND IT WORKED! 4n1c8 in ~ [26916]: cat /proc/driver/nvidia/agp/status Status: Enabled Driver: AGPGART AGP Rate: 4x Fast Writes: Disabled SBA: Disabled

no idea what i did though. ive tried the nvidia drivers before via nvidia installer. this combination hasnt been tried - nvidia 5336, compiled in agpgart, ti4600, 2.6.6 kernel

Lesson for all you people out there: Dont buy a 00:00.0 Host bridge: VIA Technologies, Inc. VT8375 [KM266/KL266] Host Bridge

| changed card to a Ti4600 (from Ti4400) - no go | | The AGP did work back in the days before windows stopped working, with | either card. | | I'm thinking this motherboard will never do it under linux, people have | mentioned agp/via problems, but are always saved by switching to nvAGP | or agpgart. | | If you want to nvidia driver to work under 2.6.6, you've got to leave 4k | stacks off, lave register arguments off and reverse a module move patch. | While it will function with register arguments on, its unstable i'm | told. i didnt try it. | | Tristan Scott wrote: | | adam@thebowery.co.uk wrote: | | | On Sun, Apr 18, 2004 at 10:54:05AM +0000, Tristan Scott wrote: | | | | | <snip> | | | Hmmmm, non-free software, I shouldn't really reply to this but then I | | also use | | | the Nvidia non-free driver on my desktop (I unfortunately have an | | addiction to | | | FPS games). | | The nv driver does double-buffering but not glx (therefore no ut2004 | et al) | | | | | Does XFree86 actually work? as in you have a GUI but no AGP? I ask as | | this | | Yup, double-buffering seems to be broken too. 3d work respectably, such | | as ut2004 in 800x600, every effect off, ~30 fps. consistent with a pci | | card methinks. | | | isn't clear in the original email. If XFree86 won't start then it | | could be | | | something to do with the glx bits needing to be installed in a | | different manner | | they seem to work. /usr/lib/tls does not exist, and nvidia 5336 via | | gentoo's portage made no mention of it. | | | to work with kernel 2.6, I had a problem with that on my Debian | | machine in the | | | past. | | | | | | Anyhow I would recommend that you checkout the Nvidia support forum | | which is | | | linked to from Nvidias site (and Nvidia people do post there) and can | | be found | | | at http://www.nvnews.net/vbulletin/forumdisplay.php?s=&forumid=14 you | | will | | | probably find a solution there, or some people who have the same | | problems. | | Will do... in the morning *looks at clock* later this morning I mean. | | | | | | Adam | | | | _______________________________________________ | main@lists.alug.org.uk | http://www.alug.org.uk/ | http://lists.alug.org.uk/mailman/listinfo/main | Unsubscribe? See message headers or the web site above! | |

_______________________________________________ main@lists.alug.org.uk http://www.alug.org.uk/ http://lists.alug.org.uk/mailman/listinfo/main Unsubscribe? See message headers or the web site above!

On Tuesday 20 April 2004 12:59, adam@thebowery.co.uk wrote:

I would have thought the lesson learned would be don't use Gentoo or buy an Nvidia card to use with Linux :)

Can't argue with the first bit, as I haven't really spent a lot of time playing with Gentoo.

But

If you want/need functional 3D with comparative performance to a Windows system, what other options are there ?

At the moment if you buy any Nvidia card there is a reasonable chance that you can make it work, yes they are closed drivers, no it doesn't work in 100% of cases (in this case I think it was more to do with AGP support on the chipset rather than the drivers) and yes if Nvidia opened their driver spec we could have nice open source drivers, but the point is it does work.

Not all ATI cards are linux friendly, the drivers they offer are also not open and no other manufacturer I can think of can offer such performance.

If Nvidia moved some IP from the driver to the card then maybe they could open the drivers, but then we would have a more expensive card and one that probably didn't get incremental performance improvements with every driver release.

I am pretty sure that if Nvidia could open up the drivers without some huge financial or intellectual cost then they would. Why would they not, why maintain linux drivers and pay people to write installers etc when there is a community that would do it for free ?

The reason I think is that there is either a. Some clever stuff going on in the driver that they don't want competitors to see or b. Some licensed technology that they are not allowed to release.

W

On Tuesday 20 April 2004 22:43, adam@thebowery.co.uk wrote:

Of course this means that a few bits will get dropped from the driver *but* what would be great if they make the kernel module free software and let the community improve the driver and do the same with the OpenGL implementation. Then if you need the non-free bits of the driver that are patent and license encumbered then you could resort to a non-free library that Nvidia could distribute seperately.

I guess that depends on what exactly are the juicy bits, I find it suspicious that the reverse engineering/OS community have completely failed to extract any sort of 3D capability from the card whatsoever, Quite complicated things have been successfully worked out in the past. So I am guessing that there is something very weird going on with the way the driver and card interact when doing 3D.

This also ties up very well with past problems Nvidia have had with their cards being fussy about chipsets/AGP settings/memory types/Power Supplies/ BIOS versions/Price of tea etc....What if say Nvidia know a way of making the AGP bus work in a more efficient way, and what if that mode is the only way you can talk 3D to the card......purely speculation I know but what if that "interface" is the bit Nvidia can't let out ? then they could only really give us what we already have with the "nv" XFree driver.

On Wed, Apr 21, 2004 at 12:36:41AM +0100, adam@thebowery.co.uk wrote:

My guess is that reverse engineering the drivers is just too hard, apparently the Linux driver is based on the Windows NT driver (as in very heavily, i did read somewhere that it even emulates parts of the Windows registry!) and I figure that people who care about the driver being free have bought an older ATI card and those that don't care just use the Nvidia driver. I could imagine that reverse engineering a driver/card to produce a driver worse than one that is already available isn't a very attractive option for most people.

Yup - swat I did, I go me an ATI Radeon 9200 (and very pleased with it I am too) for my home machine... And am just not using 3d on the Nvidea card that I got for my work machine (might get round to it one day, if I decide that it's worth having non-free, binary nastiness, drivers in my kernel).

What would be *really* nice is if someone designed some open hardware for talking GL over AGP... and then someone wrote so nice GPL drivers for it... But I think that'd involve too much outlay in the first case, and that it's probably not of the highest importance to the people in the know.

Right - back to work.

On Thursday 22 April 2004 00:13, MJ Ray wrote:

I don't think it suspicious. This hardware seems to be made more troublesome by the manufacturer producing their own inVidious proprietary drivers, which reduces the number putting effort into a free software one, as well as the common problems with lack of manufacturer documentation.

So are you saying that Nvidia producing a closed linux driver actually has a more negative impact on the community than then doing nothing for Linux at all. I'd never thought of it that way.

Maybe this is why we see so little Linux support from other manufacturers. They either can't get their heads around (or can't for technical/political or legal reasons) releasing open drivers or open hardware documentation. They think as you do that releasing proprietary drivers simply stops the development of an open solution, so figure the best they can do for linux support is nothing and let the reverse engineering gurus figure it out.

I see the logic behind your argument, but I still feel the motivation behind Nvidia not providing open drivers or proper hardware documentation is more a technical or legal one.

There is far more to the Nvidia drivers than just hardware interfacing, how else do you explain the speed improvements between releases, what if those optimisations have some relevance to other Graphics hardware for example.

On Friday 23 April 2004 3:36 pm, MJ Ray wrote:

Maybe they just removed really boneheaded errors from their drivers? I think those nVidia kernel modules were part of the motivation for the kernel taint warnings, so I'd guess the early ones were not good.

I had problems building nVidia drivers last year; after a lot of head scratching it turned out to be a handful of typos in their makefile. Jen