I want free drivers

[plasma]

(free as in freedom)

If ATI can do it why can't nvidia?

[bwkaz]

ATI has done it? Could've fooled me...

They've released info on their boards to the XFree86 driver developers, but considering how short the ATI product cycle is vs. how long it takes X to release a new version with new drivers, it really doesn't surprise me that not every ATI card works. In fact, the R350 core cards don't work at all (IIRC), and the R300's only work in 2D. With XFree86 4.2.0, up to the 7500 worked, and nothing else (in 3D mode, that is). Even though 8500's had been out for quite some time before X's 4.2.0 release.

ATI also provides closed-source drivers that they've written, that work with all of their cards (well... the 8500 and above). At least, I assume they're closed-source because you can only download RPMs. Yeah, that's swift, leave those of us that don't use RPM distros out in the cold... :rolleyes: I think I can say that for me, nVidia's driver setup is better, right now.

However, free drivers would definitely be better than what we have. I, for one, wouldn't have any of my issues with 4191 (or 2960, all those many months ago). For those of you that say "they can't! they signed NDAs with other companies!", there are ways around that. For example, put the code you can't open up into the video BIOS, and then publish the interface to it. Do that, and allow us to see your current drivers (current at the time -- obviously it'd take a while to change how all your cards work by that much).

You get free development help (from people like me that like to think (:p) we know what we're doing), your stupider competition wastes their time on cloning your current boards instead of making their own product different/better, while you can put out even better stuff because you're the original developer of it, plus you'll get a lot of Linux users as loyal customers, rather than the current state of things, where they go with nVidia merely because it seems to work. I know I for one would jump on any manufacturer that came out with truly open drivers, and probably stay with them for life. As it is, yeah nVidia's drivers work (most of the time, see 4191 for ample counterexamples though), so I'm using your cards. That kind of "loyalty" can evaporate quite quickly. Especially if your next driver release is just as crappy as 4191 seems to have been (and especially if my __divdi3() bug isn't fixed yet).

(unfortunately, I haven't thought most of these arguments up -- see ESR's The Magic Cauldron appendix (http://catb.org/~esr/writings/magic-cauldron/magic-cauldron-17.html) on why closing drivers is a bad idea)

Edit: And oh, you'll probably want to make sure overclocking using your drivers isn't possible. Otherwise you'll get quite a few of the less-intelligent users rampantly overclocking, then complaining about the card burning itself -- definitely not a good situation to be in, even if they don't have a leg to stand on with the complaint.

[hohlraum]

I'll settle for closed source nvidia drivers that don't suck like these current 4XXX ones do. They were doing great in the 3XXX series and then they release these new pieces of s**t. I really don't care if the various components on the MB have closed source drivers either. However, having close sourced MB chipset drivers is out of the question if that is infact the case.

[plasma]

Originally posted by bwkaz
ATI has done it? Could've fooled me...

They've released info on their boards to the XFree86 driver developers, but considering how short the ATI product cycle is vs. how long it takes X to release a new version with new drivers, it really doesn't surprise me that not every ATI card works. In fact, the R350 core cards don't work at all (IIRC), and the R300's only work in 2D. With XFree86 4.2.0, up to the 7500 worked, and nothing else (in 3D mode, that is). Even though 8500's had been out for quite some time before X's 4.2.0 release.

but it's a start.

ATI also provides closed-source drivers that they've written, that work with all of their cards (well... the 8500 and above). At least, I assume they're closed-source because you can only download RPMs. Yeah, that's swift, leave those of us that don't use RPM distros out in the cold... :rolleyes: I think I can say that for me, nVidia's driver setup is better, right now.

However, free drivers would definitely be better than what we have. I, for one, wouldn't have any of my issues with 4191 (or 2960, all those many months ago).

I still have issues. The most annoying is when the kernel driver gets fried - X keeps locking up (even after being restarted) and the virtual terminals are screwed meaning I have to reboot if I want to use a monitor :(. Thanks nvidia for "suporting" linux.

For those of you that say "they can't! they signed NDAs with other companies!", there are ways around that. For example, put the code you can't open up into the video BIOS, and then publish the interface to it. Do that, and allow us to see your current drivers (current at the time -- obviously it'd take a while to change how all your cards work by that much).

You get free development help (from people like me that like to think (:p) we know what we're doing), your stupider competition wastes their time on cloning your current boards instead of making their own product different/better, while you can put out even better stuff because you're the original developer of it, plus you'll get a lot of Linux users as loyal customers, rather than the current state of things, where they go with nVidia merely because it seems to work. I know I for one would jump on any manufacturer that came out with truly open drivers, and probably stay with them for life.

my 3dfx will continue to be suported by X. :D (I could use it with a mac if I wanted too..

As it is, yeah nVidia's drivers work (most of the time, see 4191 for ample counterexamples though), so I'm using your cards. That kind of "loyalty" can evaporate quite quickly. Especially if your next driver release is just as crappy as 4191 seems to have been (and especially if my __divdi3() bug isn't fixed yet).

unless this gets sorted I will never buy an nvidia anything again.

(unfortunately, I haven't thought most of these arguments up -- see ESR's The Magic Cauldron appendix (http://catb.org/~esr/writings/magic-cauldron/magic-cauldron-17.html) on why closing drivers is a bad idea)

doesn't make them invalid.. ;)

Edit: And oh, you'll probably want to make sure overclocking using your drivers isn't possible. Otherwise you'll get quite a few of the less-intelligent users rampantly overclocking, then complaining about the card burning itself -- definitely not a good situation to be in, even if they don't have a leg to stand on with the complaint.
why? the current ones are overclockable. (no I'm havn't overclocked my nvidia).

[cityhunter]

if nvidia don't want to realease source for video i can understand this.... but for nforce chipset this must be, for better compatibility, for a lot of points, ide performance (cpu use) is really strange.....
we allready use i810 audio driver for nforce apu...
think this should be the best

[cityhunter]

80 view and only 9 votes :eek: !?!,
vote even if it's to say forget about me

[captnmark]

Originally posted by cityhunter
if nvidia don't want to realease source for video i can understand this.... but for nforce chipset this must be, for better compatibility, for a lot of points, ide performance (cpu use) is really strange.....
we allready use i810 audio driver for nforce apu...
think this should be the best

Agreed. There is zero chance that I will ever buy a mobo that needs closed source linux drivers. Especially if it forces me to use Nvidia gfx card.

[cityhunter]

1) nforce doesn't force you to use nvidia card.... althought I don't know if it's possible to use two opengl board at the same time.... anyway, integrated geforce4mx is on pci 2:0:0, classical vga is on pci 1:0:0.... (it's triker since you have to change XF86Config)
but nothing seems to prevent you to use two Nvidia graphic card on your nforce board (up to 3 screens :eek: )
the only problem (if there is problem) is to include other manufacturer accelerated opengl board.......

open source is a way to garantie linux user that even if support will be no longer (because of new hardware) bugs can still be solved....

furthermore be able to compile the drivers garantie the best performances..... and even more since on compilation we can guess the type of hardware and then remove unecesary lines.....

[nutball]

Originally posted by cityhunter
80 view and only 9 votes :eek: !?!,
vote even if it's to say forget about me

That's because the poll asks the question:

"When did you stop beating your wife?"

A) Recently
B) Not yet

I'm not going to vote because none of the answers suit me!

Where's the "I don't care if the drivers are open source or not, just so long as they don't suck" option?

I do wish the open source community would get its collective head around the idea that open source is not a panacea.

Linux is a highly unstable platform, and this very often makes it very frustrating to use and to maintain. Sure it's great for hackers who love spending all day installing todays version of everything, but for those of us who want a working e-mail client, a working web browser, and a sensible GUI toolkit we can develop against it's a nightmare.

Do you really think the open source community could do a better job writing drivers for hardware they no documentation for? Do you honestly expect NVIDIA to start publishing commercial confidential information about the architecture of their latest chips which could give their competitors an advantage?

If NVIDIA have ownership of the drivers and the drivers suck I can tell them they suck and I have the right to expect them to be fixed. If the drivers are open source and there's a problem and I say so, then I get some smart arse saying "well why don't you hack the driver source and fix the problem and send us a patch" it doesn't really get me anywhere. I don't care about hacking kernel or driver source, I got all that hack nonsense out of my system twenty years ago. I just want a stable platform which works correctly so I can do my job and develop end-user applications.

(Which isn't what I have at the moment: NVIDIA your current Linux drivers suck and need fixing).

[cityhunter]

Originally posted by nutball
[B]That's because the poll asks the question:

"When did you stop beating your wife?"

A) Recently
B) Not yet

I'm not going to vote because none of the answers suit me!

Where's the "I don't care if the drivers are open source or not, just so long as they don't suck" option? the third one : forget about me : you don't care, then you're not pro or cons open source


I do wish the open source community would get its collective head around the idea that open source is not a panacea.

go say that to people that still use 3dfx boards.... they can go suck a cow if they want to find drivers for 2k or xp.... or recent ones for 98..... the result will be the same....
if one day NVidia begins to change the "unified version of the drivers" and decide to support only new boards.... if a bug is found you'll be crying that you must change your MB or CG only for a minor bug..... 3DFx user are still alive on linux thanks to open source.


Linux is a highly unstable platform, and this very often makes it very frustrating to use and to maintain. Sure it's great for hackers who love spending all day installing todays version of everything, but for those of us who want a working e-mail client, a working web browser, and a sensible GUI toolkit we can develop against it's a nightmare.
think that's too long that you ever touched a linux pc..... ask to my admin here how much time his gateway router (the one on which he is working all day long) has been running... the answer will be "I've never stoped it since about 4 years"

Do you really think the open source community could do a better job writing drivers for hardware they no documentation for? Do you honestly expect NVIDIA to start publishing commercial confidential information about the architecture of their latest chips which could give their competitors an advantage?

the architecture docs are already available, the docs don't give real advantage, a chip to the point of view of a cpu is a set of register, setting some register in some way give some results and so on..... I don't think opensource communeauty will do better but opensource communeauty isn't driven by economical law.... try to make nforce1 lan working...... you'd better suck a cow at least you'll make someone happy....

[nutball]

Originally posted by cityhunter
think that's too long that you ever touched a linux pc..... ask to my admin here how much time his gateway router (the one on which he is working all day long) has been running... the answer will be "I've never stoped it since about 4 years"
[/B]

When I said unstable I didn't mean that it crashes, I mean that the underlying software is changing very rapidly. This means that if I want to upgrade my e-mail client, or web browser, chances are I'll have to update a whole load of other stuff, and chances are that those upgrades will be incompatible with something else on my system.

[cityhunter]

you should use sources based distrib like sorcerer
if you have an unlimited connection, doing a "augur update" will rebuild the entire new stuff for you :eek:
I really love it ;)

then you'll always have a uptodate box :)

[plasma]

Originally posted by nutball
When I said unstable I didn't mean that it crashes, I mean that the underlying software is changing very rapidly. This means that if I want to upgrade my e-mail client, or web browser, chances are I'll have to update a whole load of other stuff, and chances are that those upgrades will be incompatible with something else on my system.
It's called debian (http://debian.org) ;) not as easy to setup, but very nice to maintain.


Do you honestly expect NVIDIA to start publishing commercial confidential information about the architecture of their latest chips which could give their competitors an advantage?

If it's usefull for writing good drivers, Yes. No that's not comercial deathwish.


If NVIDIA have ownership of the drivers and the drivers suck I can tell them they suck and I have the right to expect them to be fixed.

expecting isn't a guarantee.


If the drivers are open source and there's a problem and I say so, then I get some smart arse saying "well why don't you hack the driver source and fix the problem and send us a patch" it doesn't really get me anywhere. I don't care about hacking kernel or driver source, I got all that hack nonsense out of my system twenty years ago. I just want a stable platform which works correctly so I can do my job and develop end-user applications.

If you're having a problem someone else probably has too - nvidia should fix it if asked. If they don't you're not screwed (unlike the current situation :banghead: ). Because the code is free it's likely to be more stable (more people will check it, debug it and it's likely to be tidier).

(Which isn't what I have at the moment: NVIDIA your current Linux drivers suck and need fixing).

Agreed.
IMHO freeing them would be an empowering fix :)

[zoon64]

I they dont want to release the drivers maybe they should devellop there own Linux Distro made only for Nvidia products like motherboars with Nforce1/2 chipsets and there graphics range.


Can you imagine no overhead pure NVIDIA power to play with.

What dreams may come. HEHE

[jackangel]

All of you people are encouraged to ask nvidia to release all their linux/freebsd drivers as open source. I've created an online petition where you can sign it, later I'll send it to NVIDIA by snail-mail. Please, consider it seriously, it's very important issue. If we can get NVIDIA release their drivers open-source, kernel hackers will quickly fix many bugs and boost their performance, also possibly simplifying installation process or converting the driver architecture to standard DRI.

Here's the petition:
http://www.petitiononline.com/hwopennv/petition.html

[nutball]

Originally posted by jackangel
If we can get NVIDIA release their drivers open-source, kernel hackers will quickly fix many bugs and boost their performance


Maybe.

[hohlraum]

That must be why the opensource ATI drivers are so wonderful. *cough*

[jackangel]

ATI drivers may not be very good now, but the data is present, everyone with programming skills may develop it - you'll see, they'll be great in no time

[verysoft]

The most regretable decision I ever made was to believe the hype and buy an ATI radeon VE.
While everything works fine under win, you are completely stuffed under linux. ATI does not even have the courtesy to distribute their "free" drivers.
Instead you find links to some non existend 'precisioninsight.com' which folded up some years ago. Or gatos! absolutely hopeless, distributing the radeon 1.1 driver, that does absolutely nothing foer any distro. One poor lonely soul is hacking in some godforsaken CVS-tree.
I tried to get my radeon going for almost a year, downloading hundreds of Mb's, CVS's broke my debian distro, tried RH, tried mandrake -> anything but 3d!
When last XFree-4.3 was released I had hope again. recompiled everything including latest kernel module for the evasive 1.2/1.3 module! It hung up that badly that I could not find my SCSI-drive after hard reboot!
Beeing completely fed up I bought a Geforce4/Leadtek for A$98 on the weekend market. Sunday afternoon I was braindead enough to install a wrong driver, e-mailed nvidia and had a response in under 2hrs with a link to the correct software - installed and ran it.

Conclusion: anybody suggestiing even a keyhanger made by ATI for use with a linux system to me risks severe personal injury!
Recommending any ATI products under linux only shows that you have not the faintest idea, and just blab some BS that you picked up elsewhere!

However! I still got my ATI, if you like it so much, for 50 bucks its yours buddy!

[plasma]

Interesting.
According to the dri site (http://dri.sourceforge.net/dri_status.phtml) it's supported. It does mention problems with early VIA chipsets and a hack to get pci cards to work (http://dri.sourceforge.net/faq/faq_display.phtml?id=42).... (also use the "radeon" driver, not the "ati" one (http://dri.sourceforge.net/faq/faq_display.phtml?id=22)) Also I read somewhere that S3 Texture Compression is patented and consequently not implemented in these drivers.

Of course Nvidia 3d acceleration will never work out of the box with debian with the current proprietry drivers... (see the Debian Social Contract (http://www.debian.org/social_contract)). This is obviosly nvidia's problem not debian's.

ATI has made the necessary hardware and programming information available to Linux developers for the development of hardware 3D acceleration. (http://mirror.ati.com/support/faq/linux.html#3d)
So perhaps it's not ideal, but its a start and obviously something nvidia could follow (and hopefully better. eg by actually paying the driver developers, and having the drivers available nearer launch (without comprimising quality)). I don't think it's much to ask of nvidia to match ati's support of independent free driver developers.

[verysoft]

sorry if I might have sound a tad rude, but this is what exactly what I mean: you read the site, you read the reviews, and you hear another successstory. Unfortunately you be in for some nasty surprise as soon as you have the real thing.
The kernel *radeon* driver continously clash with the xfree-drivers.
The kernel only compiles up to 4.1 (check out your config) - and this is just the start of your odyssee. When I said I tried everything believe me. NVIDIA hiccups like modprobes, headers not found or compiler mismatch didn' even irritate me.

But this was not **YOUR** point, and I am sorry to elaborate that much. You are right: it would be nice to have the nvidia as open source, preferable distributed with the kernel.
Just ATI is a really bad example.

Like not everybody you smoke a joint with turns out to be a nice person, there is a lot of hype and rubbish, and just because it runs under the linux-name doesn't mean its nice or good.
The work done on the ati kernel by the community s great, but this is such a complicated and big project that it could not be undertaken without massive support from the manufacturer. It is not just good enough to release some data and don't give a damn what happen to it, and than claim you are ->the<- open source supporter.
Just for the heck of it: try to download a kernel module radeon1.2 (7000 series). considering that I tried to get the card going for 6mnth means I did not get much use out of it eventhough it is a year old by now :(

[driehuis]

I'm glad many people in this forum acknowledge that writing high performance drivers is _hard_!

The issues most folks have with the nvidia drivers require deep magic to debug. Now, I'm a staunch supporter of open source, but my practical concern with the binary drivers for FreeBSD is not that I can't debug X. I tried debugging X drivers back when it was "simply" a matter of getting the ramdac settings right, and I gave up on that. I don't even want to think about debugging AGP or 3D cores! I'm more impacted by the fact that I can't just install XFree and be done with it. I just upgraded my FreeBSD installation from 4.7 to 4.8, and consequently XFree from 4.2 to 4.3, and have been suffering hard hangs -- which means that I'm in for a fun session of twiddling knobs and see which makes it stable. Compounded by the fact that if I start X without a window manager, and then start X normally, the problem doesn't show (which of course I forget every so often, and then I get the joy of running a manual fsck to recover the file system -- and of course, the crash is so quick and so total that no log files survive the crash, so even if I had source I would be clueless where to look).

The free "nv" driver in XFree sucks to the point that it breaks even xterm! You have to disable 2D acceleration to get readable text in an xterm -- which makes my 2GHz 1GB machine with a Ti4200 feel like a 486 with an dumb frame buffer when using X.

When it comes to high performance graphics, I sometimes long for win2k. If I can make a plea to nvidia, it's to ship the drivers for the most conservative mode that is (almost :-) guaranteed not to crash a system by default. Then again, I think that is their aim already.

I'd toss my Ti4200 in a jiffy if a decent 3D board with stable XFree drivers was available; screw the investment. But I haven't seen any vendor do this better. The stable boards are slow and the fast ones are flaky on FreeBSD. And I somehow doubt that a huge talent pool exists that knows how to tackle such low level stuff as AGP, with all the motherboard differences thrown into the mix. Of course, if more sources were available, many smaller nuisances would be tackled by the community real quickly; like the libGL.so link issue I mentioned in another thread, and the "out of the box" experience that improves when integration testing can be done prior to a release such as FreeBSD 4.8.

For now, I'm happy that nvidia concentrates on getting the drivers right (and from what I hear, I don't have to feel bad that the FreeBSD drivers are still 3203 based :-)