Re: [Flightgear-users] NVIDIA driver for linux
On Wed, 5 Feb 2003 08:48:53 -0600 Curtis L. Olson [EMAIL PROTECTED] wrote: Richard Keech writes: support for 3D capabilities of NVIDIA cards is not open source. you will need binary-only kernel modules (from nvidia) to make this work. many report good success with these drivers, however binary-only drivers should be strenuously avoided. The use of these drivers has been implicated in many stabiity problems with Linux. I would view the above quoted words as a political statement. :-) First, who cares if it's binary. Yes, RS and his group will squawk and yes, I like and use Opensource wherever I can but in some cases such as this it really doesn't matter - the vendor does supply drivers that work and if he wants to protect the portion that is proprietary that's fine with me as it's his. Second, I find that many people do not read the README that Nvidia supplies and hence have problems that can be fixed. The Nvidia readme is a comprehensive, very well done document that lists many of the problems (including the AMD one), what the fixes or workarounds are and contains a lot of detail on how things work. Nvidia is obviously supporting Linux seriously as this document is more than superficial. I have a GF3 TI500 on my Linux system and I read the entire document before I installed. I was prepared to setup my XF86Config, I knew what I did or did not have to have in the kernel and it worked well. I would never discourage an open exchange of ideas here, but I disagree with much of the above. It is true that nvidia's linux driver support is not open source. But ATI's has also done their most recent driver release as binary only. This means if you ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users
Re: [Flightgear-users] NVIDIA driver for linux
On Wednesday 05 February 2003 10:26 am, brett holcomb wrote: On Wed, 5 Feb 2003 08:48:53 -0600 Curtis L. Olson [EMAIL PROTECTED] wrote: Richard Keech writes: support for 3D capabilities of NVIDIA cards is not open source. you will need binary-only kernel modules (from nvidia) to make this work. many report good success with these drivers, however binary-only drivers should be strenuously avoided. The use of these drivers has been implicated in many stabiity problems with Linux. I would view the above quoted words as a political statement. :-) First, who cares if it's binary. Yes, RS and his group will squawk and yes, I like and use Opensource wherever I can but in some cases such as this it really doesn't matter - the vendor does supply drivers that work and if he wants to protect the portion that is proprietary that's fine with me as it's his. That's a point that bears looking at again. The Nvidia drivers work. If we look at RMS history, it was a driver that -sucked- that got him started. HP wouldn't accept a patch for a printer driver IIRMFC. Nvidia' unified driver means as long as they're making chips your Nvidia card is maintained. I'll take bets that Nvidia will release the code base when they do obsolete the architecture. Second, I find that many people do not read the README that Nvidia supplies and hence have problems that can be fixed. The Nvidia readme is a comprehensive, very well done document that lists many of the problems (including the AMD one), what the fixes or workarounds are and contains a lot of detail on how things work. Nvidia is obviously supporting Linux seriously as this document is more than superficial. I have a GF3 TI500 on my Linux system and I read the entire document before I installed. I was prepared to setup my XF86Config, I knew what I did or did not have to have in the kernel and it worked well. I would never discourage an open exchange of ideas here, but I disagree with much of the above. It is true that nvidia's linux driver support is not open source. But ATI's has also done their most recent driver release as binary only. This means if you ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users
Re: [Flightgear-users] NVIDIA driver for linux
I am still mystified about the NVidia/AMD problem if it is such. I have read and done everything mentioned on the SuSE site. On SuSE 8.1 with the SuSE kernels, including the recent SuSE internal 2.4.21, there are no problems. On the 2.4.20 kernel from ftp.kernel.org, I get everything up and working, but on every VT switch or shutdown, the machine imitates a clam requiring a hardware reset. If I get the time, I must look around and see what and where the differences are.I'm using the latest NVIDIA_*-1.0-4191.src.rpm. Regards Sid. On Wed, 2003-02-05 at 14:48, Curtis L. Olson wrote: Richard Keech writes: support for 3D capabilities of NVIDIA cards is not open source. you will need binary-only kernel modules (from nvidia) to make this work. many report good success with these drivers, however binary-only drivers should be strenuously avoided. The use of these drivers has been implicated in many stabiity problems with Linux. I would view the above quoted words as a political statement. :-) I would never discourage an open exchange of ideas here, but I disagree with much of the above. It is true that nvidia's linux driver support is not open source. But ATI's has also done their most recent driver release as binary only. This means if you want to run an ATI or nVidia card you are going to have to live with binary only drivers on your machine. If you choose anything else, you take a *big* step down in quality and/or performance. In terms of stability, the main problem I've seen is with older motherboards not doing reliable agp 4x or 2x or agp at all. There is also a hardware bug that went unfixed for a *long* time that causes agp memory corruption when you mix AMD CPU's with AGP (supposedly fixed in recent linux kernels and recent AMD cpu's.) There's an easy work around in linux if you find yourself with this problem. But all of these major stability issues were beyond the scope of the drivers. The AMD/AGP one was really tricky so as people were floundering around for a year or two trying to figure it out, a lot of theories were tossed about, and there was a lot of confusion and hair pulling. I'm sure the binary only nvidia drivers took their fair share of the blame. Yes, in a perfect world, all source would be open, we would all be working hard to make the world a better place, we'd all share the fruits of that hard labor freely and equally with those that are a little less fortunate, we'd all be dancing naked through the daisies with no worries and no fears. But that, unfortunately, is not the world we live in. People have to support their families and make money, and there are unfortunately too many people out there that don't want to work at all, but they still want to share the fruits of everyone else's labor, and live an easy life. The people at nVidia want to protect their efforts and continue to build and support their company and support their employees. We may not always agree with everything they do or the way they do it, but they certainly have every right to release binary only drivers if that's the way they want to play it. I personally don't have a problem with that. Personally, I think that if you want the best 3d performance and best 3d rendering quality on Linux, look at nvidia first (and I'm told that the latest ATI hardware and binary only drivers are pretty good too, but I haven't tried them myself.) Best regards, Curt. -- Curtis Olson IVLab / HumanFIRST Program FlightGear Project Twin Cities[EMAIL PROTECTED] [EMAIL PROTECTED] Minnesota http://www.menet.umn.edu/~curt http://www.flightgear.org ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users -- Sid Boyce ... hamradio G3VBV ... Cessna/Warrior Pilot Linux only shop ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users
Re: [Flightgear-users] NVIDIA driver for linux
On 5 Feb 2003, Sid Boyce wrote: I am still mystified about the NVidia/AMD problem if it is such. I have read and done everything mentioned on the SuSE site. On SuSE 8.1 with the SuSE kernels, including the recent SuSE internal 2.4.21, there are no problems. On the 2.4.20 kernel from ftp.kernel.org, I get everything up and working, but on every VT switch or shutdown, the machine imitates a clam requiring a hardware reset. There are certain functions on AGP graphics cards which, according to the AGP specs *require* cache flushes after they've been used. If this is not done then operation after that is undefined. Part of the reason that nvidia achieve such high performance is cutting corners on things like this. It's just possible that the effect of this corner cutting is felt more on AMD hardware than on intel. -- Jon Stockill [EMAIL PROTECTED] ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users
Re: Re: [Flightgear-users] NVIDIA driver for linux
Sid, I think if you go back to earlier kernels, 2.4.10ish would certainly do it, you'd have a better chance of experiencing the problem. Jon, the AMD problem was a known bug that AMD MS put out fixes for. ---Original Message--- From: Jon Stockill [EMAIL PROTECTED] Sent: 02/05/03 11:21 AM To: [EMAIL PROTECTED] [EMAIL PROTECTED] Subject: Re: [Flightgear-users] NVIDIA driver for linux On 5 Feb 2003, Sid Boyce wrote: I am still mystified about the NVidia/AMD problem if it is such. I have read and done everything mentioned on the SuSE site. On SuSE 8.1 with the SuSE kernels, including the recent SuSE internal 2.4.21, there are no problems. On the 2.4.20 kernel from ftp.kernel.org, I get everything up and working, but on every VT switch or shutdown, the machine imitates a clam requiring a hardware reset. There are certain functions on AGP graphics cards which, according to the AGP specs *require* cache flushes after they've been used. If this is not done then operation after that is undefined. Part of the reason that nvidia achieve such high performance is cutting corners on things like this. It's just possible that the effect of this corner cutting is felt more on AMD hardware than on intel. -- Jon Stockill [EMAIL PROTECTED] ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users
Re: [Flightgear-users] NVIDIA driver for linux
On Wed, 05 Feb 2003 13:43:54 -0500 John Check [EMAIL PROTECTED] wrote: On Wednesday 05 February 2003 10:26 am, brett holcomb wrote: On Wed, 5 Feb 2003 08:48:53 -0600 First, who cares if it's binary. Yes, RS and his group will squawk and yes, I like and use Opensource wherever I can but in some cases such as this it really doesn't matter - the vendor does supply drivers that work and if he wants to protect the portion that is proprietary that's fine with me as it's his. That's a point that bears looking at again. The Nvidia drivers work. If we look at RMS history, it was a driver that -sucked- that got him started. HP wouldn't accept a patch for a printer driver IIRMFC. Nvidia' unified driver means as long as they're making chips your Nvidia card is maintained. I'll take bets that Nvidia will release the code base when they do obsolete the architecture. Maybe I should have said - many feel that only opensource is acceptable and all else is TAINTED. In an ideal world it would be nice to have all open source but that isn't here yet. I agree - I've used Nvidia based cards for a long time and love the unified architecture. ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users
Re: Re: [Flightgear-users] NVIDIA driver for linux
Tony Peden writes: Sid, I think if you go back to earlier kernels, 2.4.10ish would certainly do it, you'd have a better chance of experiencing the problem. Jon, the AMD problem was a known bug that AMD MS put out fixes for. And the problem on the Linux side was that the wording of the bug report/notification was very obscure (maybe due to embarrassment over the stupidity of the bug?) Because of the obscure language, the Linux people completely missed it and didn't realize what they were talking about. So the bug lived for a long time unaddressed in the linux world. Meanwhile people were blaming the problem on buggy agp hardware, lousy binary graphics drivers, sun spots, bad application code, bad RAM, bad graphics hardware, buggy XFree86 code, and just about anything else you could imagine. Curt. -- Curtis Olson IVLab / HumanFIRST Program FlightGear Project Twin Cities[EMAIL PROTECTED] [EMAIL PROTECTED] Minnesota http://www.menet.umn.edu/~curt http://www.flightgear.org ___ Flightgear-users mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-users