vBulletin Search Engine Optimization
| |||||||
| Register | FAQ | Members List | Calendar | Search | Today's Posts | Mark Forums Read |
| ||||
| So get this. Each time you upgrade your kernel, you get to struggle upgrading your Nvidia video drivers once again too. Isn't that nice. And they have a really difficult time installing with more than one kernel on the machine. I'm fed up. I just expect the video card to WORK. It is most fundamental. Can anyone recommend a plain vanilla video card (no TVs, etc.) that just -works- with Redhat? AGP please. Ideally, when you install Redhat, it should just install without any thought or action on a human's part. No drivers to download. NO multiple 55 page manuals to read. No 4 dozen RPM options to read about and consider. NO posting to newsgroups to make. NO emails to tech support who neither understand or answer the question.... I just want it to work. First time. Every time. You have to hand it to Mr. Bill, it's very rigourous on Windows. When you want to install a video driver, you just install the driver. Total time, about 10 minutes. And what works for one, works for almost all. Why must Nvidia's drivers tie into the base operating system kernel? Why are they not self contained, separate from the kernel? Thanks |
| |||
| In comp.os.linux.hardware linuxquestion@yahoo.com wrote: > So get this. Each time you upgrade your kernel, > you get to struggle upgrading your Nvidia video > drivers once again too. Isn't that nice. No. Every time you build a new kernel, you cd into the directory: [wherever you unpacked the nvidia driver]/NVIDIA-(version)/usr/src/nv/ Read the README in that directory, then $ make a new interface module for nVidia, nvidia.o, then copy it across to the relevant /lib/modules/(kernel version)/kernel/drivers/video/ directory. > And they have a really difficult time installing > with more than one kernel on the machine. If the kernel versions are different, ie you have a 2.4.21, a 2.4.22 then you build a 2.4.24 you'll have an nvidia.o in each place, and each will suit the relevant kernel. If you are playing with multiple kernels of the same version, you'll have the potential for module conflicts in general. That's not confined to the nvidia module. Some modules compile differently depening upon kernel configuration options. Using some with a different kernel can cause some interesting effects... > Why must Nvidia's drivers tie into the base operating > system kernel? Why are they not self contained, separate > from the kernel? Because they perform as well as they do _because_ of the way they are tied into the kernel. -- Athol <http://cust.idl.com.au/athol> Linux Registered User # 254000 I'm a Libran Engineer. I don't argue, I discuss. |
| |||
| athol <me@privacy.net> writes: >> Why must Nvidia's drivers tie into the base operating >> system kernel? Why are they not self contained, separate >> from the kernel? > > Because they perform as well as they do _because_ of the way they > are tied into the kernel. It's not about performance. The kernel drivers are there for the same reason all the 3D drivers have a kernel part. The function of the kernel module is to manage DMA transfers to the video card. The reason this needs to be done by the kernel is that the physical address of the source needs to be known, and there is no way to find this information from userspace. If the memory mapping was accessible from userspace it would perform equally well there. -- Måns Rullgård mru@kth.se |
| |||
| linuxquestion@yahoo.com wrote: > Can anyone recommend a plain vanilla video card > (no TVs, etc.) that just -works- with Redhat? > AGP please. I have a Radeon 9000 card, and it "just works" without any extra driver installation. > Why must Nvidia's drivers tie into the base operating > system kernel? Why are they not self contained, separate > from the kernel? The driver must access the card hardware directly, that's why there must be a component that runs as a kernel module. nVidia refuses to tell Linux developers how their cards work, that's why the driver can't be included in the standard kernel like the Radeon DRM module. -- Markku Kolkka markku.kolkka@iki.fi |
| |||
| On Sat, 10 Jan 2004 22:30:59 -0800, linuxquestion wrote: > So get this. Each time you upgrade your kernel, > you get to struggle upgrading your Nvidia video > drivers once again too. Isn't that nice. > > And they have a really difficult time installing > with more than one kernel on the machine. > > I'm fed up. I just expect the video card to WORK. > It is most fundamental. All the Nvidia cards I have ever owned have worked well with Red Hat out of the box just using the default nv driver that comes with XFree86. The ONLY reason you may need the Nvidia drivers is if you also need to use accelerated 3D. If you don't need that you don't need the Nvidia drivers from Nvidia. IF you DO need accelerated 3D then you DO need the Nvidia drivers. When the kernel is upgraded the driver module has to be rebuilt to match the kernel. This is as simple as running (as root): sh NVIDIA-driver-you-downloaded and answering a few yes/no questions then switching back to the GUI runlevel: telinit 5 Whole process takes about 30 seconds. To make this process even simpler I just keep the the Nvidia driver installer I downloaded in my /root directory. When I install a new kernel and reboot I tell grub to boot to runlevel 3 (console mode). Just hit "a" when the grub screen comes up and then "3" and Enter. It is really REALLY easy. Incidentally Nvidia uses the same driver for Linux as it does for Windows. The driver works well out of the box... if you want to tweak it for some reason, yes, you will have to read the excellent docs that Nvidia supplies in order to find out what the tweaks are. Again if you don't need accelerated 3D, which you only need for some games and engineering/scientific apps... then just use the stock nv diver and skip all the (minor) hassle. For me, Nvidia cards are some of the best out there. I have had very good use out of them. -DU-...etc... |
| |||
| Well, I finally got it to work. rpm -qa | grep -i kernel | sort kernel-2.4.20-18.10.1 kernel-2.4.20-18.10.1 .... kernel-source-2.4.20-18.10.1 .... kernel-2.4.9-e.3 kernel-source-2.4.9-e.3 kernel-summit-2.4.9-e.3 kernel-utils-2.4-6 Hmm. Two kernel-2.4.20-18.10.1 kernels! Remove them both. rpm -e --allmatches --nodeps kernel-2.4.20-18.10.1 Reinstall. Interestingly enough, by installing the .src file, it doesn't install the kernel. rpm -ivh --nodeps kernel-2.4.20-18.10.1.src.rpm 1:kernel ########################################### [100%] rpm -qa | grep -i kernel | sort - not there Install the binary, and source: rpm -ivh --nodeps kernel-2.4.20-18.10.1.i686.rpm Preparing... ########################################### [100%] 1:kernel ########################################### [100%] rpm -ivh --nodeps kernel-source-2.4.20-18.10.1.i386.rpm Preparing... ########################################### [100%] 1:kernel-source ########################################### [100%] -------------- I managed to find the old nvidia package: NVIDIA-Linux-x86-1.0-4496-pkg2.run I reboot to the new kernel, and run install.sh sh NVIDIA-Linux-x86-1.0-4496-pkg2.run - worked! - about time. It should have taken 10 to 20 minutes, but I finally finished about after 2 days. Somewhere between 10 to 20 hours of struggle. Directly related to Nvidia drivers. I will never buy nvidia cards intentionally again. --------------- linuxquestion@yahoo.com wrote in message news:<672ceaed.0401102230.6a383fd8@posting.google. com>... > So get this. Each time you upgrade your kernel, > you get to struggle upgrading your Nvidia video > drivers once again too. Isn't that nice. > > And they have a really difficult time installing > with more than one kernel on the machine. > > I'm fed up. I just expect the video card to WORK. > It is most fundamental. > > > Can anyone recommend a plain vanilla video card > (no TVs, etc.) that just -works- with Redhat? > AGP please. > > > Ideally, when you install Redhat, it should just > install without any thought or action on a > human's part. No drivers to download. NO multiple > 55 page manuals to read. No 4 dozen RPM options to > read about and consider. NO posting to newsgroups > to make. NO emails to tech support who neither > understand or answer the question.... I just > want it to work. First time. Every time. > > > You have to hand it to Mr. Bill, it's very rigourous > on Windows. When you want to install a video driver, > you just install the driver. Total time, about 10 minutes. > And what works for one, works for almost all. > > Why must Nvidia's drivers tie into the base operating > system kernel? Why are they not self contained, separate > from the kernel? > > > Thanks |
| |||
| > All the Nvidia cards I have ever owned have worked well with Red Hat out > of the box just using the default nv driver that comes with XFree86. The > ONLY reason you may need the Nvidia drivers is if you also need to use > accelerated 3D. If you don't need that you don't need the Nvidia drivers > from Nvidia. > > IF you DO need accelerated 3D then you DO need the Nvidia drivers. Well, all I'm trying to do is X Windows. I don't think that is 3D. But if we define it as so, it is not so simple. > When the kernel is upgraded the driver module has to be rebuilt to match > the kernel. This is as simple as running (as root): > > sh NVIDIA-driver-you-downloaded > > and answering a few yes/no questions then switching back to the GUI > runlevel: > telinit 5 > > Whole process takes about 30 seconds. How many times did you do it before you got it to this time? It's taken me between 10 to 20 hours. In retrospect, I had a problem of two kernels. But why didn't I get a better error messages? From either linux, or from the nvidia scripts? I'm glad it's "just that simple" for you. Please continue to post the short and experienced instructions for the rest of us. |
| |||
| On 10 Jan 2004 22:30:59 -0800, linuxquestion@yahoo.com Gave us: >Why must Nvidia's drivers tie into the base operating >system kernel? Why are they not self contained, separate >from the kernel? > Nvidia has single file drivers now that are one button press and play affairs. Uninstalls any existing, installs new. No Problemo. By far the best as well. |
| |||
| On 11 Jan 2004 18:02:40 -0800, linuxquestion@yahoo.com Gave us: >> All the Nvidia cards I have ever owned have worked well with Red Hat out >> of the box just using the default nv driver that comes with XFree86. The >> ONLY reason you may need the Nvidia drivers is if you also need to use >> accelerated 3D. If you don't need that you don't need the Nvidia drivers >> from Nvidia. >> >> IF you DO need accelerated 3D then you DO need the Nvidia drivers. > >Well, all I'm trying to do is X Windows. I don't think that is 3D. >But if we define it as so, it is not so simple. > > >> When the kernel is upgraded the driver module has to be rebuilt to match >> the kernel. This is as simple as running (as root): >> >> sh NVIDIA-driver-you-downloaded >> >> and answering a few yes/no questions then switching back to the GUI >> runlevel: >> telinit 5 >> >> Whole process takes about 30 seconds. > >How many times did you do it before you got it to this time? > >It's taken me between 10 to 20 hours. In retrospect, I had a problem >of two kernels. But why didn't I get a better error messages? >From either linux, or from the nvidia scripts? > >I'm glad it's "just that simple" for you. Please continue to >post the short and experienced instructions for the rest of us. I ALWAYS boot to command console, and run startx on my own. That way I can install drivers that xwindows may interfere with. Regardless, the new NVIDI drivers go in very easily. You may be thinking of the old manner they used to require for installation. Now, it is a single file. Works fine on my dual athlon. :] good luck. |
| ||||
| In comp.os.linux.hardware DarkMatter <DarkMatter@thebarattheendoftheuniverse.org> wrote: > I ALWAYS boot to command console, and run startx on my own. That > way I can install drivers that xwindows may interfere with. On my systems that _have_ X, I default to it (runlevel 4 in slackware). If I want to play with drivers, I simply feed the alternative runlevel to LILO while booting. If I'm testing a new kernel that I know won't work with the nVidia driver, I put an append line in lilo.conf to force a non-X runlevel for that kernel. -- Athol <http://cust.idl.com.au/athol> Linux Registered User # 254000 I'm a Libran Engineer. I don't argue, I discuss. |