On 2/20/22 15:57, Felix Miata wrote:
DennisG composed on 2022-02-20 14:10 (UTC-0500):
mark neidorff composed on 2022-02-17 17:41 (UTC-0500):
If the problem is not nouveau, what should I run to get more diagnostic information? Needless to say, my goal is a stable, graphical KDE environment. It used to be with a Dell desktop that if a discrete graphics card was installed on a model that includes motherboard IGP or CPU IGP, that connecting a display to the IGP output would result in a BIOS message reporting that you are not allowed to do that, and must connect your display(s) to the discrete card's output(s) instead. I've not seen a Dell BIOS anywhere near as new as yours so as to know what to suggest you might need or want to do there. Motherboards for self-built PC's provide a lot more options re graphics setup. You might want to poke in yours for a suitable change to make.
Ideally you have two ways to go:
1-Remove the discrete card 2-configure use of the discrete card to the exclusion of the IGP
Another option would be to utilize the concept of "offloading", where the IGP and GPU share the job. That's more complicated and something I've never done. Whether anyone with such experience reads this list I have no idea, but I know there is at least one person routinely on the openSUSE forums who does, so you might want to do a search or inquire there. https://forums.opensuse.org/forumdisplay.php/668-Install-Boot-Login
For option 1, it should be simple. Remove the discrete card, and let automagic do its thing without interference.
NAICT, you have a 400 series Intel chipset PC, which supports Comet Lake (10th gen) and Rocket Lake (11th gen) Intel processors. At some point we may need to know the specific CPU model you have installed to be sure of how to proceed, as with Rocket Lake there could be an added problem if you ever connect more than one display at a time. All the bugs aren't out of Rocket Lake support yet.
For option 2, you may need to blacklist i915 and/or i965, and/or change a BIOS option, to prevent interference from the IGP in using either of the two ostensibly competent FOSS display drivers for GeForce. If the GeForce cannot be made to work properly with the FOSS drivers, a bug report would be in order, as it should just work automagically. Installing *nouveau* may be necessary to enable that, as well as purging any residue from any prior attempt to install proprietary NVidia drivers that you might have made.
To be complete, there's also the proprietary NVidia driver option, which I never use, so won't comment further about. If you want to try this again anyway (I don't advise this), the packages needed are as follows. Note that there are actually 2 nouveau drivers,
On 2/17/22 19:55, Felix Miata wrote: the kernel driver which starts at boot and an X driver (xf86-video-nouveau). xf86-video-nouveau provides an /optional/ /old technology/ display driver. For best results from FOSS on NVidia GPUs, it is not usually required.
Ah, that explains why I didn't find it required by anything, although it is installed by default. Does this mean that the current (kernel) nouveau has resolved the power management issues that it had, and so that can be ruled out for causing OP's racing issue? And is it no longer required to import the mok certificate?
Before using it, you must be sure that there is no modesetting in the grub2 boot command line (if there is, remove that Modesetting is the name of a driver. It is nomodeset, along with i915.modeset=0, amdgpu.modeset=0, radeon.modeset=0, and nouveau.modeset=0 that will disable KMS, which disables all optimal FOSS display drivers.
Yes, thx. Given that distros use different syntax here, I just used an abbreviated "modesetting" to keep in simple, lest there was something left over from previous attempts.
Finally, re the integrated graphics with your Rocket Lake chip. ...You also were probably looking at the xf86-video-intel package, which provides a DRI/3D driver for X (analogous to the X driver above for nouveau), not the same animal as the main Intel i915 kernel driver. (IOW, xf86* is actually in 15.3 although not shown by the search tool.) Also like with nouveau, both drivers are required. Only the i915 kernel module (driver) is required for Rocket Lake. No DDX driver is required for either Intel or NVidia GPUs: # rpm -qa | egrep 'xf86-v|firmware|intel' | sort intel-media-driver-20.3.0-1.41.x86_64 intel-vaapi-driver-2.4.1-1.62.x86_64 kernel-firmware-i915-20210208-2.4.noarch kernel-firmware-intel-20210208-2.4.noarch libdrm_intel1-2.4.104-1.12.x86_64 ucode-intel-20210525-7.1.x86_64 xf86-video-fbdev-0.5.0-4.31.x86_64 xf86-video-vesa-2.4.0-3.31.x86_64
Noted. I think I was thrown off by xf86 nvidia and intel packages being included in 15.3 and TW.
# lsmod | grep i9 i915 2625536 4 i2c_algo_bit 16384 1 i915 drm_kms_helper 262144 1 i915 cec 65536 2 drm_kms_helper,i915 drm 614400 4 drm_kms_helper,i915 video 53248 1 i915 # inxi -SCy System: Host: ab560 Kernel: 5.3.18-150300.59.49-default x86_64 bits: 64 Desktop: Trinity R14.0.11 Distro: openSUSE Leap 15.3 CPU: Info: 6-core model: 11th Gen Intel Core i5-11400 bits: 64 type: MT MCP cache: L2: 3 MiB Speed (MHz): avg: 801 min/max: 800/4400 cores: 1: 802 2: 801 3: 801 4: 803 5: 801 6: 801 7: 800 8: 801 9: 802 10: 800 11: 801 12: 800 # inxi -Gayz Graphics: Device-1: Intel RocketLake-S GT1 [UHD Graphics 730] vendor: ASUSTeK /driver: i915/ v: kernel bus-ID: 00:02.0 chip-ID: 8086:4c8b class-ID: 0300 /Display/: x11 server: X.Org 1.20.3 /driver/: /loaded/: modesetting/ unloaded: fbdev,vesa alternate: intel display-ID: :0 screens: 1 Screen-1: 0 s-res: 2560x1440 s-dpi: 120 s-size: 541x304mm (21.3x12.0") s-diag: 621mm (24.4") Monitor-1: DP-1 res: 2560x1440 hz: 60 dpi: 109 size: 598x336mm (23.5x13.2") diag: 686mm (27") OpenGL: renderer: /Mesa Intel Graphics/ (RKL GT1) v: 4.6 Mesa 20.2.4 direct render: Yes #
Note absence of "intel" with regard to any running driver. The newer FOSS technology modesetting DIX driver is the preferred FOSS driver for all GPUs it supports, except for newer AMD GPUs that are supported by the DDX display driver provided by xf86-video-amdgpu. This support includes NVidia GPUs.
https://forums.opensuse.org/showthread.php/562680-AMD-Intel-amp-NVidia-X-gra...
The nouveau DDX display driver is old technology, and reverse-engineered. The intel DDX display driver is also old technology. It hasn't had an official release in over 7 years. It's in maintenance mode, because for ancient Intel IGPs it's the only KMS-supporting display driver, and because in some configurations it may work better than the modesetting DIX.
Again, thx for the update. Clearly I need that primer. :)
That said, as I wrote two weeks ago but you did not reply to, due to Rocket Lake being newer (March 2021) than stock 15.3 (Feb 2021), there is a very strong possibility that you need updated Intel kernel firmware I didn't for my Rocket Lake. He doesn't on account of any Rocket Lake hardware. I suppose Dell could have slipped in something else that needs newer firmware.
I'll take your word. Although it seems curious why there is a firmware update given that Xe-LP is the latest architecture. Maybe due to processor differences across the Rocket Lake series (although I would expect the same for OP's i7 as your i5). In any event, thanks Felix for taking the time with the needed corrections/clarifications. Re the OP's problem: His nvidia card has had good driver support from nvidia for 3 years, but he claims it does not work. He has reported both good, and very bad, results with nouveau. One post showed that the i915 had been loaded, but it apparently didn't work, either. At this point, then? A newer kernel, as previously advised? Or go back to square one, test each of the alternatives again verifying steps, and take a look at the X log each time (as probably should have been suggested before)? Or, if nouveau and i915 are viable alternatives, maybe just another fresh install to test each??? OP has been to this movie 3 times over the last 2 months. --dg 15.3/Plasma