| BluesKaj | Hi all | 12:34 |
|---|---|---|
| sgmoore | arraybolt3: mmikowski: RikMills: I have seen a few of these https://www.reddit.com/r/Kubuntu/comments/1c915x0/heads_up_i_hit_a_serious_bug_while_testing/ Any suggestions on video drivers we should be seeding? | 14:42 |
| RikMills | I think we should look at getting the bugs fixed in the drivers we seed at this point, rather than trying to seed less better versions | 15:09 |
| IrcsomeBot | <sgmoore> Umm, how does one 'fix' bugs in proprietary drivers? | 15:10 |
| RikMills | This is something that might need a change of mind for 24.04.1 | 15:11 |
| RikMills | sgmoore: If we are talking Nvidia, then it is either that packaging or the upstream code at fault | 15:12 |
| RikMills | I cannot belive for one minute that say Nvidia would happy with broken 24.04 LTS systems | 15:14 |
| sgmoore | nvidia is of course the most common point of failure, but I saw an amd in the mix, both proprietary. radeon seems to fix the problem. | 15:14 |
| RikMills | Oh, I only saw nvidia. appologies | 15:14 |
| sgmoore | which is why I am wondering if our seed should e a differnt version. we have 545 now and its seems problematic, one report says 550 fixes it. | 15:15 |
| RikMills | is 550 a beta/RC? | 15:16 |
| sgmoore | It has to be as it wasn't there when I updated the seed to 545 | 15:16 |
| sgmoore | yes beta. not doing that | 15:25 |
| RikMills | I think we have to accept that some issues are going to have to be a target for 24.04.1 | 15:35 |
| sgmoore | okay | 15:35 |
| RikMills | This is how LTS releases work | 15:35 |
| mmikowski | sgmoore RikMills driver support we have seen looks like this: Nvidia > Intel > AMD. The requirement for bleeding edge kernels to get less broken AMD support for the latest hardware means that most folks in that camp are no longer running stock Kubuntu, any least not the stick kernel hwe packages. | 16:42 |
| mmikowski | Intel has had some regressions (and a really bad one with screen flicker and corruption 2 years ago), but are mostly able to work around with kernel parameters. | 16:45 |
| mmikowski | Nvidia works well as long as you pick the right version. No need to change kernels. Some hassle as always with install and upgrades, but once installed, generally very good performance, reliability, and features. | 16:46 |
| mmikowski | At present we curate to 535 with pinning. | 16:47 |
| mmikowski | 545 had reports is regressions, as you noted. | 16:47 |
| mmikowski | We can test 550 drivers and see if they are a good upgrade. | 16:48 |
| mmikowski | There are also considerations for productivity support. We check for CUDA, OptiX, and ML libs for pytorch and TensorFlow. That would be a halo effect of our testing, but realize it does require a pinning package. | 16:52 |
| mmikowski | Which isn't "pure" upstream. | 16:52 |
| sgmoore | Okay, definately sounds like something to investigate for the .1 release | 16:53 |
| mmikowski | sgmoore for example, on the Ir14, there are 3 Intel-specific boot parameters that are applied for that iGPU. | 16:54 |
| mmikowski | See /etc/default/grub.d/kfocus.cfg which is a model specific symlink. | 16:55 |
| mmikowski | I assume we're looking a 550 for stock at present in the upstream? | 16:56 |
| mmikowski | If so, the testing will come as a matter if course. | 16:57 |
| sgmoore | there are a few reported issues with ( on reddit... ) with 545 which is currently seeded. It is said 550 fixes it, however, I am not comfortable shipping with a beta driver. | 16:59 |
| mmikowski | We can help with Intel and Nvidia, and already are testing a bit on metal, but more will come. AMD isn't something we have in the lab. | 16:59 |
| mmikowski | I don't recall the issue off-hand, can you remind me? | 17:00 |
| sgmoore | A pretty big one. Black screen after reboot ( new install ) | 17:01 |
| mmikowski | That's typically for to DKMS install and failing to run update-initrafs -c -k $(uname -r). | 17:02 |
| mmikowski | Not sure that's what's going on here. This is usually done by the DKMS scripts. BUT... | 17:03 |
| mmikowski | It breaks when people upgrade the kernel /and/ the drivers at the same time. DKMS then had made the initrd for the previous kernel before the reboot. | 17:04 |
| mmikowski | And that's the one huge downside to the DKMS approach. Ubuntu do offer precompiled modules in separate packages, but that has it's own problems. | 17:07 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!