13:12zfnmxt: When using the nouveau driver, I'm only able to output 4K @ 30Hz over HDMI on a 1050 Ti. Could this possibly be fixed by switching to DisplayPort?
13:13zfnmxt: (Fixed being output at 60 Hz)
13:22gnarface: it could possibly be fixed by just trying a different hdmi cable even
13:23gnarface: i couldn't tell you how likely it is though
13:26zfnmxt: Even if the issue is only present with the nouveau driver? I can get 4k@60Hz with the propietary Nvidia driver.
13:27zfnmxt: Nouveau can output 4k@60Hz with a 1050 Ti, right? I guess that's what I should clear up first.
13:30gnarface: oh if it's working right with the proprietary driver then it's not hte cable
13:31zfnmxt: Right, that's what I figured.
13:31RSpliet: zfnmxt: first thing to try, if you haven't already, is a shiny new kernel. Display stuff is under active development, recent improvements may not have made it to the kernel of your distro of choice, especially if that distro is a debian-derivative. Kernel 5.5 preferred
13:31zfnmxt: Hm. I'm on 5.4.24
13:31zfnmxt: I'll try 5.5
13:31RSpliet: Okay, 5.4 is actually not so bad
13:32zfnmxt: But I just went ahead and also bought a USB-C to DisplayPort cable...so maybe that fixes things as well. Won't find out until next week!
13:32RSpliet: Lyude might actually still remember what changed between 5.4 and 5.5 and whether that's relevant ;-) It's a generic piece of advice
13:32zfnmxt: So it's unlikely that nouveau is inherently limited to 4K@30Hz on the 1050 Ti, right?
13:33zfnmxt: That's my chief concern because if that's true, nothing I do will fix this issue :)
13:33zfnmxt: And I'm not sure how to find out if that's true or not!
13:36RSpliet: So... I *thought* nouveau used to be limited to this a while ago by virtue of hard-coding the maximum HDMI bandwidth no matter what. 4K@30Hz didn't fit the bill either way. I also thought many such limitations have been looked at and gradually lifted over the course of time.
13:36RSpliet: But being a frazzled scientist (and a retired nouveau dev at that), I lost all track of time.
13:37zfnmxt: And such a limitation wouldn't apply to DP?
13:37linkmauve: I’ve heard some drivers cheat by encoding to 4:2:0 in order to halve the bandwidth required, even if these formats aren’t part of the HDMI spec.
13:37zfnmxt: RSpliet: Heh, it's okay, I'm a frazzled scientist in the making ;-)
13:37RSpliet: Different ports had different limitations.
13:37linkmauve: zfnmxt, DP has a lot more bandwidth than HDMI typically.
13:37RSpliet: Hard-coding was in absence of knowledge on how to query the hardware for the real limits
13:37RSpliet: And possibly in absence of being able to do dual-link stuff properly
13:38zfnmxt: I see.
13:38zfnmxt: Okay, I'll try the DP cable once it arrives, then!
13:38RSpliet: zfnmxt: I suspect a 5.5 kernel won't make much of a difference, unless this patch matters: https://github.com/skeggsb/nouveau/commit/676944d53ef4a5870316b37ac4b0ae82bdc11413
13:39RSpliet: 5.6 has bigger changes.
13:40RSpliet: The residential expert on this is Lyude, but I haven't just lost track of times but also timezones. You may get a more accurate response later. Feel free to hang around
13:41zfnmxt: Building with 5.5.9 now, so we'll see.
13:45zfnmxt: Yeah, still stuck at 30Hz under 5.5.9.
13:52RSpliet: Mmm okay, sorry for making you build a brand new kernel.. But this makes you more eligible for filing a bug report should you wish to do so :-D Might be easier than IRC to get the attention from the devs who lives in different timezones
13:53zfnmxt: No worries, I'm on NixOS so rolling back is super easy. :)
13:54zfnmxt: Does this constitute a bug? Not sure where the line between limitation/bug exists.
13:56RSpliet: If NVIDIA can do it and nouveau can't, from nouveau's perspective it's either a bug or a feature request. Both go on the issue tracker, but devs get the final say on priorities ;-)
13:56zfnmxt: Right, makes sense.
13:57orbea: last time I submitted a feature request to the nouveau tracker it was closed as notabug :)
13:57orbea: granted several years ago
15:07Lyude: zfnmxt: feel free to assign it to me, and also feel free to nag me every few days if I forget about it ;)
15:08Lyude: (email is lyude at redhat ° com)
15:24karolherbst: imirkin_: I guess there would be no way to support tesselation shaders on nv50, right?
15:24karolherbst: (and also no point anyway)
19:22edgecase: hey what's vulkan support like in nouveau? or mesa/openvk... is nv50 suppoted?
19:25linkmauve: edgecase, inexistant.
19:25linkmauve: Never heard of openvk though, what is it?
19:25Lyude: living up to your name I see
19:27edgecase: ah yes, Phoronix says ENOEXIST. openvk, isn't that name of libopenvk, libGL, etc?
19:28edgecase: hmm seems to be libvulkan1
21:54zfnmxt: Where is the bug tracker for nouveau? FreeDesktop Bugzilla says it's out of commission for new bugs.
22:03zfnmxt: Lyude: https://gitlab.freedesktop.org/mesa/mesa/issues/2656
22:30Lyude: zfnmxt: psssst, that's not a mesa bug ;)
22:39zfnmxt: Lyude: Oof. I didn't see a bug tracker for nouveau-specific things...where should it go?
22:39Lyude: I actually am not sure :S, imirkin / skeggsb do you know?
22:40Lyude: usually it would be fdo's bugzilla but I don't know if we've migrated our issues to gitlab
22:40imirkin: issues were migrated
22:40imirkin: but they're not being emailed out
22:40imirkin: which basically amounts to them going nowhere
22:40imirkin: so reporting here is the latest and greatest thing
22:40zfnmxt: tl;dr I should delete that issue, right?
22:41imirkin: i don't have scrollback, but if it's a gitlab issues under xf86-video-nouveau, i doubt anyone will look at it
22:41zfnmxt: Under mesa :^)
22:41imirkin: people do look at that one, just not nouveau people
22:41zfnmxt: I'll take it down
22:42imirkin: if it's a nouveau issue, you're best off reporting it in here
22:42zfnmxt: I did :D
22:42zfnmxt: I was just told to also make a bug
22:42imirkin: have a link to point people at isn't necessarily terrible
22:42imirkin: but just doing that will get you nowhere.
22:42imirkin: so this is a hdmi 2.0 issue, right?
22:42zfnmxt: I think so.
22:42zfnmxt: I mean, it's not even that important
22:43zfnmxt: I'm starting to realize that 4K on a 27" screen on linux just sucks
22:43imirkin: gah, no way to look at the files inline? sad
22:43zfnmxt: And I should've gone for 1440p
22:43imirkin: thanks gitalb
22:43zfnmxt: Yeh, sorry.
22:43imirkin: it's like they sit in a room thinking up of additional ways to annoy me
22:43RSpliet: it's highly personal.
22:44RSpliet: Soz, was my advice to go for gitlab. Didn't know that "bug tracking" was still an unresolved issue
22:44imirkin: zfnmxt: ok, so one thing with hdmi 2.0 that i noticed is that you have to set the mode AFTER the display is pwoered on
22:44RSpliet: I'll file a bug for the "we need a solution for bug tracking" issue to make sure it gets handled.
22:44imirkin: (fwiw i'm the one who added hdmi 2.0 support, but i only tested on gm20x and gp10x)
22:45zfnmxt: imirkin: didn't help :)
22:45imirkin: ok, so looking at the edid, scdc is reported as present, which is good
22:46imirkin: and you have a 4k@60 594mhz mode, which is also common
22:47imirkin: and you don't have errors in dmesg about how enabling scdc failed
22:47imirkin: let's see... is there a way to enable a message saying it *was* enabled, or did i not think that far ahead?
22:48imirkin: no i did not. but it does scream bloody murder if the write fails
22:49zfnmxt: At some point in those logs I also tried @30Hz, if that matters.
22:49zfnmxt: I don't even know if that will appear.
22:49imirkin: oh, and this is a 1050 Ti, for some reason i read 1650 at first
22:49imirkin: so this is pascal
22:49imirkin: and i tested with pascal
22:49imirkin: so ... it should work.
22:50zfnmxt: Could it be an issue with the monitor?
22:50karolherbst: imirkin: mhhh.. I actually know how to open those attachments as new tabs, but it's not pretty :/
22:50zfnmxt: Unfortunately I have nothing else to test on =/
23:04imirkin: zfnmxt: i mean .. anything's possible
23:04imirkin: but i'd say unlikely.
23:04imirkin: esp if it works with blob
23:04imirkin: coudl be that i missed a spot on gp10x
23:04imirkin: it definitely worked with GP108 though
23:43imirkin: zfnmxt: are you able & willing to do debugging of stuff in the kernel? if so, i can point you to a few critical pieces of code
23:43imirkin: or just look at my series to add the hdmi2 support in the first place
23:54zfnmxt: imirkin: I can, but probably needs to wait until the weekend.
23:54imirkin: i'm in no rush
23:55zfnmxt: I appreciate your help, at any rate :)
23:55imirkin: a few people have come in with issues, so i expect i either missed something, or it's flaky
23:56imirkin: the way that hdmi2 works (for the high frequencies) is that you drop the tmds link rate by 40, and enable scrambling
23:56imirkin: somehow this makes the magic fly
23:56imirkin: enabling scrambling is done as an i2c write, similar to edid but a different address ("scdc")
23:56imirkin: and both the monitor and gpu have to be in agreement on the various settings
23:57imirkin: if the monitor is off when the gpu decides to flip the mode, and ignores the write (/ it errors), then they can get out of sync, leading to the sort of issue you see
23:57imirkin: however that would have triggered errors in dmesg
23:57imirkin: but perhaps there are other more subtle cases we're skipping