03:05imirkin: karolherbst: let me know when you feel comfortable that you've tested my text patch enough
03:34imirkin: karolherbst: i'd like to get this into the 19.0 branch
03:34karolherbst: imirkin: mhh, when is the branch point?
03:34imirkin: Tuesday Jan 29, 2019
03:35imirkin: [should be plenty of time]
03:35karolherbst: mhh, bad timing :/ these weekend is devconf
03:35karolherbst: well, I installed the patch system wide and run steam and tried a few games with it
03:35imirkin: and it didn't totally die presumably
03:35karolherbst: worked fine... but nothing really triggers eviction that often anyway
03:36karolherbst: nope, it didn't
03:36imirkin: dirt showdown did, when forcing it to run with bindless textures
03:36karolherbst: the webgl cts is fine with it as well
03:36imirkin: that's how i did the initial debugging
03:36imirkin: ok, that's cool
03:36karolherbst: mhh, bioshock infinite does it three times
03:36imirkin: the problem is when you have tons of contexts
03:36karolherbst: only chromium does it
03:36imirkin: which is not something i had really considered before
03:37karolherbst: heh, did bioshock increase in size? 36GB
03:38gnarface: no the save files are just huge
03:38gnarface: like 10MB each or some shit
03:38gnarface: +- some variable amount depending on the map
03:38karolherbst: how many save files do I need to have so that's visible on the download size?
03:38karolherbst: I thought it was around 20GB or something in the past.. maybe I just remember wrong
03:39gnarface: you mean in-game? i wasn't aware it ever showed up
03:39gnarface: maybe it did and i forgot
03:39karolherbst: gnarface: no, steam tells me it wants to download 36GB ;)
03:39gnarface: but i think i checked them out on-disk in wine
03:39gnarface: oh to DOWNLOAD ith
03:39karolherbst: ohh, maybe it's the shader cache thing as well
03:39karolherbst: that's new in steam
03:39gnarface: i actually only have bioshock 2
03:40karolherbst: :D (71 MB pre cached... sooo I doubt that's it
03:40karolherbst: doesn't matter anyway
03:40karolherbst: imirkin: what I would like to test is to run the entire plasma session with that patch, but... I think its fine generally
03:43imirkin: well, i'm going to push over the weekend then
03:43imirkin: if you run into issues first, let me know
07:18sigod_: anyone know of a kepler series card that doesnt have the context switching crash with acceleration like the gtx 660. Is the GTX Titan any good?
07:19sigod_: or will all kepler series cards do that without the non-free firmware for video acceleration
08:16karolherbst: soooo... finally wrote that script for our shader-db repository
08:16karolherbst: sigod_: we don't really know why it happens
08:16karolherbst: it seems to only cause issues on a few models, usually 660s
08:17karolherbst: but... no idea why, no idea how to fix it
08:17karolherbst: could affect any card for all we know
08:45sigod_: thanks karolherbst
08:46sigod_: would be good if intel supported libreboot and nvidia made their drivers open source
08:46sigod_: i mean free software
08:46HdkR: You don't want Nvidia to open source their driver. You want Nvidia to release full documentation on their hardware and to fund Nouveau
08:47sigod_: ok that sounds better
08:47sigod_: that's what i was meaning
08:48sigod_: but that is a fantasy world
08:48sigod_: where things actually work
08:50karolherbst: well, if you just want open source you get something like ROCm :/
08:51sigod_: what the heck is that
08:51karolherbst: AMDs compute stuff
08:51karolherbst: currently they are in damange control mode with the community
08:52sigod_: amd cpus and gpus need non-free firmware though don't they?
08:52karolherbst: that's not the issue
08:53karolherbst: ROCm is like propritary software open sourced, but with a free license"
08:53karolherbst: they ar ehapy to accept patches, but AMD controls everything
08:54sigod_: ok so not gpl
08:54karolherbst: the license doesn't matter
08:54karolherbst: you can fork it, but what would you get from it?
08:55sigod_: do they release hardware specs?
08:55karolherbst: well they have open documentation
08:56sigod_: so the community can't write their own driver
08:57karolherbst: they could, but that rocm stuff isn't a driver either. it's basically just the compute runtime. They use the llvm compiler for example.
08:57karolherbst: it's just not a community driven project
08:57karolherbst: and it was never intended to be one
08:57sigod_: i think i need to read up on it; it sounds confusing.
08:58sigod_: they could just release hardware specs for their latest boards and cards though
09:04sigod_: the problem with corporations is they are focused more on profit than community in the first place
09:06sigod_: The technology is actually a product of capitalism and it doesn't actually liberate us from anything
09:06sigod_: ill stop there
09:12sigod_: karolherbst, which operating system do you use?
09:12karolherbst: in which sense?
09:13sigod_: do you have a main pc
09:13karolherbst: I don't think I could say I have a "main" one
09:13sigod_: are you an arch user?
09:14sigod_: windows 10?
09:15karolherbst: why are you asking anyway?
09:15sigod_: just wondering what you tech guys use
09:15karolherbst: I don't use the same thing on all my machines as those usually serve different purposes
09:15karolherbst: everything is Linux based though
09:16HdkR: I use Ubuntu with a different DE because I'm super lazy at OS setup :P
09:17sigod_: HdkR, fair enough
09:18karolherbst: currently I use 1x fedora, 2x gentoo (but one of them I will move over to something else as those are basically just clones and I want to setup the one machine from scratch), 1x I don't know yet, have to migrate it to something where I can run containers easily on,
09:19sigod_: do you wear a fedora?
09:19karolherbst: not right now
09:20HdkR: lol, start the data mining
09:20karolherbst: the gentoo system is mainly for testing nouveau
09:20karolherbst: super easy to just patch system libraries
09:20karolherbst: everywhere else it's just painful
09:21karolherbst: on gentoo it's two commands, everywhere else it's at least 10
09:21sigod_: dont you have to compile everything on gentoo?
09:21karolherbst: sure, but that's why it's easy to patch things
09:21sigod_: or emerge
09:21sigod_: does it take ages it used to on my old 500mhz celeron
09:21karolherbst: if you need a patch to fix your package on any other distribution, you are essentially screwed
09:21HdkR: Just throw more cores at the problem :P
09:22karolherbst: yeah, more CPU :p
09:22karolherbst: I wouldn't run gentoo on such a low end system... except I'd have a build server
09:22HdkR: Personally I adore breaking apt's dependency tracking and spending hours fixing it </s>
09:23sigod_: i use parabola
09:24sigod_: but i hate the constant updating
09:25sigod_: im on win 7 right now though :o
09:25HdkR: I was using Windows recently and I purchased new hardware just to get away from it
09:26sigod_: which hardware?
09:26HdkR: Went from a Surface Book 2 to a 2019 Razer Blade 13"
09:26HdkR: SB2 has major issues under Linux that aren't easy to work around
09:28sigod_: HdkR, which video driver do you use on that?
09:28sigod_: the razer i mean
09:29HdkR: On the Razer I'm mostly using i965. Can't use Nouveau since the MX150 has PM issues. Sometimes I use the blob for testing things though
09:29sigod_: so ubuntu uses free software intel drivers?
09:30HdkR: Of course
09:30HdkR: Intel doesn't ship closed blob drivers for GEN
09:31sigod_: does the intel video have hdmi out?
09:31HdkR: On Linux
09:31HdkR: This device doesn't have a pyhsical HDMI output port. Need to use an adapter for type-C to HDMI
09:32HdkR: Can still drive three heads regardless
09:33sigod_: would a type-c to hdmi be as good as native hdmi?
09:35HdkR: Not sure. The Type-C to DP passive adapters would be
13:43denvit: Hey everyone!
13:44HdkR: Hello single one
13:44denvit: I'm having a problem with nouveau and a gm20b (NVIDIA Tegra X1)
13:44denvit: I'm currently running linux 5.0.0-rc3
13:45denvit: I've built xf86-video-nouveau and libdrm with Tegra API support
13:45denvit: but I can't seem to be able to run any GL stuff
13:45denvit: glxgears returns a blank window (but seems to render something since the framerate is shown in the console)
13:45denvit: glxinfo shows the following:
13:47denvit: Video Memory: 0 MB
13:48denvit: and dmesg shows the following error: [ 4416.391365] nouveau 57000000.gpu: gr: DATA_ERROR 0000009c  ch 5 [0400415000 glxgears] subc 0 class b197 mthd 0d78 data 00000052
13:48denvit: [ 4416.391389] nouveau 57000000.gpu: gr: DATA_ERROR 0000009c  ch 5 [0400415000 glxgears] subc 0 class b197 mthd 0d78 data 00000016
13:49denvit: This is the Xorg output: https://gist.github.com/denysvitali/70ec80839a2c12202f20e09d5462688b
13:49denvit: I'm using modesetting as a driver
14:28karolherbst: denvit: what's your version of mesa?
14:29karolherbst: ohh, no, should be fine "OpenGL renderer string: tegra"
14:29karolherbst: tagr: ^^
14:29karolherbst: for you :p
14:30karolherbst: the data error is weird though, allthough I would ignore it maybe?
14:30karolherbst: denvit: you could try if it's better with the nouveau DDX
14:30karolherbst: "Page flip failed: No such device or address" mhh
14:37imirkin_: denvit: i don't think xf86-video-nouveau can really support that use-case
14:37imirkin_: denvit: the GM20B does not support displaying anything
14:37imirkin_: there's a separate device (tegra) which works the display
14:37imirkin_: and the GM20B itself is only used for acceleration
14:37imirkin_: i THINK (but am not 100% sure) that xf86-video-modesetting should support that use-case
14:37karolherbst: imirkin_: in the X log it indicates it uses modesetting anyway
14:38karolherbst: the tegra gallium driver is just a wrapper around the PRIME buffer stuff anyway, afaik
14:38imirkin_: ok. then the other thing is that there's no vram. the nouveau mesa driver has been updated to understand that situation
14:38karolherbst: or was there more to it?
14:38imirkin_: however it's not frequently tested
14:38imirkin_: so it could be that some additional problems crept in
14:39imirkin_: oh. i didn't quite finish reading the problem description
14:39imirkin_: ok, so display works, but glxgears is blank.
14:40imirkin_: oh well, he left
14:40imirkin_: the thing to try would have been LIBGL_DRI3_DISABLE=1 glxgears
14:41imirkin_: since it's talking about present flip failures
14:43cosurgi: imirkin_: chromium is damn to slow for me. Would you recommend me some web browser which does not use OpenGL and is generally okay to use?
14:43karolherbst: cosurgi: nouveau is blacklisted in chromium since 71
14:43cosurgi: btw, I was about to ditch chromium anyway.
14:43imirkin_: unfortunately i have to use chrome
14:43imirkin_: firefox doesn't work with gmail for me
14:44imirkin_: which is ... basically the whole reason why i use a browser :)
14:44cosurgi: why? ahh. And you don't know anything else?
14:44cosurgi: okay :/
14:44cosurgi: for mail I also use claws-mail. And this one is fast!
14:44karolherbst: imirkin_: do you also get frustrated by all those overloaded webpage? I get the feeling it got really bad last year
14:44imirkin_: for whatever reason, page up/down doesn't work in the email text editors in firefox
14:44karolherbst: and this year will be even worse
14:44imirkin_: and without page up/down it's pretty useless.
14:45imirkin_: coz ... i need to be able to navigate in an email.
14:45cosurgi: heheh :)
14:45imirkin_: (i know, i have craaaazy requirements...)
14:45karolherbst: cosurgi: I doubt, because software rendering might be even slower :/
14:45karolherbst: cosurgi: blame the web pages
14:45karolherbst: fancy css effects
14:45karolherbst: and shit
14:46imirkin_: cosurgi: do give firefox a shot though, coz other than that one (enormous to me) issue, it's fine
14:46cosurgi: imirkin_: give claws-mail a try. It has nice plugins, and is nicely configurable.
14:46cosurgi: ok. I will check it out :)
14:46imirkin_: does it do gmail-style threading?
14:46cosurgi: Ctrl-t toggles threading in current view.
14:46imirkin_: mutt also has threading
14:46imirkin_: but it's quite different
14:47imirkin_: i mean specifically gmail-style threading
14:47imirkin_: i.e. single list of all emails in the thread
14:47imirkin_: with collapsing of identical sections
14:47cosurgi: is it gmail style - I am not sure. I use gmail in www also. I prefer claws-mail threading style. In fact I used claws-mail and chromium in parallel. And sometimes I just got to switch from chromium to claws-mail just to find my thread.
14:48cosurgi: single list. No. It threads how they were answered to. Now I get what you mean. If I want that in claws, then I disable trheading and sort by date. So it's not excactly the same.
14:49imirkin_: i will investigate
14:49cosurgi: heheh. My last sentence must have discouraged you ;) My bad. But try anyway ;)
14:49imirkin_: sort-by-date doesn't thread
14:50imirkin_: anyways, when gmail became a thing, it was a huge game changer, largely because of how it presented email
14:50imirkin_: it enabled me to process large quantities of email very quickly
14:50imirkin_: and threading is a big part of that
14:51imirkin_: perhaps the solution to that problem is "stop using email for everything"
14:51imirkin_: but that's a larger change :)
14:51imirkin_: denvit: does LIBGL_DRI3_DISABLE=1 glxgears help?
14:52cosurgi: imirkin_: hey, there is something like this. https://lists.claws-mail.org/pipermail/users/2015-August/013808.html "Threading in claws vs grouping in Gmail"
14:52cosurgi: I never tried this...
14:52imirkin_: cosurgi: ok, and the idea is that you hook it up to google using imap or something?
14:53cosurgi: works well with imap.
14:53karolherbst: cosurgi: gmail doesn't do imap and their imap imlementation is quite crappy
14:54cosurgi: hm. I don't know. In fact I enabled imap only for my university mail, and it works fantastically. For gmail I used pop, because I was too lazy to change old config.
14:54karolherbst: for everything besides gmail it's great
14:54karolherbst: but gmail uses this label thing
14:54karolherbst: and that just doens't map all to well to imap folders
14:56imirkin_: heh. pop.
14:56imirkin_: that brings back memories.
14:56cosurgi: that's possible. However, on iphone (apologies, anodroid has too many security holes and can't be updated) I use an app called 'spark' and it seems to work with gmail using imap. At least all the folders are there.
14:56imirkin_: [of writing a webmail client in php... i didn't say the memories were good...]
14:57denvit: karolherbst: Mesa 18.3.2
14:58denvit: karolherbst: Nouveau DDX segfaults :(
14:58imirkin_: denvit: that's expected
14:58denvit: imirkin_: Nope, it hangs the DM, just tried :(
14:58imirkin_: DM = ?
14:59imirkin_: denvit: well, this is not a path well-trodden
14:59karolherbst: yeah.. that dri3 thing should work more likely
14:59imirkin_: i believe the wayland situation is actually a lot likelier to work
14:59imirkin_: [and i'm not just some wayland fanboy saying this ... i much prefer X in general]
15:00imirkin_: denvit: another thing to try is to start X without AutoAddGPU
15:00imirkin_: since it looks like the nvidia board is coming in as G0 on there
15:00karolherbst: imirkin_: for dri2, that's required
15:00denvit: [dvitali@alarm ~]$ LIBGL_DRI3_DISABLE=1 glxgears
15:00denvit: Running synchronized to the vertical refresh. The framerate should be
15:00denvit: approximately the same as the monitor refresh rate.
15:00denvit: 133 frames in 5.0 seconds = 26.563 FPS
15:00imirkin_: which is not actually something that you want in this case, since there should be a tegra driver which wraps nouveau inside it
15:00denvit: this is what happens on SSH
15:00denvit: but actually, the screen freezes
15:01denvit: and doesn't unblock itself unless I restart lightdm
15:01imirkin_: karolherbst: with tegra_dri it's not offloading though, as far as it knows
15:02karolherbst: imirkin_: what do you think is what the tegra gallium driver does?
15:03karolherbst: it opens the nouveau renderer node and let it render stuff
15:03imirkin_: i think it opens an fd to /dev/dri/card1 behind the X server's back
15:03imirkin_: so as far as the DRI2/DRI3 stuff is concerned, it's not offloading
15:03karolherbst: ohh, none of that reaches the tegra ddx
15:03karolherbst: it's all inside gallium
15:03imirkin_: i think so, yeah.
15:03karolherbst: and just plain DRI_PRIME stuff, just.. hidden
15:04karolherbst: and maybe a little better perf
15:04karolherbst: anyway, if you do that with DRI2 you need a DDX for both sides
15:04karolherbst: with DRI3 you won't
15:04karolherbst: _but_, I think the tegra driver does what we would do with dri3 anyway
15:04karolherbst: at least it looks like that
15:04imirkin_: you're getting confused
15:05imirkin_: as far as the X server should know, there should only be one GPU
15:05imirkin_: i.e. the main one
15:05imirkin_: that's why i'm saying AutoAddGPU = false could help
15:06karolherbst: right, but that only works with the dri3 path, aka no "LIBGL_DRI3_DISABLE=1"
15:06karolherbst: (if that matters at all for the tegra driver)
15:07denvit: I'm a bit of a noob with wayland
15:07denvit: I started a weston session, but I can't run glxgears because there is no X display running (I guess?)
15:07denvit: But it looks like it's accelerated
15:08denvit: I mean, the session seems to be running on GPU
15:08karolherbst: denvit: you can enable xwayland or use that eglgears thing?
15:12imirkin_: karolherbst: what dri3 path?
15:12imirkin_: karolherbst: why would it matter?
15:12imirkin_: karolherbst: you're thinking about the offloading case again, when the offload happens in X
15:12karolherbst: it doesn't with dri3
15:12imirkin_: it still does since it has to handle the present call
15:13imirkin_: with a weird dma-buf in it
15:13karolherbst: you don't need any driver for the second gpu
15:13karolherbst: in X
15:20denvit: karolherbst: I don't have any eglgears, and xwayland (despite it's installed) doesn't seem to work
15:22denvit: But I can run alacritty, which uses GPU acceleration
15:24denvit: eglinfo works tho
15:25imirkin_: i think XWayland has to be explicitly enabled
15:30denvit: Okay, glxgears works with xwayland
15:43denvit: But feels pretty laggy
15:43denvit: there are lots of artifacts
15:43denvit: but Chromium works (displays its content instead of a black window)
16:07tagr: denvit, karolherbst: those errors look like the ones you'd get if you're using an IOMMU but aren't telling Nouveau, or vice versa
16:08karolherbst: tagr: ohh, interesting
16:08karolherbst: supporting running using an IOMMU is something I would at least consider an important thing to have
16:09tagr: commit commit b59fb482b52269977ee5de205308e5b236a03917 was supposed to fix that, and it certainly did fix it for me
16:09tagr: then again, if this is gm20b that commit shouldn't have anything to do with it
16:09tagr: I'll need to look at this more closely, but probably won't get around to it until after the weekend
16:09tagr: denvit: mind dropping me a mail about this so I don't forget?
16:11karolherbst: I never actually looked into using the IOMMU for the nv GPU under x86.. I at least have a CPU supporting Vt-d
16:17denvit: Sure thing! What's your email?
19:05cosurgi: imirkin_: no idea why I have done these screenshots, but here you are: http://wilis.pg.gda.pl:55555/tmp/claws/claws-thread1.png http://wilis.pg.gda.pl:55555/tmp/claws/claws-thread2.png
19:05cosurgi: imirkin_: http://wilis.pg.gda.pl:55555/tmp/claws/claws1.png
19:05cosurgi: maybe because I like claws so much ;)
19:10imirkin_: cosurgi: btw, if you have an nvidia kepler board lying around, you'll get much better perf with nouveau
19:11imirkin_: that would be a GTX 6xx or 7xx
19:13imirkin_: i suspect you're strongly constrained by vram bandwidth
19:13imirkin_: and on kepler we can reclock the memory
19:15karolherbst: (and maxwell1)
19:17imirkin_: GTX 750 could still be on the weaker side
19:17imirkin_: [which is maxwell1]
19:19karolherbst: it hadd GDDR5
19:20imirkin_: yeah, definitely a nice little bit
19:20karolherbst: they aren more or less equal to my 770m I used to use :p
19:21imirkin_: ah, so not too bad then
19:21imirkin_: i have a GTX 745 :)
19:21karolherbst: yeah, that one is terrible in comparison :D
19:21imirkin_: dell just throws them into computers randomly
19:21imirkin_: i take them out :)
19:21imirkin_: aka "profit"
19:22karolherbst: GT 740 with GDDR5 would be some nice thing to find
19:22karolherbst: I actually got a GT 730 with GDDR5
19:22imirkin_: yeah, that's a thing
19:22imirkin_: i have a GK208 plugged in now with DDR3 (marketed as GT 730)
19:23karolherbst: there was even a fermi one
19:23karolherbst: which was like 30% as fast
19:23imirkin_: waiting for a cheap K4000 to show up on ebay
19:24imirkin_: 2x DP + DVI is kinda perfect for me
19:25HdkR: You don't want a brand new TU102 to have 3xDP, HDMI, and a USB Type-C port? :P
19:25cosurgi: imirkin_: hm. I specifically bough this 1060 because It had three DP outputs. I gave my GTX780 to my wife afterwards. But it has no 3 DP outputs.
19:26imirkin_: GTX780 would fare much better.
19:26imirkin_: but ... 3x 4K might be tricky
19:26cosurgi: I suppose she would gladly swap the video cards. But without DisplayPort outputs that's a problem. How would I achieve 60Hz refresh?
19:26imirkin_: since each DP port can really only handle 1 4K monitor @60hz
19:26imirkin_: that said ... do you need 60hz?
19:26imirkin_: sounds like you're mostly sitting in vi
19:27HdkR: I had a 60hz panel and 30hz panel connected. Even with just text editing between the two windows it hurt when transitioning :D
19:27imirkin_: HdkR: i don't think TU102 will solve the clock performance issues with nouveau...
19:28HdkR: True, but at idle clocks it may be able to brute force some things :D
19:28imirkin_: tbh i'm not 100% sure what the issue is
19:28imirkin_: but "faster vram" rights a lot of wrongs
19:28cosurgi: heh. I didn't like 30Hz while moving windows around. They were lagging. So... I'm not sure. But there's another problem: I would need to buy new cables and thread them through the holes in two walls. My PC is in anothher room to eliminate noise, and the cables are 6 meter long. Finding a good DP cable was real difficult. I don't want to do this again for HDMI.
19:29imirkin_: yeah =/
19:29imirkin_: and of course, you can't use HDMI 2.0
19:29imirkin_: because that's only on later GPU's
19:29imirkin_: (the ones that don't have memory reclocking)
19:29imirkin_: so yeah. you might be sunk. next time, buy amd :)
19:29glennk: the expensive option is fiber transceivers
19:30imirkin_: much like faster vram rights a lot of wrongs, so does fiber :)
19:31cosurgi: in fact one of my DP cables is optical 10 meters long. No difference with the other two cables :)
19:31cosurgi: I had to try it, to make sure and test how it works. Turns out it was unnecessary.
19:32cosurgi: but for next two month or a little longer I'm out of money for such things. I have to wilve with what I have ;)
19:32cosurgi: But next time, yeah, probably fglrx. :)
19:33imirkin_: open-source drivers are much better actually
19:33cosurgi: oh, you mean radeon ? xserver-xorg-video-radeon
19:33imirkin_: well, for later ones, amdgpu
19:34imirkin_: but most important is the kernel driver in all this
19:34cosurgi: yes, I prefer open source. I forgot that fglrx were the binary blobs
19:34imirkin_: which handles things like dynamic clocks, etc
19:34imirkin_: as well as the 3d driver, which lots of people use for playing games/etc
19:35imirkin_: (and the x accel is based on that same driver, since newer amd chips don't have a real 2d unit anymore)
19:36cosurgi: what about OpenGL - is it stable and non-crashing for a year of constant light usage and a separate xserver for playing games on steam, and this one can crash as much as it wants to ?
19:37HdkR: imirkin_: Are you calling something on Nvidia a real 2D unit? :P
19:37imirkin_: cosurgi: crashes tend to be all-or-nothing
19:37imirkin_: HdkR: the GRAPH hardware has a 2d class
19:37imirkin_: which enables 2d operations
19:37cosurgi: ech not-good. but okay. So how about amdgpu OpenGL stability?
19:37HdkR: wtf is graph? Is that on something old?
19:37cosurgi: better than nouveay+mesa?
19:37imirkin_: HdkR: uhm... maybe you know it as "GR"?
19:38imirkin_: HdkR: https://github.com/envytools/envytools/blob/master/rnndb/graph/g80_2d.xml
19:38imirkin_: (don't worry about the "g80" name)
19:38imirkin_: cosurgi: it's way better than nouveau certainly
19:38imirkin_: but i don't think it has perfect error recovery
19:38HdkR: ah. that thing still exists? fancy
19:38imirkin_: intel has the best error recovery afaik
19:39imirkin_: HdkR: it's the same underlying engine, just a different "api" presented via the fifo
19:39cosurgi: imirkin_: amdgpu it is then! Next time when I have money ;) Can you recommend any particular model with three DisplayPorts ? :)
19:40HdkR: The newest and biggest you can afford? :P
19:40cosurgi: it's that simple? Good.
19:40cosurgi: different than force downgrade to kepler ;)
19:41HdkR: Vega is pretty good stuff if you can afford it
19:43HdkR: RX 580 is typically half the price though
19:43imirkin_: cosurgi: check #radeon. i don't have specific advice.
19:43cosurgi: ok :)
19:44imirkin_: my only specific advice is stop using my software ;)
19:44HdkR: "Get off my lawn!"
19:44imirkin_: nouveau's a lot more fun to hack on than to actually use
19:45HdkR: Sounds about right
22:11denvit: tagr: May I have your email, so that I can ping you about the problem w/ GM20B?
22:13pmoreau: denvit: You can see his email here: https://lists.freedesktop.org/archives/mesa-dev/2019-January/213743.html
22:14denvit: pmoreau: Thanks!