07:19 Tom^: imirkin: your patch worked in any case.
07:19 Tom^: =D
07:20 Tom^: imirkin: yea saw the article on phoronix at work, but they introduced a mouse regression tho so its pretty much unplayable it seems :P
07:21 karolherbst: Tom^: which game what issue?
07:21 Tom^: cs:go
07:22 karolherbst: k
07:43 Tom^: fix one thing break three other. https://github.com/ValveSoftware/Counter-Strike-Global-Offensive/issues/574 :P
09:00 hayspat: Hello, I'm trying to test kepler reclocking support as it explained here https://www.reddit.com/r/linux_gaming/comments/3l5tx5/successful_nouveau_reclocking_efforts_test_branch/
09:01 hayspat: there is no longer ddr5 branch, i suppose its in mainstream?
09:07 karolherbst: hayspat: yeah
09:08 karolherbst: I think it is already there in 4.3, 4.4 for sure
09:08 karolherbst: hayspat: yes, 4.3 I just looked it up
09:12 hayspat: @karolherbst before i saw that reddit post, i tried to change pstate and all failed. i have gtx660
09:12 karolherbst: yeah, gddr5 was buggy before that
09:13 karolherbst: there was a _small_ chance though it succeds
09:13 karolherbst: you might encounter volting problems though
09:13 hayspat: Is there anything i can do for debugging?
09:13 karolherbst: but this can be usually workaround by clocking to 0a first
09:13 karolherbst: hayspat: too late for that ;) try the 4.3 or 4.4 kernel
09:14 karolherbst: though there are some issues on desktop systems
09:14 karolherbst: like some screen corruptions
09:14 karolherbst: but this can be workaround by another reclock (0f -> non-0f -> 0f)
09:14 hayspat: I'm currently complinix the kernel from git lets see what happens
09:14 karolherbst: there is something missing for the display part
09:15 hayspat: was using 4.3 from arch
09:15 karolherbst: mhhh
09:15 karolherbst: it should work with 4.3 though
09:15 karolherbst: ahhh wait
09:15 karolherbst: then I know the issue
09:15 karolherbst: it isn't related to the memory then, but the voltage used for the core is too low
09:15 karolherbst: there is an easy hack for this
09:16 karolherbst: hayspat: this function: https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/tree/drivers/gpu/drm/nouveau/nvkm/subdev/volt/base.c?id=refs/tags/v4.3.3#n60
09:16 karolherbst: static int nvkm_volt_map
09:16 Tom^: is it really that safe to do .max on all the cards?
09:16 karolherbst: you could replace info.min with info.max
09:16 karolherbst: Tom^: depends actually
09:16 karolherbst: usually yes, because we never go beyond what the hardware can do
09:16 karolherbst: heat is one problem
09:17 karolherbst: drawing too much power is another one
09:17 karolherbst: but the card itself is save
09:17 Tom^: mm
09:17 karolherbst: so we don't hack this in by default
09:17 karolherbst: but as long as you keep an eye on the temperature it is fine
09:17 Tom^: probably should warn the others before letting them run with the hack then :P
09:17 karolherbst: I usually do that :D
09:18 karolherbst: hayspat: do you know if your PSU is quite strong?
09:18 hayspat: well i never changed sources, could you give me more instruductions something to google
09:18 hayspat: yes my psu is quite strong
09:18 karolherbst: hayspat: well if you compile your own kernel you should be able to modify the source too
09:18 karolherbst: hayspat: like more than 20% buffeR?
09:18 hayspat: 750w for just gtx660 probably could handle 3x gtx660
09:18 karolherbst: mhh
09:18 karolherbst: acutally I won't :D
09:18 karolherbst: gtx660 is 140W+
09:19 karolherbst: ohh well
09:19 karolherbst: yeah
09:19 karolherbst: okay, it works :D
09:19 Tom^: pfft, thats only like half of mine.
09:19 Tom^: =D
09:19 hayspat: but i think its mroe than enough :D
09:19 karolherbst: depends on the cpu though
09:19 karolherbst: but yeah
09:19 karolherbst: seems okay
09:19 karolherbst: but 750W with only one gtx660 is plenty
09:19 hayspat: I have full responsiblity if something happens dw
09:19 hayspat: i know the consequenses
09:20 karolherbst: k
09:20 karolherbst: but overvolting isn't an issue by itself anyway
09:20 karolherbst: going above 150W is one
09:20 hayspat: hmm but i have overclocked bios on my card already
09:20 hayspat: irrc
09:20 karolherbst: ahhh
09:20 karolherbst: then it makes totally sense it crashes :D
09:20 hayspat: lol never throught about it
09:20 hayspat: yea
09:21 karolherbst: usually those OC have a pretty tight safety margin for voltage
09:21 karolherbst: and nouveau uses lower voltages than nvidia
09:21 karolherbst: all those OCed cards have currently a volting problem
09:22 karolherbst: hayspat: do you invoke your make command yourself or just PGKBUILD?
09:22 hayspat: ok, im gonna flash default bios
09:22 karolherbst: hayspat: there is an easier way :D
09:22 hayspat: how can i use whisper?
09:23 karolherbst: why whisper?
09:23 karolherbst: :D
09:23 karolherbst: its the nouveau channel we can talk as much about nouveau as we want
09:23 karolherbst: :p
09:23 hayspat: oh okay
09:23 hayspat: didnt wanna disturp the channel
09:23 karolherbst: hayspat: you can always specify the firmware to use by nouveau with the nouveau.NvBios= parameter
09:23 karolherbst: but because you use an initramfs this might be a bit tricky though
09:24 karolherbst: mupuf: does nvafakebios work across boots as long as the machine stays powered on?
09:24 karolherbst: ohh I doubt it
09:24 karolherbst: ... nvm
09:25 hayspat: well i havent used nouveau till now because it was causing screen freezes
09:25 hayspat: hope using external firmware solves it
09:25 karolherbst: hayspat: well maybe flashing the stock bios is the easier way for now
09:25 hayspat: maybe oced bios was the problem for screen freezes
09:25 karolherbst: no
09:25 karolherbst: gddr5 was
09:26 karolherbst: and voltage
09:26 karolherbst: but now it is only voltage
09:26 hayspat: you can use me as your test rab im happy to help :D
09:26 hayspat: Flashing original bios then
09:27 karolherbst: hayspat: well if you have a lot of time you might look into this: https://trello.com/c/2Zd5nMU9/91-linebuffer-iso-hub
09:27 mupuf: karolherbst: nope
09:28 hayspat: I dont have second monitor
09:28 mupuf: true, I should work on nvafakebios and not spend time on adding new machines. I was discussing this with hakzsam privately. Anyway, time to go back home!
09:30 karolherbst: :D
09:30 karolherbst: hayspat: no, this also applies with one monitor already
09:30 karolherbst: I think
09:30 karolherbst: not sure though
09:30 karolherbst: but
09:30 karolherbst: there is something wrong with the reclocking leading to screen corrputions
09:31 karolherbst: mupuf: I've updated the kepler reclocking trello by the way
09:31 karolherbst: I hope I didn't miss anything
09:32 hayspat: karolherbst: how can i test this? I need some instroductions sadly didnt use trello before
09:32 karolherbst: no clue
09:32 karolherbst: I have a mobile chip
09:33 karolherbst: so this issue doesn't apply to me at all
09:33 karolherbst: maybe it even doesn't for you
09:33 karolherbst: who knows
09:33 karolherbst: best is, check if your stock vbios works for nouveau
09:33 karolherbst: and reclock like 100 times
09:33 karolherbst: if nothing bad happens then you are fine to go I guess
09:34 hayspat: okay currently waiting for kernel to compile but i guess 4.3 is good to go
09:34 karolherbst: I also just heared of a few systems where this corruption happens, so I am really not sure if that's a general thing
09:34 karolherbst: yeah, 4.3 is good
09:35 hayspat: is there a way to flash bios in linux? else I'm gonna do that in windows
09:36 karolherbst: no idea, I bet there isn't a tool for kepler gpus anyway
09:36 karolherbst: I know there was stuff for much older cards
09:37 hayspat: okay, switching to windows brb in 10mins
09:38 imirkin: huh? karolherbst, your fix is only in 4.4
09:38 karolherbst: gddr5?
09:38 karolherbst: it sure is in 4.3
09:39 imirkin: karolherbst: your patch to fix the pll dividers? only in 4.4
09:39 karolherbst: imirkin: https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/commit/drivers/gpu/drm/nouveau?id=78eaf335e4c8224e74e5d512f20ec48109db9dac
09:40 karolherbst: mhh maybe this lies to me?
09:40 imirkin: it doesn't lie
09:40 imirkin: it just doesn't show you what you think it shows
09:40 imirkin: karolherbst: http://hastebin.com/emodututiq.hs
09:40 karolherbst: ohhhhhh
09:40 karolherbst: then I browsed the git tree for the current rc :/
09:40 karolherbst: meh
09:40 karolherbst: my fault then
09:41 karolherbst: ,.,.
09:41 karolherbst: found the issue
09:41 karolherbst: when I browse a tag and click on log, the tag reference is gone int gitweb :/
09:42 karolherbst: why...
09:42 imirkin: use the cmdline
09:42 karolherbst: yeah, now I will
09:43 karolherbst: but his OC vbios won't have worked most likely anyway :/
09:59 hayspat: karolherbst: flashing original bios worked! I'm currently running on 0f state
10:00 karolherbst: hayspat: with 4.3?
10:00 hayspat: no 4.4git
10:00 karolherbst: hayspat: because I was wrong, the fix is only inside 4.4
10:00 karolherbst: ahh k
10:00 karolherbst: hayspat: did you test 4.4 with the OCed vbios?
10:00 hayspat: nope
10:00 karolherbst: well you can try it out later anyay
10:00 hayspat: but the oced bios had some voltages i dont think it will matter
10:01 karolherbst: hayspat: same voltage for same clock?
10:01 hayspat: oced bios has 993mhz core non,oc 980 with same voltages
10:01 karolherbst: ohhh 13MHz difference
10:01 karolherbst: mhhh
10:01 hayspat: :D
10:01 karolherbst: with bad luck it won't work :D
10:01 hayspat: haha
10:01 karolherbst: or it will crash 10% more likely
10:01 karolherbst: or there is a chance it will crash and with stock there is none
10:01 karolherbst: who knows
10:01 hayspat: It says there are 2 types of memory one from hynix one from samsung
10:01 hayspat: mine was samsung
10:01 hayspat: it that helps
10:02 karolherbst: this is about engine reclocking, not memory
10:02 karolherbst: hayspat: but you could try this: reclock to and from 07 to 0f
10:02 karolherbst: try that out like several times
10:02 karolherbst: and see if you get any screen stuff
10:03 karolherbst: if something odd happens, just reclock until it is gone
10:03 karolherbst: :D
10:03 hayspat: well changed it like 30 times nothing happened
10:03 hayspat: i think its stable
10:04 karolherbst: mhhh
10:04 Tom^: run unigine heaven
10:04 hayspat: kk downloading
10:04 hayspat: but i have another problem now
10:04 hayspat: Xorg crashes with nouveau whereas wayland is working fine
10:04 hayspat: I'm on wayland now
10:05 karolherbst: hayspat: while true; do echo 07 > pstate && sleep 2 && echo 0f > pstate && sleep 2; done
10:05 karolherbst: and run this for some time until something happens
10:05 hayspat: kk
10:06 Tom^: imirkin: uhm something changed in mesa recently ? im seeing a ~5fps loss in unigine heaven :o
10:06 hayspat: I'll just run it from terminal is that right?
10:06 hayspat: or create a bash file
10:06 Tom^: or this SwapLimit "2" is lowering it even tho im not vsyncing.
10:06 imirkin: swaplimit 2 is a bad idea
10:06 imirkin: you really just want swaplimit 1 :)
10:07 Tom^: i want triple buffering ;_;
10:07 karolherbst: hayspat: terminal
10:07 karolherbst: Tom^: application usually do that on their own
10:07 imirkin: Tom^: then take the 5fps loss.
10:07 karolherbst: :D
10:07 Tom^: unacceptable conditions.
10:08 imirkin: conceivable that i messed something up in my draw stuff
10:08 Tom^: yea il disable it and test again before you go hunting
10:08 imirkin: i made the single-indirect-draw case use the same macro as multidraw
10:08 hayspat: but it doesnt do that screen effect when changing via echo 0f > /sys/class/drm/card0/device/pstate command
10:08 imirkin: my theory was that it didn't matter
10:08 imirkin: since draws are slow, what's an extra handful of macro instructions to execute
10:08 hayspat: currently running the command anyway
10:08 karolherbst: hayspat: which screen effect?
10:09 hayspat: karolherbst: screen goes black for 0.1 sec
10:09 karolherbst: hayspat: everytime it reclocks?
10:10 hayspat: only when its reclocking to 0f
10:10 karolherbst: mhhh
10:10 karolherbst: mupuf: is this normal?
10:11 imirkin: screen blanking? it's not *not* normal
10:11 imirkin: it should ideally not happen
10:11 imirkin: we must mess something up with vblanks, or the reclock takes too long, or... something
10:11 imirkin: (that linebuffer thing)
10:11 hayspat: oh it happens everytime when reclocking
10:12 hayspat: from every state to state
10:12 karolherbst: k
10:12 karolherbst: imirkin: yeah I know it shouldn't happen, but maybe it is the current state with nouveau
10:12 imirkin: :)
10:12 hayspat: but it doesnt do that with your terminal command
10:12 hayspat: is there anyway to get traceback with that command
10:12 hayspat: if its working
10:12 karolherbst: hayspat: maybe you write into the wrong file then ;)
10:12 mupuf: karolherbst: what platform?
10:12 mupuf: Never saw those after tesla
10:12 karolherbst: mupuf: kepler
10:13 mupuf: imirkin: the line buffer should not be a problem ... since we reclock during vblank
10:13 karolherbst: hayspat: you should execute my while stuff inside /sys/class/drm/card0/device/
10:13 Tom^: ive never reflected on that but yes i got a tiny flicker to when going from anything to 0f
10:13 karolherbst: :D
10:13 hayspat: echo 0f > /sys/class/drm/card0/device/pstate i used this command
10:13 hayspat: yeah small flickering
10:13 karolherbst: why doesn't anybody say anything :D
10:14 mupuf: karolherbst: you wouldn't see it anyway :D
10:14 mupuf: you optimus user! :p
10:14 karolherbst: right :D
10:17 Tom^: imirkin: http://i.imgur.com/AtPm2id.png :/ in comparision to earlier http://i.imgur.com/uDLW4Gi.png
10:17 karolherbst: Tom^: disabled this x option?
10:17 Tom^: yea
10:18 karolherbst: because I also felt like nouveua got slower
10:18 imirkin: Tom^: bisect
10:18 imirkin: maybe it was the shader draw parameters
10:20 hayspat: karolherbst: reclocking command caused hanging and color corruption
10:20 hayspat: it was fine for 1min
10:20 hayspat: then boom
10:20 karolherbst: hayspat: hanging like in real hang
10:20 karolherbst: or can you still reclock?
10:20 hayspat: yea had to restart
10:20 karolherbst: k
10:20 karolherbst: then maybe there is something in the log
10:21 hayspat: where is log file located at?
10:21 karolherbst: hayspat: your nornal system log
10:21 karolherbst: hayspat: are you using systemd?
10:21 hayspat: yep
10:22 karolherbst: then if it isn't corrupted check "journalctl --boot -1"
10:23 hayspat: it isnt there
10:23 hayspat: I'll try to reproduce
10:24 hayspat: maybe it can catch then
10:24 karolherbst: hayspat: k, next time try ssh
10:24 karolherbst: hayspat: is ssh setup for the machine?
10:24 hayspat: nop
10:24 hayspat: I have an android smartphone maybe i can usethat for ssh
10:25 hayspat: but gotta work on it
10:25 karolherbst: if ssh works you can check dmesg and stuff
10:25 karolherbst: mhhh
10:25 karolherbst: so you are in irc with your phone?
10:25 hayspat: no
10:25 hayspat: with my pc
10:25 karolherbst: ahh k
10:25 karolherbst: ssh from a phone is a bit ugly :/
10:25 hayspat: well I dont have a second pc
10:25 Tom^: karolherbst: why=
10:26 karolherbst: Tom^: because it is :D
10:26 Tom^: android has a few nifty terminals that works ok :P
10:26 karolherbst: hayspat: yeah it should easy to check dmesg
10:26 Tom^: and with busybox and root you barely notice a difference ;)
10:26 karolherbst: but I find it takes some time to do usual stuff
10:26 karolherbst: Tom^: yeah when you are okay with a crappy keyboard that is :D
10:33 hayspat: karolherbst: It didnt even last 10secs this time, no logs again
10:33 Tom^: are you using the volt hack?
10:34 hayspat: no
10:34 hayspat: It kept reclocking, screen was flickering but it was frozen could use mouse etc had to restart
10:34 hayspat: couldnt*
10:34 imirkin: "it kept reclocking"?
10:35 karolherbst: imirkin: while loop
10:35 imirkin: the reclock happens once, when you run the command
10:35 imirkin: why?
10:35 karolherbst: hayspat: that's why you should use ssh to get the dmesg output before rebooting
10:35 karolherbst: imirkin: testing
10:36 hayspat: ok I'll try to setup ssh
10:36 karolherbst: hayspat: I am sure though this is due too low voltage
10:37 hayspat: I was having this issue since like 2-3 years, screen freezing with nouveau
10:37 karolherbst: hayspat: well you could try out the one volt hack, this should fix the issue
10:37 karolherbst: but it is an ugly workaround
10:37 hayspat: after some use it would lock and freeze the pc so I never could use nouveau properly
10:39 hayspat: I can try it
10:39 hayspat: lets see if its caused by low voltage
10:39 karolherbst: hayspat: can you add patches to PKGBUILD?
10:40 hayspat: never tried it but i can google it
10:40 hayspat: how hard it can be :D
10:43 hayspat: had another lock up on 07 state
10:45 karolherbst: hayspat: https://gist.githubusercontent.com/karolherbst/769c4918b3c49459d067/raw
10:45 imirkin: if you're stress testing the perf level switching stuff, you should expect lockups
10:45 Tom^: karolherbst: https://www.youtube.com/watch?v=mm9UWVrykUA the clicking is me jumping between 07 and 0f on my keybindings.
10:45 Tom^: karolherbst: :p
10:46 karolherbst: imirkin: after like 1000 reclocks, yes
10:46 karolherbst: :D
10:46 karolherbst: and 2 second pause is pretty much safe
10:46 imirkin: karolherbst: you're not driving any displays
10:46 karolherbst: ohh right, forgot
10:46 karolherbst: Tom^: then reclock like 100 times a second :D
10:47 karolherbst: yeah but this flicker is bad
10:47 karolherbst: but it looks like vblank related
10:47 Tom^: im running a non composited wm so im not vsync ing it.
10:48 karolherbst: I meant, the display reads the buffer up until one point and then it is gone or something
10:48 karolherbst: maybe running it vsynced helps with that
10:49 Tom^: doesnt happend on blob tho
10:50 hayspat: karolherbst: Had to switch to nvidia blob - so many lock ups
10:50 hayspat: how can i use max power workaround?
10:50 karolherbst: hayspat: even on 07 without reclocking I guess?
10:51 hayspat: yep even on 07 without reclocking
10:51 karolherbst: hayspat: patch your kernel with that: https://gist.githubusercontent.com/karolherbst/769c4918b3c49459d067/raw
10:54 Tom^: imirkin: hm im almost starting to suspect libdrm
10:54 karolherbst: yeah I also tried to bisect
10:54 karolherbst: but I didn't see any change on 11.1
10:54 Tom^: yea me neither
10:55 karolherbst: Tom^: maybe 10.3
10:55 karolherbst: ?
10:55 Tom^: but im relatively sure i was on 11.1+ when i got the highscore.
10:55 karolherbst: I know
10:55 karolherbst: something else changed
10:55 karolherbst: no idea what
10:56 karolherbst: maybe it is thi nvif change in libdrm
10:56 karolherbst: :/
10:56 Tom^: thats the only thing that i can think of that has changed.
10:56 karolherbst: Tom^: you are on 2.4.66?
10:56 Tom^: because im still on your Tom branch
10:56 Tom^: :P
10:58 Tom^: karolherbst: im on -git , http://cgit.freedesktop.org/mesa/drm/commit/?id=b38a4b2 at this commit i think, if i understand this pkgver right.
10:59 karolherbst: yeah this is 2.4.66
10:59 karolherbst: try 2.4.65 then
11:01 duelle-again: imirkin: Just wanted to give you an update on my issue: I just started up ubuntu live and could run the commands with xrandr without any issues. I could not join the intel-gfx because it says I am banned .. no idea why.
11:01 imirkin: congrats :)
11:02 duelle-again: More or less.. this doesnt solve my problem on arch yet.. but it verifies, that it is possible to achieve in some mysterious way
11:02 imirkin: duelle-again: this is banned on #intel-gfx: *!*@gateway/web/freenode*
11:03 imirkin: duelle-again: different versions of software
11:03 imirkin: duelle-again: now to determine which software is the relevant one
11:03 duelle-again: So I am not allowed to join from the web client? mkay. Is there anything I can check on the ubuntu live system - i am on right now - that I can check?
11:03 imirkin: also i wouldn't be surprised if ubuntu kept local patches to make things work
11:06 duelle-again: I dont know whether this is relevant, but on this system my screen is LVDS2 instead of LVDS1 (on arch).
11:06 duelle-again: The external ones have the same names.
11:06 imirkin: probably nouveau got loaded first
11:06 imirkin: doubt that'd matter
11:07 imirkin: duelle-again: check your Xorg version and kernel versions
11:09 duelle-again: imirkin: Kernel: 4.2.0-16, Xorg: 1:7.7+7ubuntu4
11:09 imirkin: i wonder what 7.7 means
11:09 imirkin: 1.17 maybe?
11:11 duelle-again: It is listed like this even here: http://packages.ubuntu.com/wily/xorg
11:12 duelle-again: imirkin: "X -version" shows: 1.17.2
11:12 imirkin: ah ok
11:14 Tom^: imirkin: when did mesa start depending on libdrm 2.4.66
11:14 imirkin: Tom^: at some point :)
11:14 imirkin: like... late december
11:15 Tom^: you dont have commit hashes memorised ? im dissapointed.
11:15 karvas: hello, I have serious xserver crashes after 10min-2hours with nvidia gtx 660 and debian 8.2 stable; Error is 'failed to idle channel 0x...' -> What I tried: updating to debian testing -> different error message ("fifo: gr engine fault on channel 5") but crashes as well
11:16 imirkin: karvas: "failed to idle channel" = "the gpu hung for an unspecified reason"
11:18 karvas: I had an older debian installation (i think wheezy) with which i had no problems, after fresh install I'm rebooting every 30mins :(
11:18 karvas: and windows runs smoothly on the same machine, so I guess it's no hardware problem
11:19 karvas: any idea what I can do?
11:19 imirkin: karvas: pastebin dmesg
11:20 karvas: ok, sec
11:20 duelle-again: Sorry for bothering you that often imirkin, but tbh I am not sure what to do with this information. It would be interesting whether there is anything they change when building the ubuntu packages and perhaps even what. Or what can I do now to (hopefully) finding a solution?
11:21 imirkin: duelle-again: start by downgrading your kernel on arch and see if that helps
11:21 imirkin: Tom^: who said i didn't remember the hash? :p
11:21 Tom^: xD
11:21 Tom^: imirkin: i found it anyways.
11:22 duelle-again: Ok, thanks - I will give it a try
11:24 karvas: http://pastebin.com/iSS4yaKT
11:24 Tom^: karolherbst: dafuq 2.4.65 and the git commit before skeggs adds the libdrm bump on mesa. still slow
11:26 Tom^: im missing something obvious..
11:27 karvas: plus latest: http://pastebin.com/GqPgbj82 - If I try to play a HD Video it crashes 100% - which i have done to create this output
11:30 karolherbst: Tom^: did you set NvBoost=3 ?
11:30 Tom^: yes
11:31 karolherbst: then something else is fishy
11:33 karolherbst: Tom^: maybe this? https://github.com/karolherbst/nouveau/commit/1cafff7370bc583a2b119f7dea6723cbf915947f
11:34 imirkin: Tom^: yay, so it's not the stuff i pushed recently :)
11:37 Tom^: karolherbst: i dont get it.
11:37 karolherbst: Tom^: try to rever that commit
11:37 Tom^: karolherbst: "2015-12-29 19:37:52Tom^didnt change my unigine score tho" thats when i tried your gddr5 branch or rather my Tom branch.
11:37 Tom^: im still on that, and im on oldass libdrm and mesa pre libdrm change.
11:37 karolherbst: Tom^: when did you rebuild nouveau the last time?
11:38 Tom^: like at that point, i havent upgraded the kernel in ages.
11:38 karolherbst: k
11:38 Tom^: is my card dying? xD
11:38 karolherbst: then it is something funny
11:38 karolherbst: don't think so
11:38 karolherbst: because I also felt like everything was slower
11:40 karolherbst: x11-libs/libXdmcp, x11-libs/libXi, x11-libs/libxcb, were updated recently :/
11:40 karolherbst: but I doub it is their fault though
11:41 karolherbst: Tom^: try 11.0.7 or 11.0.6
11:41 karolherbst: mesa
11:41 Tom^: im soon blaming nvidias new blob for putting my card in some brokenass state
11:41 karolherbst: Tom^: 361 works for you?
11:41 karolherbst: :O
11:42 Tom^: yes
11:42 karolherbst: mhhh
11:42 karolherbst: I installed it once but downgraded because it didn't work for me
11:43 karolherbst: I guess glvnd was messed up
11:46 karolherbst: wut...
11:46 karolherbst: ohh wait I load nouveau while nvidia was still loaded
11:48 karolherbst: imirkin: how likely could it be, that a newer nvidia driver could cause something like that? :/
11:48 karvas: would you recommend to install the non-free nvidia driver as a last solution?
11:48 karolherbst: karvas: your choice
11:48 imirkin: karolherbst: if you're switching between them, it's conceivable. if it's from a cold boot -> nouveau, then... unlikely.
11:48 karolherbst: debugging issues are hard if you cna't look anything up
11:49 karolherbst: imirkin: yep, cold boot
11:49 karvas: what do you mean by cold boot?
11:49 Tom^: building mesa 11.0.6 which was before phoronix even tested the 780ti or you highlighting me. if that is still slow its for sure not mesa.
11:49 Tom^: then its just libdrm and your branch left
11:50 karolherbst: Tom^: which would be weird, cause for reasons
12:01 Tom^: yep mesa 11.0.6 still low , and with libdrm 2.4.65 wich is before my testing here aswell. so only thing left is your nouveau branch, gotta go back to something from that period of time.
12:01 Tom^: otherwise the blob has really indeed set my card to some broken state
12:04 imirkin: karvas: there are apparently still people who benefit from using blob ctxsw firmware, you could try that
12:05 imirkin: karvas: i don't have the time right now to explain to you how to get it going, but perhaps someone else can
12:06 karvas: I'm trying to install nvidia drivers from repository right now, is there a difference to that?
12:06 karvas: but thanks for the hint!
12:07 Tom^: anyways im a bit tired from working today, gonna hit the bed and continue this tomorrow.
13:13 hayspat: karolherbst: max voltage workaround is working perfectly. no more lock ups. It added a new power state named AC which i cannot switch to.
13:13 karolherbst: no
13:14 karolherbst: AC means current stuff
13:14 karolherbst: it should be there since ever
13:14 hayspat: It wasnt there before :O
13:14 karolherbst: it should be :/
13:14 karlmag: AC power state? That would confuse some people ;-) (AC vs DC power, etc...)
13:14 hayspat: or i have a memory of fish :D
13:15 karolherbst: hayspat: maybe it was -- before?
13:15 karolherbst: no clue though
13:15 hayspat: but its strange that default nouveau even on 07 state locks up the gpu which makes nouveau completely unusable
13:15 karolherbst: hayspat: low voltage
13:15 karolherbst: hayspat: did it also happens when you didn't reclock at all?
13:15 hayspat: yep
13:16 hayspat: it was there for like 3 years
13:16 hayspat: since i got my gpu
13:16 karolherbst: ohh wow
13:16 karolherbst: nouveau leaves the gpu in default state as far as I know
13:16 hayspat: always had to use property blob
13:16 karolherbst: mupuf: does nouveau change voltage at boot time?
13:16 karolherbst: I thought not
13:16 hayspat: even on default state it causes hangs
13:16 karolherbst: mupuf: but then again, why is the default voltage also too low :/
13:18 hayspat: however i have another problem now, cannot use xorg with nouveau, I'm currently on wayland session
13:18 hayspat: xorg works fine with nvidia blob
13:18 karolherbst: hayspat: xorg log pleas
13:18 karolherbst: *please
13:20 hayspat: http://pastebin.com/dXEhSvsk
13:22 karolherbst: hayspat: you have to remove the nvidia generated xorg.conf file
13:22 karolherbst: because currently it uses the nvidia xorg module
13:22 karolherbst: not the nouveau one
13:23 hayspat: oh thanks!
13:51 phomes: I was playing around with nvbios from envytools. I get an error about the record size (5) in IUNK21 for all my cards. Looking at the entries they seems, and sometimes are, the same. Is the warning worthwhile to keep? And do we have any idea about the contents?
14:19 mwk: phomes: no idea wtf it is, I just wanted to raise a bell when we get unexpected size
14:20 mwk: we can throw in another expected size or whatever
14:35 phomes: mwk: the cards I tested were all old G92/G94/G200. Maybe the record size is different for later models. The version in the table is still 0x10. Maybe the record size is just not a fixed size for the version?
14:36 mwk: could be
14:37 mwk: if version number is the same, let's just accept the other size too
23:36 Tom^: yea lol mesa 11.0.6 , libdrm 2.4.65 , nouveau from same period of time as i got the scores.
23:36 Tom^: my card is dying or the blob changed it
23:46 Tom^: but it isnt slow with the blob tho :/