01:25 turner: hi, do you know how you can set options like "color vibrance" and "underscan" outside of X, maybe at boot time? they should apply to the high res framebuffer, too
01:42 cata: hi
01:43 cata: can someone help me?
01:45 karlmag: cata: no...
01:46 karlmag: cata: you have to ask a question first
01:46 karlmag: Then maybe someone can
01:46 cata: oh ok
01:47 cata: this may be a dumb question but how can I install the driver for my video card? I'm new to linux
01:48 cata: I tried with the .run stuff on nvidia website but that didn't help
01:49 karlmag: I would imagine nouveau being automatically installed by your distro.
01:50 cata: I don't think so
01:50 karlmag: If you insist on the the proprietary nvidia driver, I must admit I haven't played with that for a few years.
01:51 karlmag: which distro is this?
01:51 cata: I've tried to run cs:go to see how it works and I only got like 20-25 frames
01:51 cata: debian
01:52 cata: is there any command to check?
01:55 karlmag: You know, I think it might be more fruitful for you to check in #debian on irc.debian.org (aka irc.oftc.net) first. They should know quite specifically the quirks of debian on most matters I would think. This channel is intended for the nouveau driver related stuff (inluding development) specifically.
01:56 cata: oh ok
01:56 cata: thanks
01:56 karlmag: cata: no problem and good luck :-)
01:58 pmoreau: cata: Do you want to use Nouveau or the Nvidia driver?
01:59 martm_: pq: why do you wuassimodo spy on joss and martm, yeah same person, so what the fuck, why do you keep violating stuff, obviously its stuff that you do not understand in greater detail? what is the problem that you have?
02:01 turner: the parameters you see with xrandr --prop, can you change them without using xrandr? before you login
02:02 pmoreau: turner: With fbset maybe?
02:03 martm_: when ratherDifficult is joss, i mean why do you care, how long you intend to ban any appareance of mine, how do you think it helps you with those terroristic acts, isn't that like asking for trouble?
02:06 turner: can you use fbset for that? i think the parameters are specific to nouveau
02:07 pmoreau: You can use fbset to only modify Nouveau's fb
02:08 pmoreau: Or you mean that underscan and color vibrance are specific to Nouveau?
02:08 turner: yes, and dithering mode, scaling mode and such
02:11 martm_: pq: the conspiracy yeah was raised here against me, unluckily by persons who are not capable of doing any kind pof performance, i mean is that so complicated to understand, all shots missed all entirely wrong, it's quite yeah understandable that such project stuff and scam
02:15 pmoreau: turner: Then I don't know, sorry. imirkin was talking about using modetest tool (part of libdrm) to change the dithering.
02:16 martm_: pq: there is nothing to be afraid of, yeah i had some trouble with people, but those are all entirely nonsense incapable versions, i can not see them doing anything useful in world, i am sick cause those persons can not do anything useful
02:18 martm_: when you give human being a ball, that kick it to the destination, be that basketball or soccer ball, and human being only sabbotages behind your window, honestly every shot is missed, i mean what the hell what was ones were dealing with younger ages then, how can i deal with such?
02:23 martm_: pq: basically the biggest downside of the retard, is not that he can't do stuff, but that it wants to get rid of high seats all the time, so in real competition weaker must be cleaned off from the road, cause theyd distract stronger ones, abviously in life this can't be managed all the time
02:23 martm_: since there are more humans who can not handle simple things then those who can
02:25 martm_: its just as simple in the end, that there is no reason for me to listen to any scams, until you put the shots in the basket and prove you can understand stuff
02:28 cata: why aptitude can't find linux-headers-2.6-686
03:20 pmoreau: cata: Linux 2.6 is crazy old! It was first released in Dec. 2003.
04:32 karolherbst: mupuf: how to update the blob version on reator? I want to check if maybe the blob tells us the actual voltage (it was added in later driver versions for kepler+)
04:33 mupuf: really? Sweet
04:33 mupuf: just use grub-reboot 2
04:33 mupuf: and pacman -Suy
04:33 mupuf: it may take some time
04:34 karolherbst: mupuf: mhh it works "maybe"
04:34 karolherbst: for me it just returns 0
04:34 karolherbst: for others always 1.2V
04:34 mupuf: ah ah
04:34 karolherbst: maybe we are lucky with your maxwell
04:34 karolherbst: who knows
04:34 karolherbst: there should be also "overvoltaging" support
04:34 mupuf: well, I am all for checking my work!
04:34 karolherbst: if that works, then maybe we can get some information
04:35 mupuf: definitely!
04:35 karolherbst: who is on reator?
04:35 pmoreau: Not me
04:36 mupuf: check with w
04:39 karolherbst: there is a screen service running
04:40 karolherbst: which is attached okay
04:40 karolherbst: but w doesn't tell me much
04:40 mupuf: hakzsam: ?
04:40 hakzsam: yeah it's me
04:41 karolherbst: mupuf: lucky as we are, overvoltaging is a cli only thing :D
04:41 mupuf: gui would not have been a problem anyway
04:42 karolherbst: no, but gui over the web is a bit messy :/ I really don't like it
04:42 hakzsam: with vnc it's not really a problem
04:43 mupuf: nvidia-settings works greatr over ssh
04:43 karolherbst: mhh okay
04:43 karolherbst: I just had the feeling, that the connection went a little bad the last weeks
04:46 karolherbst: hakzsam: okay, then please tell me whe you are done, thanks :)
04:46 hakzsam: karolherbst, done for now, but I would like to use reator in few hours if possible :)
04:47 karolherbst: no problem
04:47 karolherbst: I just want to check some stuff
04:47 karolherbst: won't take me an hour
04:47 hakzsam: okay, go ahead
04:47 karolherbst: I have to reboot though :)
05:09 hakzsam: karolherbst, please, tell me when you are done, no rush :)
05:26 karolherbst: mupuf: blob still at 340 and there were no update in pacman (cuda was updated though=
05:26 mupuf: sudo pacman -S nvidia
05:26 mupuf: it will replace nvidia-340 which I probably used for its tesla support
05:27 karolherbst: "erreur : la préparation de la transaction a échoué (la satisfaction des dépendances a échoué"
05:27 karolherbst: ohh now I know what that means
05:27 karolherbst: :D
05:27 karolherbst: Supprimer means replace/remove?
05:28 karolherbst: ohh okay
05:28 karolherbst: I had to add nvidia-libgl
05:28 mupuf: right
05:28 karolherbst: okay, 352 loaded
05:28 karolherbst: nice
05:29 karolherbst: mhh okay, current voltage seems to be fixed at 1.2V
05:29 karolherbst: but
05:30 karolherbst: "The valid values for 'GPUOverVoltageOffset' are in the range 0 - 31250 (inclusive)."
05:30 karolherbst: seems to be 0.031V though :/
05:36 karolherbst: *sigh* :/
05:36 karolherbst: okay, this doesn't work on the maxwell card at least
05:40 karolherbst: wait a second....
05:40 karolherbst: ohh wrong card
05:41 karolherbst: mupuf: do you have a PWM based kepler card for reator? On my card I can't overvolt sadly :/
05:54 mupuf: karolherbst: no, I don't
05:56 karolherbst: mhh okay
06:01 mupuf: I guess they only expose the maximum voltage
06:01 mupuf: that's all
06:05 karolherbst: mhh
06:05 karolherbst: it is 0 for me though
06:18 pmoreau: I have a Maxwell (GM200) up and running with the blob, so I can check some regs if needed.
06:18 karolherbst: pmoreau: nvapeek 0x20344 && nvapeek 0x20340
06:18 karolherbst: and
06:18 karolherbst: nvidia-settings -query all | grep -i volt
06:19 karolherbst: if voltage is always 1200000, even if the peeks are different (depending on load) then mhh
06:21 pmoreau: nvidia-settings has to be run as root, right/
06:21 pmoreau: ?
06:21 karolherbst: no
06:22 pmoreau: I got this https://phabricator.pmoreau.org/P27
06:23 karolherbst: ohh nice
06:24 karolherbst: how much load is on the card?
06:24 karolherbst: maybe your card is usefull indeed
06:24 karolherbst: :D
06:24 pmoreau: It's a Titan X
06:24 karolherbst: ohh it figures
06:24 karolherbst: okay
06:24 pmoreau: Shouldn't be any load onit
06:24 karolherbst: that is much more usefull
06:24 karolherbst: mhhh
06:24 karolherbst: then the voltage is a tiny bit too hight
06:24 karolherbst: okay
06:25 karolherbst: try to run anything without vsync on the card
06:25 pmoreau: nvapeek incoming, needed to install envytools here. :)
06:25 pmoreau: 00020344: 00000028
06:25 pmoreau: 00020340: 0000006c
06:26 pmoreau: "0: 0000:01:00.0 ??? 120020a1" doesn't want to display the chipset :D
06:26 huehner: karolherbst: same from my gm206 maybe helpful: http://pastebin.com/jB4azpGK
06:27 karolherbst: mhhhh
06:27 karolherbst: these values are just too high for the reg values
06:27 karolherbst: 0x44 is the current pwm value
06:27 karolherbst: so I think it only shows the max voltage as mupuf said
06:27 karolherbst: okay, aother thing
06:27 karolherbst: please add the 16 bit to Coolbits
06:28 karolherbst: if you don't have any coolbits entry
06:28 karolherbst: "16" should be fine :)
06:28 karolherbst: it will enable the option to overvolt through nvidia-settings
06:28 karolherbst: though I didn'T see any changes on the 750 card
06:32 mupuf: pmoreau: please add support for the titan to envytools and nouveau :D
06:33 pmoreau: mupuf: :p Adding the chipset at least to envytools, so that nvalist correctly recognises it.
06:34 mupuf: yep
06:34 mupuf: 120 :)
06:34 karolherbst: pmoreau: does nvidia-smi support the titan?
06:35 pmoreau: mupuf: But, this is not my card strictly speaking, the group got it for performance measurement and use for the research, so I don't feel comfortable at all do to overclocking / reclocking / too many driver tests on it.
06:36 mupuf: pmoreau: agreed
06:36 pmoreau: karolherbst: It does return something :-)
06:36 karolherbst: pmoreau: mhh it alos return something for me, but nothing usefull
06:37 pmoreau: karolherbst: https://phabricator.pmoreau.org/P28
06:38 karolherbst: uhh
06:38 karolherbst: this is a bit more
06:39 karolherbst: nvidia-smi -q
06:39 pmoreau: mupuf: Feel like adding the chipset to the wiki page? I don't have my ids here. O:-)
06:39 hakzsam: which chipset?
06:40 pmoreau: NV120: GM200 -> Titan X
06:40 pmoreau: karolherbst: https://phabricator.pmoreau.org/P29
06:42 karolherbst: Max Clock seems to be boost
06:42 karolherbst: pmoreau: want to verify that with nvbios?
06:42 karolherbst: should be in the boost table
06:42 pmoreau: Sure
06:42 hakzsam: pmoreau, done
06:42 karolherbst: the clocks are x2 though
06:43 pmoreau: hakzsam: Thanks! :-)
06:43 karolherbst: pmoreau: the temperature part is also interessting
06:43 karolherbst: could help finding the right values out of the vbios?
06:43 karolherbst: or at least set them in nouveau
06:44 karolherbst: sadly there is no voltage :/
06:44 pmoreau: Attempt to extract the vbios from card 0 (nv120) using PROM.
06:44 pmoreau: Card has second bios
06:45 pmoreau: Really?! Two bioses?
06:45 karolherbst: pmoreau: nvidia-smi -q -u
06:46 karolherbst: pmoreau: one for linux, one for windows of course :p
06:46 pmoreau: O.O
06:46 pmoreau: What about OS X? :D
06:46 karolherbst: they have their own driver anyway
06:46 karolherbst: :p
06:46 pmoreau: They still need a VBIOS
06:46 pmoreau: HIC Info : N/A
06:47 pmoreau: Attached Units : 0
06:47 pmoreau: With the -u flag
06:47 karolherbst: :/
06:47 karolherbst: "For all S-class products." whatever they are
06:47 pmoreau: core-dumped on parsing the vbios
06:48 hakzsam: you need to fix that ;)
06:48 pmoreau: karolherbst: https://phabricator.pmoreau.org/F9979 the non-parsed VBIOS
06:49 pmoreau: hakzsam: Or hope for others to do it ;)
06:49 hakzsam: :)
06:49 kubast2: I have an gtx 650[GK107] ,is there some way I can contribute to the project[unfortunately ,I don't know programing languages/am not a driver developer]?
06:49 pmoreau: karolherbst: Got an unknown CSTEP, POWER BUDGET, and some others
06:50 karolherbst: kubast2: gddr5 card?
06:50 kubast2: yes
06:50 kubast2: 1gb
06:50 karolherbst: kubast2: test my patch
06:50 karolherbst: kubast2: are you using nouveau?
06:51 karolherbst: and tried 0f pstates and the card just crashed?
06:51 kubast2: no I'm using nvidia-355.11 on linux mint 17.2
06:51 karolherbst: ohhh
06:51 karolherbst: :/
06:51 kubast2: I can try to make bootable arch linux through
06:51 kubast2: *usb bootable
06:53 mupuf: hakzsam: thanks for adding the Titan to the list
06:53 hakzsam: np
06:54 kubast2: karolherbst[the nouveau version on linux mint 17.2 might be outdated]
06:56 kubast2: [1:1.0.10-1ubuntu2] it's installed by default and it worked fine for desktop/office usage as far as I can tell.
06:58 pmoreau: It's the parsing of the cstep which is failing
07:01 karolherbst: mupuf: should I update the nouveau partiton, too when I am at it?
07:01 karolherbst: then the blob can be loaded there too
07:02 mupuf: I am using a custom kernel
07:02 karolherbst: ohh okay
07:02 mupuf: so as we can compile nouveau
07:02 mupuf: the blob always lags behind
07:02 karolherbst: wait a second
07:02 karolherbst: you also have this soc/tegra issue?
07:02 mupuf: ?
07:02 mupuf:has no tegra yet
07:02 karolherbst: we found out yesterday or earlier, that soc/tegra headers are missing
07:02 karolherbst: and nouveau can't get compiled
07:03 karolherbst: on stock arch linux package
07:03 mupuf: yeah, that seems familiar
07:03 karolherbst: so you also have this problem
07:03 karolherbst: I see
07:03 karolherbst: though removing this includes works fine
07:03 karolherbst: *include
07:03 mupuf: well, nouveau needs to be compiled against drm-next
07:03 mupuf: so that's what I am running
07:03 karolherbst: okay
07:03 karolherbst: then I will leave it
07:04 mupuf: when nouveau will be rebased again on top of drm-next, I will update it again
07:04 karolherbst: hakzsam: you can do your thing now
07:05 hakzsam: karolherbst, thanks but I actually can't do my stuff right now (will do in 3/4 hours if you still want to use reator)
07:05 karolherbst: pmoreau: do you want to check with coolbits if changing voltage offset changes the pwm reg?
07:06 pmoreau: karolherbst: Don't feel really comfortable with it, sorry. :/
07:08 karolherbst: k
07:12 karolherbst: kubast2: mhhh changing to nouveau might be a bit messy on ubuntu machines though without changing packages
07:13 kubast2: I have just changed back to nouveau
07:13 kubast2: *I was disconnected for a little bit[reboot ,after chainging driver]
07:13 kubast2: *through it was through the official tool and I bet chainging it without deb packages will be messy just as you said
07:14 kubast2: karolherbst In other words I would need to: Clone the repo ,compile the driver and then manually install it
07:14 karolherbst: kubast2: can you add nouveau.pstate=1 to the kernel boot args?
07:15 karolherbst: and then reboot
07:15 karolherbst: there should be a /sys/class/drm/card0/pstate file then
07:15 karolherbst: if you want to help with reclocking that is
07:17 kubast2: wait I need to add GRUB_CMDLINE_LINUX="nouveau.pstate=1" ?
07:17 karolherbst: I think so, never used grub myself :)
07:20 pmoreau: Yes
07:21 pecisk: kubast2: yes, that's correct
07:21 pecisk: kubast2: and then rebuild configuration file for grub
07:22 kubast2: pecisk I have a UEFI install of mint ,so I'm not sure If after I change /etc/default/grub I'll need to apply it to both /boot/grub/grub.cfg and /boot/efi/grub.cfg
07:22 kubast2: with an grub-mkconfig command
07:23 kubast2: Adding boot menu entry for EFI firmware configuration
07:23 kubast2: done
07:23 kubast2: looks like I don't
07:23 kubast2: rebooting
07:24 kubast2: booted
07:25 kubast2: turns out /boot/efi/grub.cfg only points to /boot/grub/grub.cfg
07:25 pecisk: kubast2: you need to apply to one that will be used :) and I guess it is efi one
07:26 pecisk: ahhh
07:26 pecisk: makes sense
07:26 karolherbst: kubast2: so do you have that pstate file?
07:26 kubast2: I have applied it to /boot/grub/grub.cfg
07:27 kubast2: Kuba-Pc ~ # ls /sys/class/drm/card0/pstate
07:27 kubast2: ls: cannot access /sys/class/drm/card0/pstate: No such file or directory
07:27 pecisk: nouveau is loaded?
07:28 kubast2: yes
07:28 karolherbst: kubast2: cat /proc/cmdline
07:28 kubast2: BOOT_IMAGE=/boot/vmlinuz-3.16.0-38-generic root=UUID=8825e5b5-9491-4ff1-b034-42e10cf89608 ro nouveau.pstate=1 quiet splash vt.handoff=7
07:29 karolherbst: kubast2: lsmod | grep nouveau
07:29 kubast2: wait
07:29 kubast2: http://pastebin.com/0jHkPzUf
07:29 karolherbst: okay seems right
07:29 karolherbst: ohh 3.16
07:30 karolherbst: that's a tiny bit, ancient :/
07:30 karolherbst: it was different back then I think
07:31 kubast2: the newest in repo is 3.19.0-28
07:31 karolherbst: yeah, that one should be much better
07:31 kubast2: updating kernel then
07:34 kubast2: So it have just replaced kernel to newer one.
07:34 pmoreau: Will see tonight if I can figure out the layout used for CSTEP table in this version.
07:37 kubast2: nope still nothing.
07:37 pmoreau: Are you running the correct kernel? :) `uname -r`
07:37 kubast2: Linux Kuba-Pc 3.19.0-28-generic #30~14.04.1-Ubuntu SMP Tue Sep 1 09:32:55 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux
07:38 pmoreau: Do you have an Intel integrated GPU maybe?
07:38 kubast2: well I have
07:38 kubast2: but it never works
07:38 pmoreau: Look at card1 then
07:38 kubast2: when nvidia one is choosen as primary one
07:38 kubast2: it must be card0
07:38 kubast2: because my nvidia card have dvi hdmi and vga
07:39 kubast2: Kuba-Pc ~ # ls /sys/class/drm/card0
07:39 kubast2: card0-DVI-D-1 card0-VGA-1 device subsystem
07:39 kubast2: card0-HDMI-A-1 dev power uevent
07:39 kubast2: Kuba-Pc ~ # ls /sys/class/drm/card1
07:39 kubast2: ls: cannot access /sys/class/drm/card1: No such file or directory
07:39 pmoreau: Hum...
07:39 kubast2: My motherboard is only able to make one card at a time work
07:39 kubast2: even on w7/w10
07:39 pmoreau: Ok
07:40 pmoreau: Don't know why you don't get the file then...
07:40 pmoreau: You're running the command as root, right?
07:40 kubast2: yes
07:40 kubast2: hmm
07:41 kubast2: gonna try su root this time
07:41 kubast2: nope same
07:41 pmoreau: It's in card0/device! :D
07:42 kubast2: yes it is :D
07:42 pmoreau: karolherbst: :p
07:42 kubast2: ls /sys/class/drm/card0/device/pstate
07:42 kubast2: returns "/sys/class/drm/card0/device/pstate"
07:42 pmoreau: It's a file
07:42 pmoreau: You cat it to get the possible values
07:42 pmoreau: and echo the value to it to change to that state
07:42 kubast2: 07: core 324 MHz memory 648 MHz
07:42 kubast2: 0a: core 540 MHz memory 1620 MHz
07:42 kubast2: 0f: core 1058 MHz memory 5000 MHz
07:42 kubast2: AC: core 324 MHz memory 648 MHz
07:43 kubast2: let me just launch cs:go to test it out
07:43 pmoreau: like `echo 0a > /sys/class/...` to move to 0a perflvl
07:43 karolherbst: pmoreau: I said card1
07:43 karolherbst: *card0
07:43 kubast2: with nouveau without doing anything on bots it works at 20 fps
07:44 karolherbst: kubast2: go to 0a
07:44 karolherbst: it should be more then
07:44 pmoreau: karolherbst: Right, but you said card0/pstate, rather than card0/device/pstate ;)
07:44 karolherbst: ohhh
07:44 karolherbst: right
07:44 karolherbst: I don't use this old interface anymore anyway :D
07:44 kubast2: just did that
07:45 pmoreau: bbl
07:46 kubast2: it seems to affect the framerate[went with 0f]
07:46 karolherbst: kubast2: and did you get more performance?
07:46 kubast2: so it seems to be working
07:46 kubast2: doubled the performance
07:46 kubast2: from 20 fps
07:46 karolherbst: and 0f?
07:46 kubast2: to 40 fps
07:46 kubast2: 0f was 40 fps
07:46 karolherbst: ohh I see
07:46 kubast2: 07 is 20 fps
07:46 karolherbst: can you give me cat pstate
07:47 kubast2: 07: core 324 MHz memory 648 MHz AC DC *
07:47 kubast2: 0a: core 540 MHz memory 1620 MHz
07:47 kubast2: 0f: core 1058 MHz memory 5000 MHz
07:47 kubast2: AC: core 324 MHz memory 648 MHz
07:47 karolherbst: and how much is 0a?
07:47 AndChat|341361: So
07:48 karolherbst: you could try out to switch between 0a and 0f a few times until your desktop crashes ;)
07:48 AndChat|341361: I had to hard reset pc
07:48 karolherbst: :D
07:48 karolherbst: I see
07:48 kubast2: back
07:48 karolherbst: kubast2: 0f was messy? or what did happen
07:48 kubast2: After chainging pstate 3 times
07:48 kubast2: it rebooted
07:48 kubast2: 0f was working fine
07:48 kubast2: but after I chainged pstate from the lowest back to the biggest
07:48 karolherbst: it shouldn't reboot though
07:49 karolherbst: yeah
07:49 karolherbst: okay, then 0f is a bit unstable
07:49 kubast2: it went all black
07:49 karolherbst: okay, do you want to test a patch of mine, where it should never happen again?
07:49 kubast2: yeah sure
07:49 karolherbst: okay
07:49 karolherbst: do you have git installed?
07:50 kubast2: it's like 3346 KB
07:50 kubast2: I'll
07:50 kubast2: I will have to clone some repo ?
07:50 karolherbst: you need to install a custom nouveau module then, so it might be a bit difficult, but you can try
07:50 karolherbst: yes, mine
07:50 karolherbst: I hope it compiles against 3.19 though
07:51 kubast2: I also need build essentials right?
07:51 karolherbst: yeah
07:51 kubast2: "build-essential dpkg-dev g++ g++-4.8 libstdc++-4.8-dev" Any other required packages
07:52 karolherbst: build-essnetials should be enough
07:52 karolherbst: you don't need the c++ parts
07:52 kubast2: I know but "apt-get install build-essential" have added them
07:52 karolherbst: okay
07:52 karolherbst: then it is fine
07:53 karolherbst: go somewhere where you want to have the repository and do "git clone -b gddr5 https://github.com/karolherbst/nouveau.git"
07:53 kubast2: got it
07:53 kubast2: doing make -j4
07:53 karolherbst: wait
07:53 karolherbst: cd into drm
07:53 kubast2: yes
07:53 karolherbst: then do make
07:54 kubast2: at the moment no make errors
07:54 karolherbst: it will be in the last files somehwere if it won't work
07:55 kubast2: karolherbst can I add 2>/Errors.txt ?
07:55 kubast2: to catch the compiling errors
07:55 karolherbst: no, doesn't matter
07:55 karolherbst: if it won't work with 3.19 then you need a newer kernel, which is a little messy
07:56 kubast2: so the compiling is done
07:57 kubast2: and there was something about fbcon.c
07:57 kubast2: http://pastebin.com/EPJ4TQ0H
07:57 kubast2: the module was made through
07:57 kubast2: /home/kubast/nouveau/nouveau/drm/nouveau/nouveau.ko
07:58 karolherbst: wait
07:58 karolherbst: I didn't tough that fle, but it shouldn't matter
07:58 karolherbst: this is pretty harmless
07:58 karolherbst: okay
07:58 karolherbst: mhhh
07:59 karolherbst: now you have to find nouveau.ko(.xz) inside /lib/modules
07:59 karolherbst: or wherever the kernel modules are installed on your system
07:59 karolherbst: then you have to rename the installed module to something with .old at the and
08:00 karolherbst: and copy the compiled moudule where the isntalled on is
08:00 kubast2: wait gonna use grep + recursive ls
08:00 karolherbst: mhhh
08:00 karolherbst: find
08:00 karolherbst: find /lib/modules -iname "nouveau*"
08:00 kubast2: /lib/modules/3.19.0-28-generic/kernel/drivers/gpu/drm/nouveau:
08:01 karolherbst: this is the source folder
08:01 karolherbst: you should try with find
08:02 kubast2: cd /lib/modules/3.19.0-28-generic/kernel/drivers/gpu/drm/nouveau && mv nouveau.ko nouveau.ko.old
08:02 kubast2: /home/kubast/nouveau/nouveau/drm/nouveau/nouveau.ko ./nouveau.ko
08:02 kubast2: did that
08:02 karolherbst: ohh okay
08:02 karolherbst: should be fine then
08:03 karolherbst: try to reboot
08:03 AndChat|341361: Booted
08:04 kubast2: gonna now test the patch
08:04 karolherbst: mhh
08:04 karolherbst: after you echo 0f into pstate, please cat pstate
08:04 karolherbst: and give me the last line
08:05 kubast2: 07: core 324 MHz memory 648 MHz
08:05 kubast2: 0a: core 540 MHz memory 1620 MHz
08:05 kubast2: 0f: core 1058 MHz memory 5000 MHz AC DC *
08:05 kubast2: AC: core 324 MHz memory 5022 MHz
08:05 karolherbst: mhh
08:05 karolherbst: this is too high for some reasons
08:05 kubast2: the AC one
08:06 karolherbst: yeah
08:06 karolherbst: usually that means, my module wasn't loaded
08:06 karolherbst: please do dmesg | grep -i nouveau
08:06 kubast2: 07: core 324 MHz memory 648 MHz
08:06 kubast2: 0a: core 540 MHz memory 1620 MHz
08:06 kubast2: 0f: core 1058 MHz memory 5000 MHz
08:06 kubast2: AC: core 324 MHz memory 648 MHz
08:06 kubast2: the other one
08:06 kubast2: before applying 0f
08:07 karolherbst: this is fine, usually with my patch applies it should be really close to 5000
08:07 kubast2: http://pastebin.com/amBrESWQ
08:07 kubast2: [ 129.065970] nouveau E[ CLK][0000:01:00.0] failed to raise voltage: -22
08:07 kubast2: [ 129.065975] nouveau E[ CLK][0000:01:00.0] error setting pstate 2: -22
08:07 karolherbst: yeah, the old version is loaded
08:07 karolherbst: it should say nouveau 1.3
08:07 karolherbst: ohh right
08:08 karolherbst: you need to update your initram fs
08:08 kubast2: how can I do that
08:08 karolherbst: sadly, I never used one, sooo I have to guess how you do that :D
08:08 karolherbst: update-initramfs maybe
08:08 kubast2: well I have been once modifying one ,but on android device
08:08 karolherbst: or mkinitramfs
08:08 kubast2: well there is update-initramfs
08:09 kubast2: update an existing initramfs
08:09 karolherbst: did it work running as root?
08:09 kubast2: "-u"
08:09 kubast2: all the time
08:09 kubast2: su root
08:09 karolherbst: did you run update-initramfs ?
08:09 kubast2: *I'm aware at the moment what most of the commands do
08:09 kubast2: yes it's running now
08:09 karolherbst: okay
08:10 karolherbst: so it seems to run fine without any args
08:10 kubast2: it done it's work
08:10 karolherbst: if it is done, then reboot :)
08:11 kubast2: [drm] Initialized nouveau 1.3.0
08:12 karolherbst: :)
08:12 karolherbst: much better
08:12 karolherbst: okay, now go into 0f pstate
08:12 karolherbst: if the mem clock is 6001 or 5999 or 6000 it is good
08:12 kubast2: mine memory is 500 mhz
08:12 kubast2: 07: core 324 MHz memory 648 MHz
08:12 kubast2: 0a: core 540 MHz memory 1620 MHz
08:12 kubast2: 0f: core 1058 MHz memory 5000 MHz AC DC *
08:12 kubast2: AC: core 324 MHz memory 5000 MHz
08:13 kubast2: *5000mhz
08:13 karolherbst: okay
08:13 karolherbst: now run something at full load and try to switch a few times
08:13 karolherbst: when it doesn't crash, it works
08:13 karolherbst: when the screen gets corrupted, then just switch between a lower pstate and 0f again
08:13 karolherbst: until it looks clean again :D
08:13 kubast2: launching cs:go
08:14 AndChat|341361: Pstates 07 0a 0f
08:15 kubast2: made it go full load
08:16 karolherbst: kubast2: then switch until something bad happens :)
08:17 kubast2: nothing bad happens at themo ment
08:18 karolherbst: kubast2: so if you want, you can use nouveau a little bit at 0f for a while
08:18 kubast2: 64 times
08:18 karolherbst: and if anything comes up, just complain
08:18 karolherbst: but it seems to work better than before :D
08:19 karolherbst: only your core clock doesn't go up at 0f, which is a problem we have to take care off now
08:19 karolherbst: :/
08:19 karolherbst: kubast2: switching to 0a first should give you 540 MHz core on 0f right?
08:20 kubast2: 07: core 324 MHz memory 648 MHz
08:20 kubast2: 0a: core 540 MHz memory 1620 MHz AC DC *
08:20 kubast2: 0f: core 1058 MHz memory 5000 MHz
08:20 kubast2: AC: core 324 MHz memory 1620 MHz
08:20 kubast2: yup
08:20 kubast2: 45 fps on 0a
08:21 kubast2: 55 fps on 0f
08:21 karolherbst: mupuf: I really would like to know what goes wrong with the old PLL setup on a hardware level :/
08:21 karolherbst: because I currently don't understand why we have to respect these bounds
08:22 karolherbst: kubast2: mhh +25% only because of memory clock isn't that great, but it is something :)
08:22 mupuf: karolherbst: can you elaborate on your question?
08:23 karolherbst: don't know, how can I do that?
08:23 mupuf: I do not understand your question
08:23 karolherbst: ohh okay
08:23 mupuf: what do you mean by old PLL setup?
08:23 mupuf: are you talking about the pll limit table?
08:23 karolherbst: the issue with stock ouveau is, that for the high gddr5 clocks (above 2.4GHz) there comes a second PLL into play
08:23 karolherbst: and this second PLL was configured wrongly
08:23 karolherbst: like M has to be always 1
08:24 karolherbst: but nouveau also used 2 or 4
08:24 mupuf: right, your patch fixes that
08:24 karolherbst: and I don't know why this is bad (to use 2 or 4)
08:24 karolherbst: yeah
08:24 karolherbst: the lower pll doesn't seem to care what it has, it seems always stable
08:24 karolherbst: tried with 405MHz reflcok once, and it was fine
08:24 mupuf: maybe we wrongly assume this is a PLL, instead of a DLL
08:25 mupuf: DLLs can only lower the frequency
08:25 karolherbst: lower?
08:25 mupuf: have you tried setting M to 3 and D to 4?
08:26 mupuf: lower as in, produce an output freq lower than the input freq
08:26 karolherbst: what is D? :D
08:26 mupuf: divider
08:26 mupuf: N
08:26 karolherbst: ohh no, the output should be higher anyway
08:26 karolherbst: the input freq with my patch is something around 200MHz
08:26 mupuf: M is the divider or the multiplicator?
08:26 karolherbst: mhh post devider?
08:26 karolherbst: I don't know
08:26 karolherbst: it seems to be the same as P
08:27 karolherbst: since gt215
08:27 mupuf: is it?
08:27 karolherbst: it is handled the same way
08:27 mupuf: hmm, I need to have a look at this code again
08:27 karolherbst: / (M*P)
08:27 karolherbst: output = ((clk * N) + (((u16)(fN + 4096) * input) >> 13)) / (M * P);
08:28 mupuf: I seem to remember the pll formula to be output = (input * n / m) >> p;
08:28 karolherbst: P has to be 1 or 2 by the way
08:28 karolherbst: so we basically use the second PLL to do 0.5x steps
08:28 karolherbst: except I use P=1 for everything because it is simplier
08:28 karolherbst: though I have an experimental patch for adding P=2 as an option
08:28 mupuf: ack
08:29 karolherbst: doesn't change much though
08:29 karolherbst: except we would be a bit closer to the blob in some cases
08:29 karolherbst: the blob has a tolerance of around 6MHz
08:31 karolherbst: anyway, I really would like to know if there is a PLL limit somewhere for this
08:31 karolherbst: and I would like to know why we can only use the second PLL for x0.5a multiplications
08:32 karolherbst: this seems to be the "right" entry: "-- VCO1 - freq [2000-4000]MHz, inputfreq [150-250]MHz, M [1-255], N [8-255], P [1-63] --"
08:32 karolherbst: but
08:33 karolherbst: well it doesn't help
08:33 mupuf: well, that indeed sucks that the limit table does not work for this
08:33 kubast2: To be honest in original nvidia drivers there are 3 states. 0[=07]/1[07-0a]/2[0f].
08:33 karolherbst: kubast2: yeah
08:33 karolherbst: I know
08:34 karolherbst: mupuf: right, so I had to hardcode those limits
08:36 karolherbst: sadly tegra source doesn't help here :D
08:37 mupuf: I wonder why :D
08:37 hakzsam: karolherbst, try to ask NVIDIA :p
08:38 karolherbst: I think this is just a special PLL, because it has to putput pretty high freqs and technically there had to be some restrictions
08:44 daemon32: karolherbst: Why do I see things like "gpc: 2404000" in my cstates? Does that really mean 2.404 GHz, or am I reading that wrong?
08:44 karolherbst: daemon32: you have to devide by 2
08:44 kubast2: 1.202 ghz ?
08:44 karolherbst: yeah
08:44 kubast2: seems right
08:45 mlankhorst: /5
08:45 mlankhorst: oops
08:47 daemon32: Ohh, thanks xD
08:48 daemon32: karolherbst: And I didn't see that other branch you had until yesterday (the one with the cstates) so that's why I was wondering on reddit
08:50 karolherbst: daemon32: ohh right
08:50 karolherbst: there is a "master_karol_stable" branch now
08:50 karolherbst: this is usually safe to use
08:51 karolherbst: but can contain outdated patches
08:51 kubast2: https://github.com/karolherbst/nouveau/tree/master_karol_stable
08:51 kubast2: https://github.com/karolherbst/nouveau/tree/master_karol
08:52 karolherbst: the first one, but well
08:53 kubast2: the other one contains experimental patches?
08:53 karolherbst: everything contains experimental patches
08:53 karolherbst: but with the one branch you won't do something bad for your card
09:01 karolherbst: kubast2: but if you want you can test out the branch, it should add pcie speed changes and you could test if you get more perf on 0f, than with the gddr5 brach
09:01 karolherbst: *branch
09:02 kubast2: sure thing
09:03 daemon32: I'll test it now :)
09:04 karolherbst: allthough when you install envytools you can also switch the speed on the fly
09:05 karolherbst: I doubt though that pcie speed changes performance on a desktop system much
09:05 kubast2: module compiled 2 "warnings" one in nouveau_fbcon.c and the other nouveau_debugfs.c
09:05 daemon32: Does that have to be git, or will a release version work?
09:05 karolherbst: daemon32: it has to be one of my branches
09:05 karolherbst: kubast2: this is fine
09:06 daemon32: karolherbst: pcie2_reg?
09:06 karolherbst: ohh envytools
09:06 karolherbst: no, there master is fine
09:06 daemon32: Oh, okay xD
09:08 kubast2: rebooting
09:10 kubast2: [ 2.313235] [drm] Initialized nouveau 1.3.0 20120801
09:10 karolherbst: kubast2: the pstate file location changed though
09:11 karolherbst: it is in /sys/kernel/debug/dri/0/ now
09:11 karolherbst: then go into pstate 0f
09:11 karolherbst: and cat cstate
09:12 kubast2: well
09:12 kubast2: the previous module have been launched
09:12 kubast2: Kuba-Pc kubast # ls /sys/class/drm/card0/device/pstate
09:12 kubast2: Kuba-Pc kubast # ls /sys/kernel/debug/dri/0
09:12 kubast2: bufs clients DVI-D-1 gem_names HDMI-A-1 name vbios.rom VGA-1 vm vma
09:12 kubast2: Kuba-Pc kubast # cat /sys/kernel/debug/dri/0/cstate
09:12 kubast2: cat: /sys/kernel/debug/dri/0/cstate: No such file or directory
09:12 karolherbst: you need to update initramfs :)
09:12 kubast2: right
09:13 kubast2: I think I'll just make bash script to do some dirty work for me xd
09:14 kubast: maybe not dirty
09:14 kubast: but things I forget
09:16 karolherbst: if you get into pstate 0f, please cat cstate and run "lspci -s 01:00.0 -vv | grep LnkSta"
09:16 daemon32: karolherbst: Will do
09:16 kubast2: oh ok
09:17 kubast2: pstate is near the cstate
09:17 kubast2: current pstate: 0x0f
09:17 kubast2: available clockings: 1058 - 1058 MHz
09:17 kubast2: available cstates for current pstate:
09:17 kubast2: 0: {gpc: 2117000, mem: 2500000, voltage: 4}
09:17 kubast2: current core: 648000
09:17 kubast2: current memory: 2500000
09:17 kubast2: LnkSta: Speed 8GT/s, Width x16, TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-
09:17 kubast2: LnkSta2: Current De-emphasis Level: -3.5dB, EqualizationComplete+, EqualizationPhase1+
09:18 karolherbst: :O
09:18 karolherbst: what the heck
09:18 karolherbst: one cstate for 0f pstate?
09:18 karolherbst: mhhhhh
09:18 karolherbst: kubast2: please upload the vbios.rom file somewhere
09:19 kubast2: will do
09:19 karolherbst: I think you have to cat it somewhere first though
09:19 karolherbst: cat vbios.rom > /tmp/vbios.rom
09:19 kubast2: it's in /sys/kernel/debug/dri/0/
09:19 karolherbst: yeah
09:19 karolherbst: but it is no "file" in the way you think it is
09:19 karolherbst: I don't think you can read this one
09:19 kubast2: I know
09:19 karolherbst: okay :)
09:20 kubast2: it's binary file
09:20 karolherbst: that's not what I meant
09:21 kubast2: *and that it's from /sys
09:21 kubast2: so it's not a file I can cp
09:21 kubast2: http://www11.zippyshare.com/v/OFZGaIYP/file.html
09:22 karolherbst: wow
09:23 karolherbst: mupuf: that is the entire cstate table: https://gist.github.com/karolherbst/b0e9302f931fa3bdad28
09:23 karolherbst: GK107
09:23 karolherbst: and 8 entries in teh voltage map table
09:23 mupuf: hmm
09:24 karolherbst: what a card
09:24 karolherbst: I will upload the vbios
09:24 mupuf: along with the strap peek!
09:24 karolherbst: which was the reg again?
09:25 mupuf: 10100
09:25 mupuf: 101000
09:25 karolherbst: kubast2: please do nvapeek 101000
09:25 kubast2: now ok
09:25 kubast2: nvapeek command not found
09:25 kubast2: what I need to install
09:25 karolherbst: ohh you need envytools for that
09:25 kubast2: E: Nie udało się odnaleźć pakietu envytools
09:25 kubast2: ppa deb package ?
09:26 hakzsam: karolherbst, are you using reator?
09:26 kubast2: or the source code ?
09:26 karolherbst: hakzsam: nope
09:26 karolherbst: kubast2: you can use the code
09:26 karolherbst: it is pretty easy to compile
09:26 kubast2: https://github.com/envytools/envytools/ ?
09:26 karolherbst: yes
09:27 kubast2: it will take awhile
09:27 kubast2: my internet speed dropped
09:27 kubast2: from 700 kbps to 30
09:28 karolherbst: --depth 1
09:28 karolherbst: add this to your git clone command
09:29 kubast2: downloaded
09:29 karolherbst: mkdir build
09:29 kubast2: this one uses cmake
09:29 karolherbst: cd build
09:29 karolherbst: cmake ..
09:29 karolherbst: yeah
09:30 kubast2: doing cmake "errors occurred!"
09:30 karolherbst: it may be that you need some dev packages
09:32 kubast2: http://pastebin.com/rGYRsaD6 Cmake error file
09:32 karolherbst: glibc-dev
09:32 kubast2: E:
09:33 kubast2: need to compile it too
09:33 kubast2: or find ppa with glic-dev
09:33 karolherbst: no
09:33 kubast2: perhaps it's called diffrentlly
09:33 karolherbst: I just don't know the package name
09:33 kubast2: libc
09:33 slock83: already in your software sources
09:33 karolherbst: libc6-dev
09:33 kubast2: libc6 or something like that
09:33 slock83: libc6-dev
09:33 kubast2: it's allready installed
09:34 kubast2: libc6-dev is already the newest version.
09:34 karolherbst: yeah, because cmake should use libpthread.so
09:34 karolherbst: not pthreads
09:34 karolherbst: o
09:35 karolherbst: the error file won't help
09:35 karolherbst: please just give the command output
09:35 kubast2: of cmake
09:35 karolherbst: yes
09:35 kubast2: http://pastebin.com/mypH4kVF
09:36 karolherbst: install libxml2-dev
09:36 kubast2: done
09:36 kubast2: cleaning cmake output and retrying
09:37 karolherbst: you don't neet do clean
09:38 kubast2: so builds is done
09:38 karolherbst: there is nva/nvapeek
09:38 karolherbst: mhh
09:38 karolherbst: yes
09:38 kubast2: no
09:38 kubast2: wait
09:38 kubast2: Warning: nva won't be built because of un-met dependencies (pciaccess)
09:39 karolherbst: install libpciaccess-dev
09:39 karolherbst: rerun cmake and make
09:40 kubast2: CMakeFiles cmake_install.cmake CTestTestfile.cmake Makefile
09:41 karolherbst: ?
09:41 kubast2: wait
09:41 kubast2: http://pastebin.com/v8UVJZNi
09:41 karolherbst: this looks nice though
09:42 karolherbst: nva should build
09:42 kubast2: wait
09:42 kubast2: oh
09:42 hakzsam: mupuf, I checked all of those regs, the blob performs writes instead of masks, you were right
09:42 kubast2: cmake && make
09:43 kubast2: *facepalms hard
09:43 kubast2: Kuba-Pc nva # ls | grep nvapeak
09:43 kubast2: Kuba-Pc nva #
09:43 kubast2: nvm
09:43 karolherbst: nvapeek
09:43 karolherbst: :)
09:44 kubast2: 00101000: 8040989a
09:44 hakzsam: mupuf, actually no, sorry, it uses masks for some regs and writes for others
09:45 mupuf: hakzsam: good that I asked you to check ;)
09:45 hakzsam: sure
09:45 karolherbst: mupuf: pushed
09:45 karolherbst: wow this vbios is really funny
09:45 karolherbst: mupuf: you will like this most: "-- Mode GPIO, Base voltage 0 µV, voltage step 0 µV, acceptable range [0, 0] µV --"
09:46 mupuf: yeah baby, yeah!
09:46 karolherbst: kubast2: you should give mupuf your card :p
09:47 kubast2: it propablly was one of the cheapest gtxs 650[bought a whole computer e years ago instead of building it]
09:47 kubast2: it's an gainward gtx 650 non-ti
09:47 hakzsam: mupuf, http://hastebin.com/iqimizoded.md
09:47 karolherbst: kubast2: yeah, but this is really unusual
09:47 karolherbst: even the VID gpios are not ordered
09:47 karolherbst: what a mess
09:48 karolherbst: one really cared about this card
09:48 karolherbst: ohh unkonw con
09:48 slock83: anything I can do to help ? got a gtx780m (with optimus...), and can do things both on 'dos and linux
09:48 karolherbst: "HDMI-C (Mini) connector"
09:49 kubast2: yeah
09:49 karolherbst: this is totally missing in envytools
09:49 karolherbst: kubast2: does it work with nouveau?
09:49 karolherbst: could you run xrandr
09:49 kubast2: the card
09:49 kubast2: I connect through vga
09:49 karolherbst: and check if there is HDMI
09:49 kubast2: Screen 0: minimum 320 x 200, current 1920 x 1080, maximum 8192 x 8192
09:49 kubast2: VGA-1 connected 1920x1080+0+0 (normal left inverted right x axis y axis) 480mm x 268mm
09:49 kubast2: 1920x1080 60.0*+
09:49 kubast2: 1680x1050 60.0
09:49 kubast2: 1280x1024 75.0 60.0
09:49 kubast2: 1440x900 75.0 59.9
09:49 kubast2: 1024x768 75.1 60.0
09:49 kubast2: 800x600 75.0 60.3
09:49 kubast2: 640x480 75.0 72.8 66.7 60.0
09:49 kubast2: 720x400 70.1
09:49 kubast2: DVI-D-1 disconnected (normal left inverted right x axis y axis)
09:49 kubast2: HDMI-1 disconnected (normal left inverted right x axis y axis)
09:49 karolherbst: kubast2: please don't post such commands in the future ;)
09:49 kubast2: there is
09:50 kubast2: okey
09:50 karolherbst: yeah okay, seems like it is in the kernel, but not in envytools
09:50 karolherbst: a few lines are fine, but never more than 5 :D
09:50 karolherbst: slock83: it depends
09:51 karolherbst: slock83: 780m is GK104 right?
09:52 slock83: http://nouveau.freedesktop.org/wiki/CodeNames/ seems to say this
09:52 kubast2: according to wiki
09:52 kubast2: yeah
09:52 karolherbst: it is fine, I don't think we can much do for your card if everything works
09:52 kubast2: *wikipedia
09:52 karolherbst: so your card is mostly in a state that most of the stuff should work out of the box
09:52 karolherbst: except 0f pstate, but this will be done in the near future
09:53 karolherbst: slock83: so if you have any problems, then ask about them
09:54 slock83: Last time I tried it with nouveau, it turned my laptop into a frying pan, but it might be because I was running a rather old version (the one bundled with Ubuntu 14.04)
09:55 kubast2: karolherbst tbh even using nvidia-smi -q with original drivers didn't show the frequencies xd
09:55 karolherbst: slock83: could be
09:55 karolherbst: slock83: I have a 770m and this works pretty nicely
09:56 karolherbst: kubast2: this is fine, I don't care about nvidia-smi except on cards where it should work, like quadro cards
09:57 slock83: well I'll try with the latest version :) The only things that matter for me on linux is battery life, cuda, and external display, I'm gonna check that
09:57 karolherbst: slock83: mhh okay
09:57 karolherbst: then you should stick with nvidia
09:58 hakzsam: mupuf, the gf117 case was good, the blob uses masks
10:00 slock83: ok, thanks !
10:00 mupuf: hakzsam: ok
10:00 karolherbst: except you especially want to help nouveua, but cuda will be most likely never be done
10:01 mupuf: karolherbst: to some extent, we already have cuda support
10:01 karolherbst: mupuf: really? I thought cuda is like closed closed
10:01 mupuf: right
10:01 imirkin: mupuf: to the extent that you can use nvidia tools to generate kernels?
10:01 mupuf: yes :D
10:02 hakzsam: cuda requires compute support and we have it for some chipsets ;)
10:02 karolherbst: ohh okay
10:02 karolherbst: yeah well
10:02 karolherbst: since kepler the cores are the same anyway
10:02 mupuf: hakzsam: we have it in mesa, but there is another project outside of mesa that can run kernels generated by cuda
10:02 karolherbst: so I am not suprised, that you can push binaries into the card and it does something usefull
10:02 hakzsam: yeah, gdev
10:02 mupuf: yep
10:03 hakzsam: I used it to implement compute support for fermi, do you remember? :)
10:03 karolherbst: mupuf: how is the perf this way?
10:03 karolherbst: mhh someone could use the nvidia tools to compile glsl code :D
10:03 karolherbst: but I think this is not that easy
10:04 mupuf: karolherbst: you need to read the articles about gde
10:04 mupuf: v
10:04 mupuf: there were some nice performance values
10:05 mupuf: have to go, see you
10:06 karolherbst: cya
10:11 slock83: well, seeing this I think I will switch back to nouveau ! one last question : How is the battery life compared to nvidia drivers ?
10:11 imirkin: worse.
10:12 imirkin: pretty much everything about nouveau is worse than the proprietary driver
10:13 imirkin: but it's (a) open, (b) plays nice with the rest of the open-source stack
10:14 RSpliet: imirkin: security wise nouveau might have an edge though
10:14 imirkin: RSpliet: sorta. hard to review the blob's security. but you know for sure that nouveau is insecure ;)
10:16 imirkin: first off, nouveau is so easy to use for a DoS, secondly it hands back non-zero'd chunks of vram to random processes
10:17 karolherbst: imirkin: any idea how to fix the latter except 0ing buffers?
10:18 imirkin: you could do the zero'ing ahead of time
10:18 karolherbst: at free time then
10:18 karolherbst: mhhh
10:18 imirkin: e.g. you could keep a pool of N pages of zero'd memory
10:18 imirkin: and if that gets too low, then you zero some more
10:18 karolherbst: or zero everything else when the card doesn't do anything
10:18 imirkin: so... not exactly at free time
10:18 imirkin: but not at alloc time either
10:20 RSpliet: can't we catch this with zero-copy fills?
10:25 karolherbst: RSpliet: I think performance is a problem
10:26 karolherbst: RSpliet: ohh zero-copy?
10:26 karolherbst: how can you do that
10:37 RSpliet: I think the 10000ft idea is to mark a memory block as empty in the caches
10:37 RSpliet: I don't know what the requirements are for such a memory region though
10:39 karolherbst: I don't know if that effects every region
10:39 karolherbst: but everything which was used as framebuffer should be clreared
10:39 karolherbst: so
10:40 karolherbst: but the blob has the same issue afaik
10:41 karolherbst: ...
10:41 karolherbst: didn't notice that you were gone :D
10:42 karolherbst: "but everything which was used as framebuffer should be clreared" and " but the blob has the same issue afaik"
11:02 daemon32: karolherbst: LnkSta:Speed 5GT/s, Width x8, TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-
11:02 daemon32: karolherbst: What was it that I was supposed to do with envytools again?
11:03 imirkin_: daemon32: is it an x8 slot?
11:05 daemon32: imirkin_: Nope, it's a PCIe 2.0 x16 slot
11:05 imirkin_: hmmmm... odd!
11:05 imirkin_: do you have 2 gpu's perhaps? sometimes that causes the lanes to be split
11:05 imirkin_: or is it a GK208, in which case it only has 8 lanes to go around in the first place
11:06 kubast2: any russian speaker here ?
11:06 daemon32: imirkin_: Nope, it's on the dedicated 16 lane slot from my CPU and I have only PCI cards in the machine
11:06 daemon32: imirkin_: It's a GK104 I think
11:07 daemon32: imirkin_: Yeah, GK104
11:07 imirkin_: karolherbst: --^ very odd, looks like it comes up as x8, so makes sense to upgrade to x16
11:07 karolherbst: ohhohohoh :O
11:07 karolherbst: daemon32: vbios.rom please
11:07 karolherbst: okay this is good
11:07 karolherbst: I can work with that
11:08 daemon32: karolherbst:
11:08 karolherbst: daemon32: nvapeek 0x8c080
11:08 imirkin_: daemon32: this isn't some funny laptop right?
11:09 karolherbst: daemon32: connection refused
11:09 karolherbst: daemon32: and nvapeek 0x101000
11:10 daemon32: karolherbst: Sorry, just switched to OpenRC from systemd :P
11:10 karolherbst: daemon32: no worries
11:11 daemon32: karolherbst: 00101000: 80405c9a
11:11 daemon32: karolherbst: It should be up now, unless I screwed something else up
11:12 karolherbst: its fine
11:12 karolherbst: daemon32: what card is it?
11:12 daemon32: It's a GTX 760, if you want the specific model, I'll go look
11:13 karolherbst: no it is fine
11:13 karolherbst: this should be indeed a x8 card
11:13 karolherbst: *x16
11:13 karolherbst: what is wrong
11:13 karolherbst: okay
11:13 karolherbst: anyay
11:13 karolherbst: nvapeek 0x8c080 please
11:14 daemon32: karolherbst: 0008c080: 00000808
11:14 karolherbst: okay
11:14 karolherbst: mhhh
11:14 karolherbst: imirkin_: this reg is bothering me :D
11:15 imirkin_: didn't we figure out how to change lane widths? or was that only for tesla/fermi?
11:15 karolherbst: daemon32: are you able to run the blob?
11:15 karolherbst: imirkin_: tesla yes
11:15 daemon32: karolherbst: Sure
11:15 karolherbst: on kepler the reg is gone
11:15 imirkin_: karolherbst: and fermi too then
11:15 karolherbst: daemon32: okay, then boot with blob and open nvidia-settings
11:15 karolherbst: imirkin_: I guess so
11:15 daemon32: karolherbst: Alrighty
11:15 imirkin_: daemon32: first off, check that blob upgrades it to x16
11:15 imirkin_: if not, there's some sort of physical issue in your system
11:16 imirkin_: if it does, then on a *clean boot*, grab a mmiotrace of your board getting initialized by nvidia
11:16 daemon32: imirkin_: Will do (just in case) but I swear that it was x16 lane in windows anyway
11:16 karolherbst: yeah
11:16 karolherbst: if the blob upgrades the card to x16, we need you
11:16 karolherbst: :)
11:16 daemon32: Alright xD
11:16 karolherbst: or your card
11:16 imirkin_: probably mobo + card combo
11:17 karolherbst: yeah
11:17 karolherbst: but this gives me less clue what this reg is about
11:17 karolherbst: it does show the width
11:17 karolherbst: but
11:17 karolherbst: why twice
11:17 karolherbst: both values are always the same
11:17 karolherbst: and this reg is read only
11:17 imirkin_: for luck :)
11:21 karolherbst: imirkin_: I found a note of mine: "0x08814c maybe limits?"
11:22 karolherbst: 0x88140 is the width reg which is empty here
11:23 LordShadowWing: karolherbst, are you by chance related to the Erbst Family?
11:25 karolherbst: no clue
11:25 karolherbst: who are they?
11:33 daemon32: karolherbst: False alarm, apparently it's stuck at x8
11:34 daemon32: But I have no idea why o.o
11:35 karolherbst: me neither
11:35 karolherbst: daemon32: try to add some load
11:35 karolherbst: ohh what is the maximum pcie link width in nvidia-settings?
11:36 LordShadowWing: Karolherbst: Erbst happens to be my Last name
11:37 karolherbst: LordShadowWing: ohh okay, don*t know maybe, maybe not, but the name is so generic, that I highly doubt that there was anything special :D
11:38 karolherbst: daemon32: your gpu is the only pcie card there?
11:40 LordShadowWing: Karolherbst: http://www.geni.com/surnames/people/erbst
11:41 kubast2: "karolherbst> kubast2: you should give mupuf your card :p" Heck ,I would[but first I would need the money to buy one for myself xd].
11:43 NewtoM: Hey
11:44 karolherbst: daemon32: so, anything new?
11:44 NewtoM: Nouveau seems to have problems using my (well, pretty old, but cool :]) Geforce4 420 Go GPU. Does anyone happen to have information on that? Should that work?
11:44 imirkin_: NewtoM: is that a NV28M?
11:44 imirkin_: NewtoM: lspci -nn -d 10de:
11:45 NewtoM: that's NV17M
11:45 imirkin_: NewtoM: what sort of problem? afaik that should be fine...
11:46 imirkin_: the 3d driver leaves something to be desired, but it does in general work
11:46 NewtoM: lspci returned: "NVIDIA Corporation NV17M [GeForce4 420 Go] [10de:0175] (rev a3)
11:46 NewtoM: well, my system is a Elementary OS Luna 0.3.1
11:47 karolherbst: NewtoM: yeah, but what is the issue you have?
11:47 NewtoM: The system installs and starts OK. Then it comes to the (graphical) log-in screen.
11:48 karolherbst: and?
11:48 NewtoM: When I type user and pass and hit enter, the graphic mode changes several times, then I see horizontal lines, then mode changes again, for split second something like a destop, and then back to the log-in screen (which does not use the native resolution)
11:48 karolherbst: mhh
11:48 karolherbst: there should be X logs in /var/log
11:48 NewtoM: I do not even know where to look for a log
11:48 NewtoM: mkey
11:48 karolherbst: I think you need Xorg.0.log.old
11:49 karolherbst: because the X server restarts
11:49 karolherbst: but this sounds like some issue of your DE
11:49 NewtoM: DE? (Are you guys still here in 30 minutes?)
11:49 karolherbst: deskopt environment
11:49 imirkin_: NewtoM: sounds like your userspace does something odd. please provide dmesg and xorg logs
11:50 NewtoM: (Freya is being reinstalled in the moment. Then we can go to the depths of FRESH logs.)
11:50 karolherbst: NewtoM: why not the old one?
11:50 karolherbst: ohh mhhh
11:50 karolherbst: well
11:50 karolherbst: please do not reinstall before seeking help in the future :D
11:50 NewtoM: because I was so frustrated that I already started a reinstall ...
11:50 NewtoM: OK, I got it
11:50 NewtoM: actually, I was seeking for help
11:50 NewtoM: but not here ...
11:51 karolherbst: because that just removes a faulty configuration, we may should debug and add fixes for that
11:51 NewtoM: but now I seem to be at the right place
11:51 karolherbst: but this is generally for reinstalls, because reinstalls are fixing nothing
11:51 NewtoM: I agree in general.
11:52 NewtoM: But this time I tried millions of things from plenty of sources and I don't think that that was a good idea ...
11:52 karolherbst: ohh okay
11:52 NewtoM: So the system was in a very wracked state
11:52 karolherbst: so you already messed up your system
11:52 NewtoM: YEP
11:52 karolherbst: then it makes sense
11:53 NewtoM: OK, the installation still needs some time. I'll be back once it has finished. Thanks in advance'
11:57 karolherbst: imirkin_: I know that kscreen was able to leave me with a black screen, because it just applies an invalid xrandr mode :D but this was on intel
11:58 daemon32: karolherbst: Sorry, something came up
11:59 pmoreau: karolherbst: Interested by the small pieces parsed from the Titan X VBIOS?
11:59 daemon32: karolherbst: I'll put a load other than vblank_mode=0 glxgears and see xD
11:59 pmoreau: There's no CSTEP though yet, I'm going to fix that.
11:59 karolherbst: daemon32: the env var is different for blob
11:59 daemon32: Huh...
11:59 karolherbst: daemon32: __GL_SYNC_TO_VBLANK=
11:59 daemon32: I still got like 25k frames
11:59 karolherbst: ohh strange
11:59 karolherbst: okay anyway
12:00 karolherbst: nvidia-settings should tell you the maximum pcie width
12:00 karolherbst: pmoreau: mhhhh a little
12:00 karolherbst: the vbios just seems like that it was hacked together
12:00 qq[IrcCity]: hello. I have «/dev/fb0: No such file or directory» after successful «modprobe nouveau config=NvForcePost=1». what am I doing wrong?
12:00 pmoreau: karolherbst: There are a lot of voltage entries :-)
12:00 karolherbst: pmoreau: well
12:00 daemon32: karolherbst: I figured, it's a crappy PNY one xD
12:00 karolherbst: the parsing isn't right
12:00 karolherbst: daemon32: you could test if the clocks are scaling
12:01 karolherbst: ohh wait
12:01 karolherbst: better idea
12:01 karolherbst: daemon32: you see the powermizer section?
12:01 karolherbst: there are min/max values for each pstate
12:01 karolherbst: I think we would kind of need them
12:02 karolherbst: I think mupuf was right though
12:02 karolherbst: the blob doesn't give a damn about cstates
12:03 karolherbst: ohh wait
12:03 karolherbst: it wasn't you daemon32 :D
12:04 karolherbst: kubast2: still there?
12:04 kubast2: yeah
12:05 qq[IrcCity]: how can I check whether is fbdev enabled for nouveau? (and if isn’t, then why)
12:05 karolherbst: kubast2: could you boot with blob, open nvidia-settings and check the min/max clocks in the powerMizer section?
12:05 imirkin_: qq[IrcCity]: dmesg | grep nouveaufb
12:05 kubast2: yeah sure
12:06 karolherbst: daemon32: you should look up where you bought the card and check the specs
12:09 kubast2: installed
12:09 NewtoM: So, I am back. The new Freya is up and running. I could even log, but the Graphics does not feel right. How can I check if Nouveau is working?
12:10 NewtoM: (i mean "I could even log in")
12:10 imirkin_: NewtoM: pastebin dmesg and xorg log
12:10 karolherbst: do glxinfo :)
12:10 imirkin_: oh yeah, throw that in too
12:10 NewtoM: glxinfo is currently not installed ...
12:10 imirkin_: note that UIs designed for 2015 don't work very well on GPUs sold in 2001
12:10 NewtoM: :D so, apt-get first
12:10 karolherbst: :)
12:11 NewtoM: copy paste will not work, I am chatting on a different laptop
12:11 kubast: level: 0 core: min 324 mhz max 324 mhz memory: min: 648 mhz max: 648 mhz
12:11 imirkin_: er, sorry -- 2002.
12:11 kubast: level: 1 core: min 540 mhz max 540 mhz memory: min: 1620 mhz max: 1620 mhz
12:11 imirkin_: NewtoM: ssh?
12:12 kubast: level: 2 core: min 551 mhz max 1058 mhz memory: min: 5000 mhz max: 5000 mhz
12:12 kubast: turns out I remembered it uncorrectlly
12:12 NewtoM: I will find a way
12:13 imirkin_: http://www.nvidia.com/page/geforce4go.html -- blast from the past :)
12:13 kubast2: Level 0 Current PCIe Link Width x16 link speed :2.5GT/s . Level 2 x16 8.0 GT/s
12:13 kubast2: level: 0 core: min 324 mhz max 324 mhz memory: min: 648 mhz max: 648 mhz level: 1 core: min 540 mhz max 540 mhz memory: min: 1620 mhz max: 1620 mhz level: 2 core: min 551 mhz max 1058 mhz memory: min: 5000 mhz max: 5000 mhz
12:13 kubast2: karolherbst ^
12:14 kubast2: 2.5GT/s on level: 1
12:14 karolherbst: mhh okay
12:14 karolherbst: so the blob respects the boost table at least
12:17 kubast2: *by blob you meant nvidia drivers[blob]
12:21 NewtoM: OK
12:21 NewtoM: here it comes: dmesg http://pastebin.com/ZRdHhQFH, glxinfo http://pastebin.com/tCZzhLqg, Xorg.0.log http://pastebin.com/qgwG7Tpq
12:22 imirkin_: NewtoM: looks like you're using vesafb
12:22 NewtoM: I don't know. I don't want to.
12:23 NewtoM: I would like to make use of my great AGP card :)
12:23 imirkin_: NewtoM: nomodeset
12:23 imirkin_: that makes sure that nouveau doesn't load
12:23 imirkin_: remove that from your kernel cmdline
12:23 NewtoM: helpme
12:23 imirkin_: [ 0.000000] Kernel command line: BOOT_IMAGE=/boot/vmlinuz-3.19.0-26-generic root=UUID=419cf448-40bc-4b3a-a8cc-56b897107571 ro noapic edd=on nodmraid nomodeset quiet splash vt.handoff=7
12:23 imirkin_: you need to remove nomodeset from there
12:24 kubast2: NewtoM you are using grub ?
12:24 NewtoM: I think yes, but I did not have a choice at installation time
12:24 kubast2: sudo nano /etc/default/grub
12:24 NewtoM: i have a grub.d in etc
12:24 kubast2: find the nomodeset
12:25 NewtoM: ok i have it
12:25 kubast2: delete the "nomodest" in GRUB_CMDLINE_LINUX=
12:25 kubast2: save
12:25 NewtoM: removed it, do i need grub config
12:25 NewtoM: or how it is called ...
12:25 kubast2: then grub2-mkconfig -o /boot/grub2/grub.cfg or grub-mkconfig -o /boot/grub2/grub.cfg
12:26 kubast2: *
12:26 kubast2: then grub2-mkconfig -o /boot/grub2/grub.cfg or grub-mkconfig -o /boot/grub/grub.cfg
12:26 NewtoM: :) thx
12:26 noucro: heyho I got a gtx 560 SE (GF114) and everytime I open my old firefox session my system hangs itself up completly and dmesg tells me this:
12:26 noucro: Sep 24 18:51:54 pluto kernel: nouveau E[ PFIFO][0000:01:00.0] write fault at 0x000d800000 [PAGE_NOT_PRESENT] from PGRAPH/GPC1/PROP on channel 0x003fd30000 [Xorg[1492]]
12:26 noucro: Sep 24 18:51:54 pluto kernel: nouveau E[ PFIFO][0000:01:00.0] PGRAPH engine fault on channel 2, recovering...
12:26 noucro: any hoops I can jump through to eliminate this problem?
12:26 NewtoM: I only have grub-* and not grub2-*, is that ok?
12:27 kubast2: yes
12:27 imirkin_: noucro: what kernel version?
12:27 kubast2: you choose the line that works for you
12:27 noucro: imirkin_: oh sorry forgot to mention 4.1.6
12:27 imirkin_: odd.
12:27 noucro: oh yeah running archlinux stock kernel
12:27 imirkin_: noucro: are there any other messages before that?
12:28 NewtoM: (rebooting)
12:28 imirkin_: like validation fail or somesuch
12:28 karolherbst: imirkin_: we have to remove all mentions of "nomodeset" from all wikis and remove them :D
12:28 noucro: not really but here hope this helps (wgetpaste of journalctl|grep nouveau -i)
12:28 noucro: http://paste.pound-python.org/show/JReBJlEZ80MAyXGF4xRJ/
12:29 imirkin_: noucro: that's a G98, not a GF114...
12:29 imirkin_: oh, i see you switched
12:30 noucro: yeah the G98 is my laptop gpu I switch the harddisk when I work at home
12:30 NewtoM: Yep, now it is Nouveau and now I have that problem of not being able to log-in, again. So I will look for the logs once again.
12:30 imirkin_: noucro: odd :(
12:31 NewtoM: do you need dmesg again, or only xorg.log?
12:31 imirkin_: NewtoM: both
12:31 NewtoM: mkey
12:32 noucro: imirkin_: anything I could test? some module option to set? any debugging I could do?could this be because my (simple) multi monitor setup? (even though I don't think so or better said I hope not)
12:33 imirkin_: noucro: most likely something firefox does that nothing else does.
12:33 imirkin_: noucro: i tend to mostly use chrome, so i might not have noticed it...
12:34 m3n3chm0:nasZ
12:34 NewtoM: OK, here it comes again. Dmesg: http://pastebin.com/5tpBDifA , Xorg.0.log: http://pastebin.com/HdMgh0wi
12:35 noucro: chrome somehow gives me an uncanny shitty user experience (since it slows down to a crawl when I only have 20+ tabs open), but there must be some way to take look at what firefox does without having to cast black magic or not?
12:36 imirkin_: NewtoM: hmmmm.... am i correct in understanding it's a 1024x758 panel?
12:36 NewtoM: it should be 1024x768, I guess. And not 758 ... Let me check online.
12:36 imirkin_: typo
12:36 imirkin_: i meant 768
12:37 NewtoM: yep :)
12:37 NewtoM: is it a problem?
12:37 imirkin_: well.... the kernel seems perfectly happy
12:37 imirkin_: as does X
12:38 imirkin_: can you stick LIBGL_ALWAYS_SOFTWARE=1 into /etc/environment and try again?
12:38 noucro: imirkin_: sure
12:38 imirkin_: noucro: that was for NewtoM
12:38 noucro: :P oops
12:38 NewtoM: yep, sure
12:38 noucro: kk then I wont
12:39 NewtoM: imirkin_: do i need to reboot?
12:39 imirkin_: NewtoM: just restart X and so on
12:40 imirkin_: NewtoM: oh, and log out/back in
12:41 NewtoM: now I can log-in, but it's sooooooooo slow. Oh well, and also the log-in screen feelds slow, which it wasn't until now.
12:42 imirkin_: NewtoM: ok, so i'm guessing the issue is that you have a setup that requires modern GL features
12:42 NewtoM: I have 2 fps whatever I do :)
12:42 imirkin_: my recommendation would be to switch to something that doesn't
12:42 NewtoM: oh, does this GPU not support OpenGL?
12:43 imirkin_: no, it does
12:43 imirkin_: but only GL 1.2
12:43 imirkin_: whereas most things nowadays want shaders, etc
12:44 imirkin_: your GPU has something known as texture combiners, which are a sort of very early form of shaders, but the open-source drivers don't support it, and i'm fairly sure that nothing outside of very old games uses it
12:44 NewtoM: ok, i see
12:44 imirkin_: that was my point about not trying to use a 2015 ui on a 2002 gpu
12:44 NewtoM: Let me get that right, is it LightDM, which would need those new features?
12:45 imirkin_: lightdm is the login manager... iirc it shouldn't be too heavy
12:45 imirkin_: but if there's anything to do with gnome in there, chances are it's very heavy
12:45 NewtoM: Well, if people on Elementary OS's site say, it can be used on old hardware ...
12:45 NewtoM: ... i thought I give it a try.
12:45 NewtoM: And now I am sad it won't work.
12:49 NewtoM: If I still want to give it a try, what should I use instead of gnome?
12:49 NewtoM: (and how to change it, but for that, I can google once more ... :])
12:49 karolherbst: NewtoM: window manager should ru with compositing disabled
12:49 karolherbst: mhh
12:50 karolherbst: there is xfce and lxdm/lxqt
12:50 kubast2: xfce :3
12:50 karolherbst: I think xfce is most likely the best if you don't use a gpu?
12:50 kubast2: xfce+numix theme
12:50 karolherbst: don't know
12:50 karolherbst: lxqt uses qt5 now, and they want to use a heavy windows manager like kwin or something
12:50 kubast2: http://i1.kym-cdn.com/photos/images/newsfeed/000/591/928/94f.png
12:50 NewtoM: I would go with xfce, it's site says it's lightweight
12:50 karolherbst: lxqt is also lightweight
12:50 kubast2: lxde is too
12:51 karolherbst: yeah but lxde is old
12:51 NewtoM: well, guys, I have the least idea
12:51 NewtoM: :)
12:51 NewtoM: Just tell me which one to pick :D
12:51 karolherbst: lxde is like, superceeded by lxqt ;)
12:51 karolherbst: you should try out xfce and lxqt
12:51 karolherbst: and use what you like most
12:51 NewtoM: ok, sounds fair
12:51 NewtoM: thanks
12:52 kubast2: gonna send you my xfce desktop
12:52 karolherbst: imirkin_: do you think LIBGL_ALWAYS_SOFTWARE=1 is the best thing to do? I don't really now about the implications, but using llvmpipe on an old cpu is a bit messy
12:52 imirkin_: karolherbst: it was a quick way to avoid using the nouveau_vieux driver.
12:52 NewtoM: Is there any preferred and easy way to do the swap from Gnome to xfce?
12:52 karolherbst: if applications doesn't check if context creation fails or something, it is their fault
12:53 imirkin_: NewtoM: ask the support channel for your distro?
12:53 karolherbst: NewtoM: install a xfce meta package
12:53 imirkin_: karolherbst: until someone wants to try to log in :)
12:54 karolherbst: :D
12:54 karolherbst: but I kind of figured, that gnome simply doesn't care
12:54 karolherbst: I think they want to use GL more and more
12:54 karolherbst: NewtoM: you could also try to use kde, it may run fine with compositing disabled
12:55 karolherbst: at least kwin starts for me, even when mesa is totally messed up
12:55 NewtoM: :)
12:55 karolherbst: never used lxqt though
12:56 karolherbst: it really depends on the gpu, usually when compositiing runs fine, xfce felt better than lxde back then
12:56 karolherbst: imirkin_: how much fun is anything Xrender based on his gpu?
12:56 imirkin_: EXA should work ok
12:56 NewtoM: Ok, I just hit "install" for xfce4 meta package. I am curious what happens next.
12:56 imirkin_: there were some issues in old nouveau versions on nv1x, but i fixed them
12:57 karolherbst: NewtoM: you may have to choose the DE in lightdm
12:57 karolherbst: imirkin_: okay
12:57 karolherbst: kwin has still its Xrender backend
12:57 karolherbst: and I am using 5.4
12:57 NewtoM: (what is DE?)
12:57 karolherbst: desktop environment
12:57 NewtoM: oic
12:57 imirkin_: stupid thing that slows your system down while providing no actual value
12:57 NewtoM: :D
12:58 karolherbst: :)
12:59 karolherbst: I remember the time where I got less battery lifetime with xfce compared to kde
12:59 NewtoM: Shall I remove that LIBGL line before trying to log-in again (hopefully with xfce)?
12:59 NewtoM: Or does it make no difference for xfce?
12:59 karolherbst: NewtoM: yeah I would at least try it out
12:59 karolherbst: NewtoM: you can disable compositing in xfwm
13:00 karolherbst: but you should check if it is better with it enabled
13:00 karolherbst: but I don't think it actually uses OpenGL
13:01 NewtoM: wait a minute ... xfce or xfwm ??/
13:01 NewtoM: I am lost.
13:01 karolherbst: wm: window manager
13:01 karolherbst: de: desktop manager
13:01 karolherbst: lm: login manager
13:02 karolherbst: *dm
13:02 imirkin_: window manager = thing that lets you move windows around
13:02 karolherbst: usually the term dm is used for login manager
13:02 imirkin_: desktop manager = thing that draws the idiotic icons on your background
13:02 karolherbst: light-dm
13:02 karolherbst: sd-dm
13:02 karolherbst: g-dm, k-dm
13:02 karolherbst: and so on
13:02 karolherbst: for window manager it is more rare though :/
13:02 karolherbst: isn't xfwm the only one with "wm" in its name? :D
13:03 karolherbst: ohh right, twm
13:03 karolherbst: twm, awesome thing :)
13:03 NewtoM: Thanks :) I like your "idiotic icons" :)
13:03 Karlton: I use dwm
13:03 karolherbst: NewtoM: so if even xfwm doesn't work there is always twm
13:04 karolherbst: Karlton: does it require OpengL=
13:04 imirkin_: twm is a bit annoying coz it makes you click to create new windows
13:04 NewtoM: so, now I can choose between xfce and pantheon (?!) at log-in
13:04 karolherbst: right
13:04 karolherbst: twm sucks :D
13:04 karolherbst: NewtoM: use xfce
13:04 NewtoM: what is pantheon ... ?
13:04 imirkin_: NewtoM: just use xdm and move on with life.
13:04 NewtoM: well, xfce works pretty smooth
13:04 crazedpsyc: NewtoM: pantheon is elementaryos's custom stuff
13:05 NewtoM: will suffice fully
13:05 karolherbst: pantehon is the dm for elementary os
13:05 kubast2: pantheon is DE right?
13:05 NewtoM: and the LIBGL line is already removed
13:05 karolherbst: NewtoM: okay
13:05 karolherbst: open the xfce settings now
13:05 NewtoM: i have xfce4-settings-editor and -manager
13:06 Karlton: karolherbst: it just requires x11 deps and it's written in C in less than 2000 SLOC
13:06 karolherbst: I don't really know where it is now
13:06 karolherbst: Karlton: okay, so mhh what about tearing? :D
13:06 NewtoM: wow
13:06 NewtoM: xfce4-display-settings, maybe?
13:06 karolherbst: maybe
13:06 NewtoM: i'll tru
13:06 NewtoM: y
13:06 karolherbst: there is a xfwm tweak tool somehow
13:06 karolherbst: but there should be an option like enable compositing
13:06 NewtoM: what are we looking for? what setting?
13:07 NewtoM: ok
13:07 NewtoM: i'll check
13:07 NewtoM: there are about 20 config tools ...
13:07 karolherbst: usually that is what you want to use, but with your gpu, not sure about that
13:07 kubast2: go to settings manager
13:07 Karlton: karolherbst: you can use compton but I really don't notice much tearing so I don't care to use it
13:07 karolherbst: compton *cough*
13:07 kubast2: "tweaks of window manager"
13:08 kubast2: "compositions"
13:08 karolherbst: I don't think he wants to use compton, because it requires glx
13:08 karolherbst: I think
13:09 NewtoM: so, should I enable "display composing" or not ? :)
13:09 imirkin_: disable
13:09 karolherbst: enable it and try it out
13:09 NewtoM: ??
13:09 NewtoM: ok
13:09 karolherbst: imirkin_: I doubt that xfwm uses glx
13:09 imirkin_: i guess you can try it
13:09 NewtoM: the display flickered, but it is back and working (with ON)
13:10 karolherbst: NewtoM: does anything feel different?
13:10 karolherbst: moving windows, stuff
13:10 NewtoM: nope :)
13:10 karolherbst: :D
13:10 NewtoM: if i click shadow under windows, there is no difference at all
13:10 karolherbst: then you can leave that, because it is the new deafult
13:10 karolherbst: mhhh
13:10 karolherbst: then it just failed to enablecompositing
13:10 NewtoM: wait a minute, it says shadows under pop-ups
13:10 karolherbst: ohh
13:11 NewtoM: where should I get a pop-up now?
13:11 karolherbst: NewtoM: you should get some tearing though, but I think this is too advanced for your card? don't know
13:11 NewtoM: YES, it works
13:11 NewtoM: there IS a shadow
13:11 NewtoM: unbelievable
13:11 NewtoM: now I can do the same as winXP could ... :P
13:12 NewtoM: sorry for that
13:12 NewtoM: I am indeed happy that Linux works
13:12 imirkin_: i don't think XP had real composition...
13:12 imirkin_: that was vista+ as i recall
13:12 karolherbst: right
13:12 karolherbst: XP was software only rendering
13:12 NewtoM: I don't know but it was responsive
13:12 karolherbst: right
13:12 karolherbst: they did a good job indeed
13:12 karolherbst: but they got overtaken by os x pretty fast
13:12 imirkin_: the responsiveness is just related to not using software that expects 2015 hardware capabilities
13:12 NewtoM: so
13:12 karolherbst: os x was the first OS with gpu compositing by default?
13:13 NewtoM: even with full Shadows (under all windows), it is still pretty fast and enjoyable with xfce4
13:13 karolherbst: nice :)
13:13 karolherbst: I always trust xfce for old hardware :)
13:13 NewtoM: Thank you all! I wish Elementary OS builders would make it first choice for old hardware.
13:13 imirkin_: you should let them know
13:14 mlankhorst: no idea if it was first
13:14 mlankhorst: lets see if wikipedia knows..
13:14 imirkin_: mlankhorst: and if it doesn't, we can always fix it ;)
13:14 karolherbst: mlankhorst: I am pretty sure though
13:14 karolherbst: mlankhorst: I think with 10.2 they started their ps desktop backend and throw everything on the gpu
13:15 karolherbst: which was like 2003
13:15 karolherbst: right Quartz Extreme it was
13:15 NewtoM: now if I can disable "click by touch" on the touchpad, then I am ready to use this OS
13:15 mlankhorst: according to wikipedia there was a demo for windows too, though I guess it got released in 2007 with vista or something
13:15 NewtoM: :)))
13:16 NewtoM: GOT IT
13:16 karolherbst: AGP x2 was the requiernment though :D
13:16 NewtoM: great :)
13:16 NewtoM: Thank you really much for your precious time and help!
13:16 karolherbst: and 16MB vram?
13:16 kubast2: damn
13:16 mlankhorst: composite extension was added in 2004 to x11
13:16 karolherbst: ohh right, Quartz extreme was gpu stuff
13:17 kubast2: even my old gf3 ti-200 had 64
13:17 karolherbst: QuartzGl (added with Tiger) used OpenGL
13:17 karolherbst: there are some differences, but
13:17 mlankhorst: v10.2 was in 2002, so I guess that one wins..
13:17 karolherbst: 10.4 with OpenGL based compositing
13:17 karolherbst: 2005 was 10.4
13:18 karolherbst: mlankhorst: yeah, but this was a bit of cheating :D
13:18 karolherbst: because not everything was done on the gpu
13:18 karolherbst: only the final scene compositing
13:18 NewtoM: One more thing (well, there is always one more ...): irrespective of what X application I start, there is now a Gtk-message: Failed to load module "pantheon-filechooser-module"
13:18 karolherbst: with 10.4 everything was then moved to the gpu
13:18 NewtoM: Is that bad?
13:18 karolherbst: NewtoM: ignore that
13:18 karolherbst: you want to use the xfce one anyway
13:19 karolherbst: I guess there are some config leftovers
13:19 mlankhorst: can't find when aiglx was required..
13:20 mlankhorst: fedora core 6 had it as feature in 2006, hmph
13:20 mlankhorst: mm wobbly windows on a desktop cube
13:20 NewtoM: anyone here?
13:21 karolherbst: NewtoM: nope
13:21 imirkin_: NewtoM: this channel is about nouveau, not $random distro
13:22 karolherbst: mlankhorst: but do you know what is awesome about OS X? switcing gpus while gl applications are running
13:22 karolherbst: I have no clue how they do it, but I guess they somehow manage the context in driver independent parts
13:22 karolherbst: and just recreate some hardware context
13:23 karolherbst: would be nice to add support for that on linux
13:24 karolherbst: pmoreau: you had this macbook pro right? which was it?
13:25 pmoreau: MBP5,3
13:25 karolherbst: pmoreau: please the year name :D
13:25 pmoreau: Mid 2009
13:25 karolherbst: nice
13:26 karolherbst: do you know what GeForce Boost is?
13:26 pmoreau: Nope
13:26 karolherbst: you have a hybrid SLI config
13:26 karolherbst: "Spreads the graphics-processing over both GeForce GPUs (discrete GPU and motherboard GPU) in a Hybrid SLI notebook, increasing gaming performance up to 80%."
13:26 pmoreau: Really?! O.O
13:26 karolherbst: yes
13:26 karolherbst: http://www.geforce.com/Hardware/NotebookGPUs/geforce-9400mg
13:26 karolherbst: ohh link is dead
13:26 karolherbst: sad
13:26 pmoreau: Hum...
13:26 karolherbst: this was confirmed by nvidia though
13:26 karolherbst: pmoreau: http://www.everymac.com/systems/apple/macbook_pro/macbook-pro-unibody-faq/macbook-pro-unibody-switching-between-graphics-processors.html
13:26 karolherbst: "Switching Graphics without Logout - Earlier Models" and later
13:27 karolherbst: pmoreau: did you at any time encountered anything sli specific?
13:27 pmoreau: Oh oh!
13:27 pmoreau: I think I saw some SLI regs being touched by the blob
13:28 karolherbst: :)
13:28 karolherbst: there ya go
13:28 pmoreau: But other than that, first time I'm ever hearing about that for my laptop
13:28 karolherbst: http://www.nvidia.com/object/hybridsli_notebook.html
13:28 pmoreau: I keep getting things on my todo list... --"
13:28 imirkin_: the idea of SLI is that you can send largely identical cmd streams to multiple GPUs
13:28 imirkin_: with only slight differences which are expressed via pushbuf flags
13:29 imirkin_: where e.g. the viewports might be done differently for each SLI bit
13:29 imirkin_: so you don't have to compute 2 diff cmd streams
13:29 imirkin_: and when the SLI linkage is properly initiated
13:29 imirkin_: each GPU knows if it's SLI slave or master
13:29 pmoreau: There was something in the blob trace (can't rememeber which command) which had a field for a card id from what I understood
13:29 imirkin_: and knows which commands to ignore
13:30 NewtoM: hello back
13:31 pmoreau: Could this be hacked to allow switch betwen both cards without logging out?
13:32 imirkin_: pmoreau: well there's no great technological reason for it
13:33 imirkin_: pmoreau: but there's lots of bad reasons that make it tricky to switch gpu's in the middle of a session
13:33 karolherbst: pmoreau: what do you mean by that?
13:33 karolherbst: ahh
13:33 karolherbst: pmoreau: it already works on mac os x with newer laptops
13:33 pmoreau: Because they have softmux
13:33 karolherbst: gpu switches are on the fly
13:33 pmoreau: Mine has a hardmux
13:33 karolherbst: no application restart needed
13:33 imirkin_: pmoreau: so what?
13:34 karolherbst: pmoreau: doesn't matter with SLI as far as I know
13:34 imirkin_: pmoreau: it's not like any real state is kept on the gpu
13:34 karolherbst: because the gpu with the display connected to it is the master, and the other the slaves?
13:34 pmoreau: I guess you could send the command stream to both cards, switch the mux, power off the other one
13:34 imirkin_: pmoreau: X should just be able to disconnect from one hw device and attach to another
13:34 imirkin_: pmoreau: but in practice it can't
13:34 pmoreau: --"
13:34 karolherbst: imirkin_: yeah but this will mess up applciation states? or not
13:35 imirkin_: karolherbst: what do applications know?
13:35 karolherbst: right
13:35 imirkin_: karolherbst: i guess if they're doing GLX, that would suck
13:35 karolherbst: so how "difficutl" would it be to add support in mesa to switch to another gpu
13:35 imirkin_: but with the various robustness extensions it's supportable
13:36 imirkin_: but this is all wild fantasy
13:36 imirkin_: it'd take years of hacking to get this all to work properly
13:36 karolherbst: :/
13:37 karolherbst: it kind of bothers that users have to decide which gpu to use
13:37 karolherbst: this should be changed
13:37 karolherbst: allthough doing it on application loading time like on windows should be fine already
13:38 karolherbst: but how should libgl know how intense an application is
13:49 NewtoM: Thanks once more for your help.
13:49 NewtoM: Good night!
13:54 jefferym: ?
13:55 pmoreau: Meh, not sure I want to fill in the blanks of DCB parsing in nvbios...
13:59 karolherbst: pmoreau: then do SLI on your laptop :)
14:00 pmoreau: I already have the SPIR-V thing
14:00 karolherbst: I see
14:00 pmoreau: It's just that by fixing nvbios not to crash on the Titan X VBIOS, I started looking at what failed
14:01 pmoreau: And from there, realise that they were some unk that can be de-unk using the DCB doc from Nvidia
14:01 karolherbst: I see
14:01 karolherbst: I also work on something for that
14:01 karolherbst: so,
14:01 karolherbst: but in the conn table
14:02 karolherbst: is hdmi C always mini?
14:02 karolherbst: okay c is mini
14:03 imirkin_: https://en.wikipedia.org/wiki/HDMI#Connectors
14:04 imirkin_: karolherbst: https://en.wikipedia.org/wiki/HDMI#/media/File:HDMI_Connector.jpg
14:05 karolherbst: can I push that directly to master? https://github.com/karolherbst/envytools/commit/adc2f1f384750044f4a0e5cba014691bfb473d61
14:05 karolherbst: ohh whitespace
14:06 imirkin_: karolherbst: fix it up, and sure.
14:11 karolherbst: this vbios is so strange
14:12 pmoreau: Not sure if I should push just the first one, both or merge them: https://phabricator.pmoreau.org/P30 https://phabricator.pmoreau.org/P31
14:12 karolherbst: pmoreau: does the first one causes a segafult otherwise?
14:13 karolherbst: I don
14:13 pmoreau: Yep!
14:13 pmoreau: If you have an unknown version of CSTEP
14:13 karolherbst: 't know, it would be better to figure out what the table is about, but well
14:13 pmoreau: Like it was the case with the GM200
14:14 pmoreau: The probmel is not about understanding the table.
14:14 karolherbst: pmoreau: which version do you have?
14:14 mwk: what is HDMI-
14:14 mwk: C?
14:14 karolherbst: mwk: mini HDMI
14:14 mwk: oh
14:14 pmoreau: Currently, if you have an unknown version, ent2 is not allocated
14:15 mwk: ... is there HDMI-A and -B?
14:15 karolherbst: yes
14:15 karolherbst: "normal" HMDI is A
14:15 pmoreau: karolherbst: CSTEP version 0x8
14:15 karolherbst: B is unsued
14:15 karolherbst: D is micro
14:15 karolherbst: and E is for cars
14:15 karolherbst: I think?
14:16 karolherbst: ohh no, it is a safe versio of A
14:16 mwk: cars.
14:16 mwk: safe?
14:16 karolherbst: it won't get out by vibrations
14:16 karolherbst: http://www.hdmi.org/manufacturer/hdmi_1_4/automotive_connection.aspx
14:23 hakzsam: any ideas how heaven is able to read and display current clocks (ie. graphics, mem and processor) while benchmarking the blob?
14:23 imirkin_: hakzsam: NV_memory_info or whatever
14:24 imirkin_: hakzsam: https://www.opengl.org/registry/specs/NVX/gpu_memory_info.txt
14:24 imirkin_: i guess that doesn't have clocks....
14:24 hakzsam: nope
14:24 hakzsam: not this one
14:24 imirkin_: NV_something_else then :)
14:26 mwk: huh.
14:30 hakzsam: I doubt heaven uses a GL ext because I can't find any relevant stuff here https://developer.nvidia.com/nvidia-opengl-specs
14:31 hakzsam: or with nvidia-smi? but when I try 'nvidia-smi -q -d clock', all clocks are N/A :)
14:31 imirkin_: might be over NV-CONTROL
14:31 imirkin_: which is an X extension they use
14:31 imirkin_: (which is how nvidia-settings works)
14:34 hakzsam: imirkin_, right
14:34 hakzsam: https://chromium.googlesource.com/chromium/src.git/+/master/third_party/libXNVCtrl/NVCtrl.h#2932
14:34 karolherbst: hakzsam: usually nvidia-settings packages are providing the XNVCtrl lib
14:35 karolherbst: which is mainly open source
14:35 hakzsam: yep
14:35 daemon32: karolherbst: Yes, the clocks scale, and also the PNY website says "PCI-Express 3.0 16x" for my card
14:36 karolherbst: daemon32: mhhh
14:36 karolherbst: daemon32: what motherboard do you have?
14:36 daemon32: I have a gigabyte GA-P55-UD3R (Intel Series 5/P55 chipset)
14:37 daemon32: And it uses DMI instead of QPI for perhipherals in the PCH, so I'm not sure if maybe the lanes are being put there instead?
14:37 karolherbst: mhh
14:37 karolherbst: it looks right, but something is bothering me
14:37 karolherbst: the second x16 slot is only running at x4
14:38 daemon32: I don't remember why that is, I'll see if I wrote something somewhere...
14:38 karolherbst: daemon32: do you want to check the other slot with your gpu?
14:38 daemon32: karolherbst: You mean move it to the x4 slot?
14:38 daemon32: karolherbst: And the x4 slot only physically has x4 pins, even though it's a full-sized slot
14:38 karolherbst: I know
14:39 karolherbst: but you said on windows it is x16?
14:39 karolherbst: you can also check there
14:39 karolherbst: maybe there is something odd with the kernel itself
14:39 karolherbst: not nouveau
14:39 daemon32: Possibly, I'll see
14:40 karolherbst: it can't be nouveau, because nouveau doesn't handle that yet anyway :D and it was also faulty with nvidia
14:40 daemon32: Well, when I last checked, I actually had a Fermi card (GTX 460)
14:40 daemon32: So that's why I was concerned
14:40 daemon32: Yeah...
14:41 karolherbst: ohhh
14:41 karolherbst: I read some strange things about your board
14:41 daemon32: Oh really?
14:41 daemon32: Maybe it's time to upgrade anyway xD
14:42 karolherbst: daemon32: what is your cpu?
14:42 karolherbst: daemon32: https://forums.whirlpool.net.au/archive/1406988
14:42 karolherbst: :D
14:42 daemon32: karolherbst: Core i5-750 (@3.36 GHz, in case there's something I overlooked with the BCLK affecting something)
14:42 karolherbst: what the heck
14:43 karolherbst: 16 lanes :O
14:43 karolherbst: okay
14:43 karolherbst: this sounds ugly eough
14:43 daemon32: Yeah... I'm probably crazy
14:43 karolherbst: daemon32: okay, please disable usb-3 support in your vbios
14:43 karolherbst: then boot into linux
14:43 karolherbst: and you should get x16
14:43 hakzsam: karolherbst, GPLv2 btw https://github.com/NVIDIA/nvidia-settings/blob/master/COPYING
14:44 daemon32: karolherbst: I don't recall it having USB 3, but I'll look
14:44 karolherbst: daemon32: I think the board is really crippled with lanes, and if sata and usb are running at full speed and stuff is enabled, it may not be enough
14:44 hakzsam: nvidia releases open source softwares, incredible :D
14:44 daemon32: karolherbst: I only have EHCI "USB2 Enhanced Host Controller"
14:45 karolherbst: ohh okay
14:45 daemon32: karolherbst: Probably, I'll turn off everything and use an external USB cd drive from now on
14:45 daemon32: karolherbst: The ATA interface uses the weird 'GSATA' stuff for some reason
14:45 karolherbst: daemon32: try sata-2 only
14:59 karolherbst: okay
14:59 karolherbst: first conclusion: there is nothing like a list of cstates for each pstate
15:00 karolherbst: pstates simple have on cstates they are supose to run
15:03 karolherbst: second conclusion: cstates don't have memory clocks, because they are core only
15:04 jefferym: the freedesktop bug id #73445 mentions "we nuked MSI on all the nv4x IGPs", when were these nuked?
15:09 karolherbst: mupuf: this voltage table is completly different than the others
15:10 karolherbst: jefferym_: which bug tracker?
15:11 karolherbst: ohh found it
15:11 karolherbst: jefferym_: and why do you ask?
15:11 jefferym_: this one; https://bugs.freedesktop.org/show_bug.cgi?id=73445 comment 4, I have one of these NV4E chips and am looking into some issues
15:12 karolherbst: jefferym_: what issue do you have
15:12 imirkin_: jefferym_: 3.15ish
15:12 jefferym_: I am seeing some flickering on the screen
15:12 imirkin_: jefferym_: upstream commit 4761703bd04
15:13 imirkin_: jefferym_: which was cherry-picked into some stable trees i think
15:13 jefferym_: great, thanks imirkin
15:13 imirkin_: whoa, looks like only in 3.19
15:13 imirkin_: jefferym_: iirc the issue was a lot worse than flickering though
15:14 karolherbst: yeah I would guess this is someting else
15:14 jefferym_: ok, I am now trying to look at the existing bugs in the trackers to see if anything seems useful. I'll have more information on it tommorow
15:14 karolherbst: jefferym_: flickering since always or since a special kernel version?
15:15 jefferym_: I have only used this in recent kernels, starting with 4.0.5 or so, I have fedora on it now and am upgrading to its most recent kernel
15:17 imirkin_: jefferym_: are you using gnome-shell?
15:17 jefferym_: I am on here
15:17 imirkin_: i'd strongly advise against that
15:18 jefferym_: any particular reason? just becuase of old hardware and the overhead?
15:18 imirkin_: because all problems in the universe are due to gnome-shell :)
15:18 jefferym_: I see :)
15:19 imirkin_: it's just a good thing to rule out
15:19 imirkin_: the nv30 3d driver is kinda shit
15:19 jefferym_: I am seeing the flickering at the moment in run mode 3 without X running, just the virtual terminal
15:19 imirkin_: ah ok
15:19 imirkin_: that's much worse
15:19 imirkin_: try booting with nouveau.pstate=1 and clocking up? it might not be able to keep up if you have a high resolution
15:21 jefferym_: Thanks, I'll try that out and see how it works. Unfortunately I will have to wait until the morning
15:21 imirkin_: what's the resolution on the screen?
15:21 jefferym_: it's at 1920x1080
15:21 imirkin_: ah ok, that qualifies as higher resolution
15:22 jefferym_: I'll drop the res and see if it clears up as well, a reasonable thing to try
15:23 karolherbst: jefferym_: you could try out another window manager as well
15:23 karolherbst: install xfwm4 for example and start it with xwfm4 --replace
15:25 imirkin_: jefferym_: not so easy to do with fbcon
15:28 jefferym_: I will try that in the morning and report back, thanks for the help
15:37 daemon32: karolherbst:
15:39 daemon32: karolherbst: Still happens in windows :P
15:39 daemon32: I don't have an IRC client in windows so I'm using telnet, so excuse my broken messages xD
15:46 daemon32: Do I suffer greatly by not having all 16 lanes?
16:15 karolherbst: imirkin_: strange a x16 card having only 8 lanes on a x16 system?
16:16 karolherbst: something is odd
16:16 imirkin_: probably hw-level issue
16:16 imirkin_: like it's plugged into the second x16 slot which only has 8 lanes hooked up or something like that
16:16 karolherbst: he said he know that one slot has x4
16:16 karolherbst: so he used the x16 one
16:17 karolherbst: still the card should tell us it can do x16
16:17 imirkin_: or there's some sort of physical issue in place, e.g. some of the back connectors are broken/etc
16:17 karolherbst: mhhh
16:17 karolherbst: maybe the reg tells us slot width, enabled with
16:17 karolherbst: and the slot is dirty
16:18 karolherbst: don't know
16:18 karolherbst: but he didn't tell me what nvidia-settings was saying for maximum width so
16:19 karolherbst: hakzsam: still on reator?
16:19 karolherbst: doesn't seem so
16:22 karolherbst: ha
16:22 karolherbst: the blob knows
16:22 karolherbst: the blob knows the card max, and what the current width is
16:23 karolherbst: this is good
16:28 karolherbst: hey, this stupid godamn maxwell cards
16:31 karolherbst: imirkin_: okay, this pcie width reg is also there on fermi
16:31 karolherbst: but not on kepler+
16:45 karolherbst: mhhh
16:45 karolherbst: I think lookup is kind of broken
16:46 daemon32: karolherbst: If you don't mind me asking, what does "mhhh" mean?
16:46 karolherbst: nothing special
16:46 karolherbst: daemon32: you card is strange by the way :D
16:46 daemon32: karolherbst: Thanks xD
16:47 karolherbst: daemon32: did you check in nvidia-settings?
16:47 karolherbst: it should tell you if the card can do x8 or x16
16:47 daemon32: karolherbst: I couldn't get it to go x16 in windows in linux by turning off anything in the bios
16:47 daemon32: karolherbst: And it reports itself as PCIe GEN 2 x16
16:47 daemon32: karolherbst: In both windows and linux
16:47 karolherbst: check "PCIe maximum width"
16:48 karolherbst: in nvidia-settings
16:48 karolherbst: when you select the gpu
16:48 daemon32: karolherbst: Alright
16:48 karolherbst: maybe the card is x8 only for whatever reason
16:49 daemon32: I could put a GTS 250 or a GTX 460 in to see if that goes x16 for some reason
16:50 karolherbst: what does nvidia-settings say?
16:52 karolherbst: mhhh
17:22 daemon32: karolherbst: "Maximum PCIe Link Width: x16"
17:23 daemon32: karolherbst: "Current PCIe Link Width: x8" and I can make "Current PCIe Link Speed:" 2.5 GT/s and 5 GT/s
17:23 daemon32: But that's it
17:24 karolherbst: mhhh okay
17:24 karolherbst: strange
17:24 karolherbst: so the card seems to be able to do x16
17:24 daemon32: I blame the board
17:25 daemon32: I might try my old core 2 duo board (P45)
17:25 daemon32: It only has one 16 lane PCIe 2.0 slot on it
17:25 daemon32: I don't mean to waste your time, sorry
17:25 karolherbst: nono, actually such a board is nice
17:25 karolherbst: mhhh
17:26 karolherbst: I am working on getting this pcie stuff to work in nouveau anyway
17:26 karolherbst: and such corner cases help a lot
17:26 daemon32: Nice as in for debugging, I assume? xD
17:26 karolherbst: it is an kepler card, right?
17:26 daemon32: Yeah, it's a PNY GTX 760 (GK104)
17:26 karolherbst: envytools installed?
17:27 daemon32: Yup
17:27 daemon32: I'm running the blob right now though
17:27 karolherbst: nvapeek 0x8c080
17:27 karolherbst: doesn't matter
17:28 daemon32: 0008c080: 00000808
17:28 karolherbst: okay, that makes sense
17:28 karolherbst: then
17:28 karolherbst: nvapeek 0x88000 0x1000
17:28 karolherbst: plase paste it on the web somewhere
17:28 daemon32: Got it
17:29 karolherbst: I need the link though ;)
17:30 daemon32: http://pastebin.com/5H0bTcJg
17:30 karolherbst: and then nvapeek 0x8c000 0x1000
17:31 daemon32: http://pastebin.com/x41YDRFD
17:32 karolherbst: thanks
17:32 daemon32: No problem :)
17:33 karolherbst: and then lspci -s 01:00.0 -vv
17:34 daemon32: http://pastebin.com/CxbW8YYJ
17:34 karolherbst: thanks
17:34 daemon32: You're welcome
17:34 imirkin_: LnkCap: Port #0, Speed 5GT/s, Width x16, LnkSta: Speed 5GT/s, Width x8,
17:34 daemon32: That's what I've been seeing all day
17:34 imirkin_: some issue on the mobo =/
17:35 daemon32: Which is what I thought
17:35 daemon32: This board is crap anyway, I can't go past BCLK 166 without it becoming unstable
17:35 karolherbst: mhh
17:36 karolherbst: so the card can actually do x16, this is for sure
17:36 karolherbst: daemon32: maybe you put it into the wrong slot?
17:37 daemon32: karolherbst: I put it into the one closest to the CPU
17:37 daemon32: karolherbst: http://www.gigabyte.com/fileupload/product/2/3162/1206.jpg
17:37 daemon32: karolherbst: Which is the only slot that has x16 lanes
17:37 daemon32: karolherbst: Closest blue slot*
17:37 karolherbst: sadly the res is too low
17:37 karolherbst: yeah
17:37 karolherbst: you see it
17:38 karolherbst: the lower slot just don't have the connectors
17:38 daemon32: Yeah, which is why I didn't populate it
17:38 daemon32: And I've got one conventional PCI card, if that helps
17:38 karolherbst: why do you hve this board? :D
17:38 daemon32: I got it on ebay for cheap
17:38 karolherbst: I see
17:38 karolherbst: what kind of pci card?
17:38 daemon32: It's a sound card
17:38 karolherbst: ad what chipset is on the board?
17:38 daemon32: I needed it for midi stuff a long time ago
17:39 daemon32: The board is an Intel P55 chipset board
17:39 daemon32: AKA Series 5
17:39 karolherbst: :O
17:39 daemon32: ICH10R
17:39 karolherbst: what the heck
17:39 daemon32: ?
17:39 karolherbst: that can't work at all
17:39 karolherbst: okay
17:39 karolherbst: the board chip itself only supports pcie v1
17:40 karolherbst: 8 lanes
17:40 daemon32: o.o
17:40 karolherbst: I mean it supports v2
17:40 karolherbst: but only at 2.5 speed
17:40 karolherbst: imirkin_: !!!
17:40 daemon32: Wat
17:40 karolherbst: there is a v2 board with 2.5 speed only
17:40 karolherbst: okay
17:40 karolherbst: this is propbably connected to the x1 and x4 slot
17:40 daemon32: I thought that the CPU had the primary PCIe controller?
17:40 karolherbst: I mean
17:40 karolherbst: x1 and x8
17:40 karolherbst: it has
17:40 daemon32: And that the PCH had the x4 slot?
17:40 karolherbst: what cpu did you have?
17:40 daemon32: Core i5-750
17:41 karolherbst: okay, which is the main problem
17:41 karolherbst: it has only 16 lanes
17:41 karolherbst: now thing
17:41 karolherbst: usb, sata stuff is also connected to pcie
17:41 karolherbst: if any of that uses one lane from the cpu
17:41 karolherbst: you can't add 16 to your pcie slot
17:41 karolherbst: so it falls back to 8
17:41 daemon32: Huh... I thought that was what DMI was for, and I know they don't do x9-x15
17:41 karolherbst: usually intel cpus have 20 lanes for that
17:42 daemon32: Well shoot
17:42 karolherbst: but it is more complicated then that
17:43 daemon32: I'll have to draw out all of the traces of the board -_-
17:43 daemon32: xD
17:45 karolherbst: lspci please
17:45 daemon32: Plain or -vv?
17:45 karolherbst: plain
17:46 karolherbst: imirkin_: I think my hwids are gonna crazy: "PCI bridge: Intel Corporation Xeon E3-1200 v3/4th Gen Core Processor PCI Express x16 Controller"
17:46 daemon32: http://pastebin.com/VUuhxJtg
17:46 daemon32: karolherbst: You got a xeon e3-1200? xD
17:46 karolherbst: nope
17:46 karolherbst: 4700M
17:46 karolherbst: q
17:46 daemon32: Oh? Huh...
17:47 karolherbst: my laptop also has aspm support and stuff, pretty weird
17:47 karolherbst: allthough that should be nomral
17:47 karolherbst: *normal
17:47 daemon32: I thought that ASPM was made for laptops particularly, no?
17:47 karolherbst: don't know
17:48 karolherbst: I think so
17:48 karolherbst: but not everybody uses it
17:48 daemon32: I tried disabling the IDE controller to no avail, btw
17:49 daemon32: I wouldn't be willing to turn off my ethernet for x16 graphics xD
17:49 daemon32: I'll brb, steam doesn't want to launch with lib32-nvidia-libgl right now, and I don't feel like fixing it
17:49 karolherbst: daemon32: you are on gentoo?
17:49 karolherbst: eselect opengl
17:49 daemon32: karolherbst: No, arch linux
17:50 karolherbst: ohhhh
17:50 karolherbst: then the other one was on gentoo
17:50 karolherbst: mhhh
17:50 daemon32: karolherbst: I was thinking of going gentoo or manjaro
17:50 karolherbst: yeah, libgl is messy everywhere else :D
17:50 daemon32: karolherbst: Yeah, portage is amazing xD
17:50 karolherbst: though this is eselect
17:51 karolherbst: okay, that is nice
17:51 karolherbst: now lspci -vv
17:51 daemon32: karolherbst: But you need portage for eselect?
17:51 karolherbst: daemon32: no
17:51 karolherbst: why should you?
17:51 karolherbst: eselect is just some bash scripting
17:51 daemon32: Never tried it outside of gentoo, so I didn't look to see if it had any dependency on portage
17:51 karolherbst: it needs a specific layout of files, yes
17:52 karolherbst: anyway I know what I am looking for now
17:52 daemon32: karolherbst: And that's where gentoo and arch are divided, as arch unifies /bin and /usr/bin and they have /usr/lib32 instead of /usr/lib64
17:53 daemon32: karolherbst: http://pastebin.com/FWfETX35
17:55 karolherbst: ha
17:55 daemon32: My fault or their fault?
17:55 daemon32: xD
17:55 karolherbst: it is your ethernet controller
17:55 daemon32: WOW
17:55 daemon32: I've got a PCI one... :P
17:55 karolherbst: it uses 1 lane from port #0
17:55 daemon32: ............
17:55 karolherbst: and your nvidia card is also on port #0
17:55 daemon32: THE IRONY
17:56 karolherbst: you could remove it and check if the card is at x16 mode then, but let me check further
17:56 karolherbst: yeah
17:56 karolherbst: should be the ethernet card
17:57 daemon32: Well, I'll try PCI ethernet then (my onboard controller sucks anyway)
17:57 daemon32: brb
17:57 karolherbst: wow
17:57 karolherbst: :D
18:00 karolherbst: imirkin_: should we warn the user about stuff like that?
18:00 karolherbst: though I don't know where to get the other lanes from
18:01 karolherbst: but basically we just need to complain when current width < support width, right?
18:06 imirkin: karolherbst: i think it's port 0 of a diff root complex or something
18:09 karolherbst: mhhh
18:10 karolherbst: we will see
18:11 karolherbst: but I don't know what it could be else
18:15 daemon32: karolherbst: Nope, disabling the ethernet controller had no effect
18:15 karolherbst: oaky
18:15 karolherbst: mhhh maybe your slot is dirty or something, I don't know
18:15 daemon32: Eh, I was looking for an excuse to upgrade anyway xD
18:16 daemon32: Maybe skylake, or maybe when zen comes out..?
18:16 daemon32: *Throws money at AMD and hopes for the best*
18:19 daemon32: karolherbst: Know of any modern boards with PS/2? I use a Model M
18:19 karolherbst: why PS/2?
18:19 imirkin: look into the das keyboard
18:19 karolherbst: is there any benefit in PS/2 anyway?
18:19 daemon32: I know a guy with those, but they cost a bunch last time I checked (I got my Model M for free)
18:20 imirkin: karolherbst: yeah, all the good keyboards are PS/2 :)
18:20 karolherbst: serisouly?
18:20 karolherbst: maybe 10 years ago :p
18:20 imirkin: daemon32: yeah, it's pricey. but really nice to type on.
18:20 imirkin: karolherbst: if you like the old ibm keyboards, yes. they're just not made anymore. das keyboard is the only one that comes close
18:20 karolherbst: ohh okay
18:21 imirkin: good thing you don't need an AT port ;)
18:21 imirkin: (or, *gasp*, XT)
18:21 karolherbst: usualyl I always know what keyboard I want to have, but its too expensive
18:21 daemon32: imirkin: I've got an old XT board somewhere... xD
18:22 karolherbst: http://www.artlebedev.com/everything/optimus/maximus/ :)
18:22 daemon32: karolherbst: http://webwit.nl/input/kbsim/
18:22 daemon32: karolherbst: That one is exactly like mine
18:22 karolherbst: nothing compares to the optimus maximus
18:23 daemon32: karolherbst: Great, now I can finally switch to dvorak (and go back to QWERTY any time :P)
18:23 daemon32: Huh, that keyboard has funky XT-looking keys on the side
18:24 karolherbst: as I said, nothing compares to that keyboard
18:25 karolherbst: I think it is like 1.6k $?
18:25 karolherbst: not sure
18:30 karolherbst: daemon32: but you really should check if there is somethign strange with you board
19:06 Karlton: I just avoid wireless keyboards and keyboards that have fancy lights and stuff.
19:07 Karlton: Das keyboard is for sure the best quality keyboard though very expensive
21:19 daemon32: Oh, one more thing...
21:19 daemon32: "nouveau 0000:01:00.0: clk: failed to raise voltage: -22"
21:19 daemon32: I was told (by karol) that I could come here to help debug that
21:28 daemon32: Is it normal to get this spamming your dmesg? xD "[ 8818.593926] nouveau 0000:01:00.0: therm: FAN target request: 0%"