02:05 mlankhorst: tagr_: where can I find xusb.bin?
04:16 angelsl: my GT430's fan seems to be stuck at max speed
04:17 angelsl: ..or at least some value that makes it run very loud
04:17 angelsl: dmesg doesn't indicate anything wrong with power management
04:17 angelsl: [ 1.729938] nouveau [ PTHERM][0000:01:00.0] FAN control: PWM; [ 1.729958] nouveau [ PTHERM][0000:01:00.0] fan management: automatic; [ 1.729997] nouveau [ PTHERM][0000:01:00.0] internal sensor: yes
04:20 karolherbst: angelsl: did you check sensors?
04:20 angelsl: checking
04:20 karolherbst: mupuf: your pwm work works awesomle good with my gpu now
04:21 karolherbst: *awesomly
04:21 angelsl: https://www.irccloud.com/pastebin/cqV2yFta/
04:21 karolherbst: didn't had any voltage related crashes for days now
04:21 karolherbst: angelsl: maybe nouveau can't detecht your fan and that's why it is at full speed or something
04:21 karolherbst: angelsl: you could reboot with nouveau.debug=debug
04:21 karolherbst: and give me your dmesg
04:23 mupuf: karolherbst: what did you change?
04:23 mupuf: worked on the voltage mapping table?
04:24 karolherbst: mupuf: no, I just added your fix
04:24 karolherbst: wrong commit bit
04:24 mupuf: my fix?
04:24 mupuf: ...
04:25 mupuf: yeah, this epic brown bag moment...
04:25 karolherbst: since then I don't even need my hack for higher pwm values
04:25 angelsl: karolherbst: aight, hold on
04:25 angelsl: and thanks
04:25 karolherbst: but I am really surprised that my gpu didn't even care that much about the voltage
04:25 angelsl: i just tried setting pwm1 directly, didn't help :s
04:25 mupuf: yes! That also means everything you were testing was just wron
04:25 mupuf: g
04:25 mupuf: and pure noise
04:26 karolherbst: mupuf: yes
04:26 karolherbst: mupuf: but it works now, and this is good
04:26 mupuf: let's make this a good lesson though :D
04:26 karolherbst: mupuf: I have a stupid prototpye for pmu based dynamic reclocking now and it changes cstates a lot with antichamber
04:26 karolherbst: so even all the otehr cstates seems to work fine
04:26 mupuf: ack, but you know what we need to do :p
04:27 karolherbst: yes
04:27 karolherbst: I want to get the pmu thing stabel first though
04:27 karolherbst: this is what I mainly do currently
04:27 karolherbst: also I get some fifo: LB_ERROR
04:27 karolherbst: or MEMACK_TIMEOUT
04:29 karolherbst: while working on a kernel journald is really terrible :/ I just got my journal corrupted like 5 times yesterday :D
04:29 karolherbst: and most of the messegas were just gone :/
04:30 RSpliet: karolherbst: on GT21x mem reclocking is not stable. The values are fine but the "engine pausing" mechanism is unreliable. Could lead to crashes when reclocking under load. Not sure what the status is on Kepler
04:30 karolherbst: RSpliet: well games doesn't care about htat
04:32 karolherbst: mupuf: sometimes my card just crashes and takes the nouveau driver with it. nviord waits on a reply, but doesn't get one and kworkers are starting to hang and later even my network goes down and my usb stuff :/
04:32 angelsl: karolherbst: https://usercontent.irccloud-cdn.com/file/ERB3LHi6/dmesg
04:33 karolherbst: mupuf: I want to find out what happens, but it can take hours until it happens
04:33 angelsl: sorry, that seems to force a download
04:33 mupuf: yes, I know the feeling, but RSpliet is right
04:33 karolherbst: no worries
04:33 karolherbst: mupuf: this happens also when memory isn't reclocked
04:33 mupuf: you should try reclocking without changing the memory clock speed
04:34 mupuf: then you know it is the pausing of the engines
04:34 mupuf: and it is a pain to get right
04:34 angelsl: it seems like it's actually working
04:34 angelsl: ..just that it targets 67%, which is very loud
04:34 karolherbst: mupuf: https://gist.github.com/karolherbst/22799d6be4db1732c661
04:34 karolherbst: dmesg was with debug=debug
04:35 karolherbst: mupuf: before the first line is only some fan stuff
04:36 karolherbst: or wait, maybe I didn't see something
04:36 karolherbst: mhh, no idea
04:37 karolherbst: "parsing the fan table failed"
04:37 karolherbst: mupuf: is this well, bad? :D
04:38 mupuf: can't remember
04:38 karolherbst: but this 67% seems to be something constant, because nouveau requests that fan speed all the time
04:39 karolherbst: but there is no temperature as well
04:40 karolherbst: angelsl: could you go to /sys/class/graphics/fb0/device/hwmon
04:40 karolherbst: angelsl: ohh no
04:41 karolherbst: I am blind
04:41 karolherbst: sorry
04:41 karolherbst: angelsl: could you give me your vbios.rom in /sys/kernel/debug/dri/0/ please?
04:41 karolherbst: mupuf: but if the memory messes up, shouldn't the ptimer stuff still work?
04:41 karolherbst: or a simple reg read?
04:42 karolherbst: mupuf: because that's also totally messed up then
04:42 mupuf: can you please start from the begining, I have NO idea what you are talking about :s
04:42 karolherbst: but that could be, because nouveau decides to power off the gpu :/
04:43 karolherbst: mupuf: okay
04:43 karolherbst: at some point in my tests the gpu comes unresponsive, totally unresponsive
04:43 karolherbst: even ptimer can't be read anymore
04:43 karolherbst: it returns bad00110 or something for the timer regs
04:44 karolherbst: this leads to an infinite loop here: https://github.com/karolherbst/nouveau/blob/master/drm/nouveau/nvkm/subdev/timer/nv04.c#L51
04:44 karolherbst: later nouveau decides to just power off the gpu, because it's not in use anymore (for whatever reason)
04:44 mupuf: ok
04:44 karolherbst: this leads to a hang in nvkm_rd32 and a dead kworker basically
04:44 karolherbst: later more and more kworker are dying any my system messes totally up
04:44 mupuf: that is interesting
04:45 karolherbst: network broken, usb broken,
04:45 karolherbst: stuff
04:45 mupuf: the ibus may be hung
04:45 karolherbst: may be
04:45 karolherbst: I try with runpm=0 now
04:45 mupuf: anyway, it means our pausing / restoring is broken
04:45 karolherbst: could be
04:46 karolherbst: or we just lost some PMU IRQs
04:46 karolherbst: and get old replies while pausing
04:46 karolherbst: or stuff like that
04:47 karolherbst: angelsl: well I suspect there is a missing temp => fan speed mapping missing
04:47 karolherbst: that's why I want your vbios to check why it fails reading the fan table
04:48 mupuf: your second theory sounds insane
04:49 karolherbst: mupuf: ohh that happens
04:49 karolherbst: I actually saw it already happening
04:50 karolherbst: when we lost an IRQ and we don*t handle stalled entries in the queue
04:50 karolherbst: some memory relcoking pmu stuff reads the old message
04:50 karolherbst: and can't do much with it
04:51 angelsl: karolherbst: https://usercontent.irccloud-cdn.com/file/0l1C8LW3/vbios.rom
04:52 angelsl: sorry, was eating dinner
04:55 karolherbst: angelsl: do you have envytools installed?
04:56 angelsl: can get it
04:56 angelsl: yay AUR :>
04:57 karolherbst: mupuf: mhhh, how likely is it, that there is no fan table?
04:58 mupuf: what do you mean?
04:58 angelsl: karolherbst: okay, what do i do now?
04:58 mupuf: is it there or not?
04:59 karolherbst: the only fan related thing I saw is this: 0x6899: 22 3c 64 id = 0x22, data = 0x643c -- fan_min = 60, fan_max = 100
04:59 karolherbst: and fan bump and fan slow
05:00 karolherbst: angelsl: as root: nvapeek 101000
05:01 angelsl: 00101000: 8040448a
05:03 mupuf: karolherbst: what gen?
05:03 mupuf: or chipset
05:03 angelsl: hm
05:03 angelsl: fan control works, i can set it to zero
05:03 angelsl: but, auto doesn't
05:06 karolherbst: gf108
05:06 angelsl: if i set pwm1_max to 10 and pwm1_min to 0 it seems to work well
05:06 angelsl: also, pwm1 = 10 and pwm1 = 100 has no audible difference
05:07 karolherbst: mupuf: seems like the min/max value is wrongly read out of the vbios?
05:07 karolherbst: brb, launch
05:09 mupuf: angelsl: you have a noisy fan, right?
05:09 angelsl: yeah
05:09 angelsl: ..by default at least
05:09 mupuf: yeah, I do not understand how I can detect that and clamp the values to a more usable range
05:09 angelsl: tweaking the values helps, but well
05:09 mupuf: it is annoying, to say the least
05:17 tagr_: mlankhorst: it should be in linux-firmware.git
05:25 karolherbst: angelsl: try min: 28 and max: 68
05:26 karolherbst: mupuf: but this should be easy to check, shouldn't it? I mean, the blob should respect those values and if not, it respects something else
05:26 mlankhorst: tagr_: ok excellent
05:27 mupuf: karolherbst: easy to check?
05:27 mupuf: we are not talking about the same bug here :ds
05:29 angelsl: karolherbst: still pretty loud
05:30 angelsl: the nvidia blob is practically silent
05:31 karolherbst: musca: ohh do we
05:31 karolherbst: mupuf: ^
05:31 karolherbst: musca: sorry
05:31 karolherbst: mupuf: well I thought nouveau just reads wrong min/max values for the fan
05:33 mlankhorst: tagr_: ah right :)
05:36 mlankhorst: wonder what isa xusb.bin has..
05:44 tagr_: mlankhorst: I don't think it's a secret that it's some sort of Falcon, wouldn't be surprised if it was closely related to the ones you know already
06:09 mlankhorst: yeah looked kind of familiar
06:30 kubast2: ok so now I have some free time
06:32 mlankhorst: tagr_: I have xusb.bin but fails to load with -ENOENT :(
06:32 mlankhorst: ah..
06:36 mlankhorst: needs firmware in initramfs
06:39 kubast2: After the voltage git https://github.com/karolherbst/nouveau/commit/dad43df25177179e1ee48aa49af041eafc52d2a3 ,this is the cat /proc/kmsg[the screen stops after 3 reclockings from performance 1 to 0 state ,the screen froozed.
06:40 kubast2: *sry no cat /proc/kmsg
06:40 kubast2: And no /proc/last_kmsg avalible
06:45 kubast2: Gonna do cat /proc/kmsg > kmsg.log ,and then do reboot pretending screen isn't froozed
06:55 RSpliet: froozed == frozen? or is this a verb that indicates visual corruption?
06:56 RSpliet: hopefully that system is still accessible through SSH, that might help obtain logs. Also, the magic sysrq sync sometimes does wonders
07:02 kubast2_: the screen doesn't change that's what I meant
07:03 kubast2_: It doesn't react to ctrl+alt+f2 etc.
07:03 kubast2_: it only happens under heavier load
07:03 RSpliet: ah, frozen it is
07:04 karolherbst: back
07:05 karolherbst: kubast2: than please use the nve7_vita_cell_voltage branch
07:05 karolherbst: then we have to look at reclocking itself for now
07:12 kubast2: ok I put the new module and created new initramsfs ,rebooting
07:18 kubast2_: What should I do now ?
07:21 kubast2: AC profile is a bit diffrent now "AC: core 324 MHz memory 648 MHz" instead of "AC: core 324 MHz memory 5000 MHz"
07:23 kubast2: ok got ssh running
07:24 kubast2: launching something gpu intensive
07:26 kubast2: ok so the profile changes from 07 to 0a when chainging window focus.
07:26 kubast2: *from opengl app to non-opengl app
07:28 kubast2-android: Core went to 255
07:28 kubast2-android: 0a pstate is active
07:29 kubast2-android: The screen frozed
07:29 kubast2-android: Ssh is running
07:29 kubast2-android: TRAP CH 9 in dmesg
07:30 kubast2-android: I will reboot pc and send you dnesg
07:30 kubast2-android: Dmesg
07:34 karolherbst: kubast2-android: it shouldn't change anymore with the other branch
07:34 kubast2: http://dpaste.com/2HXJ89T
07:34 kubast2: it did
07:34 karolherbst: yeah, well you still use the old branch
07:34 karolherbst: okay
07:34 karolherbst: anyway, it breaks after memory was reclocked
07:34 kubast2: git checkout -b branchname changes the branch right ?
07:35 kubast2: or
07:35 karolherbst: no
07:35 kubast2: it creates locally a new one
07:35 karolherbst: this creates a new branch from the current tree
07:35 kubast2: ok
07:35 karolherbst: named branchname
07:36 karolherbst: something is odd with the memory load :/
07:36 karolherbst: mupuf: I fear that we can't soly rely on the pdaemon counters for memory load :(
07:38 kubast2: well my isp gotta be kidding me
07:38 kubast2: 1024 bytes/s
07:39 AndChat|341361: Hope it will download somehow
07:44 kubast2-android: 70% of objects left
07:44 kubast2-android: With --depth 1
07:45 karolherbst: :/
07:45 karolherbst: kubast2-android: you don't have to clone again
07:45 kubast2-android: Well now I have to. Deleted the nouveau folder ,awhile later internet slowed down
07:46 kubast2-android: 1-7 KB/s
07:46 kubast2-android: I'm running irc on data so it downloads slightlly faster
07:47 kubast2-android: And I almost have no data left
07:47 kubast2-android: So I'm sunning irc on 8Kb/s
07:50 kubast2-android: Great "earlly EOF" at 93%
07:50 kubast2-android: Gonna connect my phone
07:51 mupuf: karolherbst: what else do you want to use?
07:53 kubast2-android: Ok cloned
08:05 kubast2: So it works now.
08:05 kubast2: Without the auto reclocking
08:06 karolherbst: kubast2: okay
08:06 karolherbst: then you could try out if you can reclock the pstates
08:06 karolherbst: it's in sysfs now again
08:10 kubast2: /sys/class/drm/card0/ where now
08:11 kubast2: it's definietlly not device/
08:12 karolherbst: kubast2: device/
08:12 karolherbst: kubast2: you need nouveau.pstate=1
08:13 karolherbst: imirkin: you saved me the troubles writing an own commnt :D
08:13 karolherbst: thanks you very much
08:13 imirkin: :)
08:14 karolherbst: weird though, he gets so much more perf than I do :/
08:14 imirkin: probably also worded more politically than you might have :p
08:14 imirkin: GTX 780 Ti is a monster...
08:14 karolherbst: yeah
08:14 karolherbst: my 770M is close to a 660 Ti only :/
08:15 karolherbst: ohhh no
08:15 karolherbst: I looked at the wrong benchmark
08:15 karolherbst: I thoguht he had like 120 fps in valley
08:16 kubast2: Ok the game loaded ,gotta load a map and then change pstates and see if it works
08:17 karolherbst: imirkin: something is really wrong with his setup though
08:17 karolherbst: imirkin: I bet the core doesn't even reclock at 0a
08:18 kubast2: ok loaded
08:18 kubast2: 30 fps 07
08:18 karolherbst: imirkin: 3 fps ultra full hd 8xmsaa on 06 and 11 fps on 0f
08:19 karolherbst: *07
08:19 kubast2: 70 fps 0a
08:19 kubast2: test 0f ?
08:19 karolherbst: kubast2: "should" work
08:20 karolherbst: imirkin: but he is right, memory doesn't do anything good for valley
08:21 AndChat|341361: 0f triggered the screen freeze
08:22 kubast2_: The screen freeze was very similar to previous ones
08:22 kubast2_: keyboard doesn't work ,apps don't react ,mouse movment works
08:23 karolherbst: mhhh
08:24 karolherbst: kubast2: which branch are you using now?
08:24 kubast2: the one you told me to
08:24 karolherbst: k
08:24 kubast2: "git clone https://github.com/karolherbst/nouveau -b nve7_vita_cell_voltage --depth 1"
08:24 karolherbst: k
08:25 karolherbst: kubast2: can you ssh into the machine when it happens?
08:25 kubast2: yes
08:25 karolherbst: there should be some stuff in dmesg
08:25 kubast2: ok gotta trigger the thing again
08:26 AndChat|341361: Ok got it
08:26 AndChat|341361: 0f triggered
08:27 AndChat|341361: Dmesg dumped rebooting
08:28 karolherbst: imirkin: uhh it is a nvf0 card :/
08:28 AndChat|341361: Weird
08:29 karolherbst: AndChat|341361: what?
08:29 AndChat|341361: Empty
08:29 AndChat|341361: Empty dmesg
08:29 karolherbst: mhh?
08:29 karolherbst: dmesg should never be empty
08:29 AndChat|341361: I did "dmesg > dmesg.log"
08:29 AndChat|341361: As root
08:29 AndChat|341361: It's empty when I "su root" too
08:30 karolherbst: what does plain dmesg give you?
08:30 AndChat|341361: Gonna do the thing again
08:30 AndChat|341361: And use sync
08:31 AndChat|341361: Trap ch 7
08:32 AndChat|341361: kwin_x11
08:32 karolherbst: AndChat|341361: complete dmesg please
08:32 AndChat|341361: In a sec
08:33 AndChat|341361: I have it dumped
08:33 kubast2: ok
08:34 karolherbst: imirkin: do you know what bothers me most? That 0f is actually faster with 07 core clock, that means that the core doesn't reclock for every pstate
08:34 kubast2: http://dpaste.com/1TZSKYP
08:34 karolherbst: imirkin: on my system 0f with lower core clocks is slower than 0a at full speed
08:35 karolherbst: kubast2: okay, that's because of the core then
08:37 karolherbst: kubast2: yeah, I think the core voltage is just wrongly set :/
08:39 kubast2: My gtx 650 is powered from the pcie 6xpin connector
08:41 imirkin: not that voltage :)
08:41 kubast2: the voltage of the core ,something like UV etc. in android
08:41 karolherbst: kubast2: open drm/nouveau/nvkm/subdev/volt/base.c
08:41 kubast2: *in general userspace voltage
08:42 kubast2: yup
08:42 karolherbst: you see that info.min += ret; in line 82?
08:42 kubast2: it's only on one line
08:43 kubast2: I don't have line counter in my editor
08:43 karolherbst: k
08:43 karolherbst: replace min with max
08:43 karolherbst: also in return info.min;
08:44 karolherbst: kubast2: this will run your gpu a little hotter, but as long as you stay at 07 it doesn't change a thing
08:44 kubast2: Changed only this 2 lines
08:44 kubast2: http://dpaste.com/0KQR376
08:44 karolherbst: yes
08:44 karolherbst: then recompile and install the module
08:45 karolherbst: after you reclock to 0a and 0f you should keep an eye on the temperature through sensors
08:45 kubast2: How can I do that
08:45 karolherbst: sensors is a command
08:45 kubast2: oh
08:50 kubast2: ok the sensor works[temp on 0a is +30*C ,on 07 + 29]
08:51 AndChat|341361: Black screen
08:51 AndChat|341361: Turned off pc just incase
09:01 kubast2: The screen wasn't white ,it was rather white this time
09:02 kubast2: Sending dmesg log in a sec
09:02 karolherbst: kubast2: next time reboot with nouveau.debug=debug
09:06 AndChat|341361: Internet is completlly down now
09:07 kubast2: Or not okey
09:12 kubast2_: http://dpaste.com/1VXHY07.txt
09:13 kubast2_: [ 157.613167] nouveau 0000:01:00.0: clk: setting performance state 2
09:14 kubast2_: dmesg was took right after chainging 0f state
09:15 kubast2_: If you need longer then I can supply you with one
09:15 karolherbst: its fine
09:22 karolherbst: kubast2_away: and what comes after it?
09:22 karolherbst: or does it work this way?
09:23 kubast2_away: the dmesg is right after chainging to 0f
09:23 karolherbst: ohh
09:23 karolherbst: so it is still running?
09:23 kubast2_away: The screen is quiet bright
09:23 kubast2_away: yeah
09:23 karolherbst: okay
09:23 karolherbst: then it should work
09:23 kubast2_away: I turn off the pc manually
09:23 karolherbst: just keepan eye on the temperature
09:23 karolherbst: but everything below 70 is okay
09:23 karolherbst: and even 90 °C wouldn't be dangerous normally
09:24 kubast2_away: so I should just run 0f for awhile with completlly white screen?
09:24 karolherbst: kubast2_away: nah, play something :D
09:24 kubast2_away: Well the screen gets completlly white
09:24 karolherbst: mhhh
09:24 kubast2_away: after chainging to 0f
09:24 karolherbst: then just change back to 0a
09:24 kubast2_away: well
09:24 karolherbst: ssh?
09:25 kubast2_away: I changed back to 07
09:25 kubast2_away: with ssh
09:25 karolherbst: ohh, still white?
09:25 kubast2_away: and it's still white
09:25 kubast2_away: yeah
09:25 karolherbst: try 0f agian :D
09:25 kubast2_away: lol
09:25 karolherbst: or switch to tty
09:25 karolherbst: I think the buffer may just be messed up or something
09:25 kubast2_away: hmm
09:26 karolherbst: this _can_ happen while switching pstates, because there is a bit still missing for memory reclocking
09:26 kubast2_away: gonna try ctrl+alt+f2[tty] ctrl+alt+f1[x session]
09:26 karolherbst: but it shouldn't always be happening
09:27 kubast2_away: ok
09:27 kubast2_away: catting pstates
09:29 kubast2: Nope ctrl+alt+f1 +f2 didn't worked
09:31 karolherbst: kubast2: then reboot and try again :)
09:31 kubast2_: I'm gonna put the script that change beetween 0f 07 each after few secounds
09:33 kubast2_: karolherbst, Are you sure it's safe?
09:33 kubast2_: Gonna do the trick with pstates chainging cople of times
09:34 karolherbst: kubast2_: well, usually it shouldn't mess up that bad
09:34 kubast2_: ok launching the script
09:35 kubast2: Hmm
09:35 kubast2: Takes quiet a while
09:35 kubast2: Set to 1s
09:35 kubast2: Ssh froozed
09:38 kubast2__: hmm gonna change dmesg to cat /proc/kmsg
09:39 kubast2__: and redeuce switches times
09:39 kubast2__: the last time I try this thing
09:39 karolherbst: kubast2__: well maybe you should first check if it still fails after reboot
09:40 karolherbst: usually the screen shouldn't turn just white on pstate change
09:40 kubast2__: you mean "reboot"
09:40 kubast2__: command
09:41 karolherbst: through ssh, yeah
09:41 kubast2: Turned black this time
09:42 kubast2: Still black
09:42 karolherbst: and there is nothing in dmesg?
09:43 kubast2: Wait gonna do it without a script
09:45 karolherbst: kubast2: check if switching without any monitor connected works
09:45 kubast2: Ok
09:46 kubast2: Same
09:46 karolherbst: kubast2: maybe you have to disable it in xrandr explicitly too :/ not quite sure
09:46 karolherbst: kubast2: after a clean reboot I meant
09:47 kubast2: The screen stays like that for quiet awhile
09:47 kubast2: Longer than usual reboot
09:50 kubast2: Ok did a clean reboot(by not chainging pstates)
09:51 kubast2: The pc isn't rebooting
09:51 kubast2: When the pstate was changed
09:51 kubast2: To 0f
09:51 karolherbst: did you try reboot?
09:51 karolherbst: as the command?
09:52 kubast2: Yes
09:52 karolherbst: mhh the driver my hang somewhere
09:54 kubast2: cat /proc/kmsg & > kmsg.log
09:55 kubast2: Hmm perhaps I will do that
09:55 karolherbst: trying to repair my mice: first conclustion, the screwes are pretty hard, broke my cheap screwdriver :D
09:55 kubast2: And then type in reboot
09:55 kubast2: First without chainging pstates
09:55 kubast2: Then with chainging pstates
09:56 kubast2: Ok
09:56 kubast2: Not like that
09:57 kubast2: The & should be after specifing stdout
10:00 kubast2: Ok so the first one is full of fan target request 30%
10:02 kubast2: Btw the secound one I will hard reset pc after a minute
10:03 karolherbst: kubast2: so clocking to 0f never succeeds without problems?
10:03 kubast2: Never
10:03 karolherbst: I guess easier to track down the problem hten
10:04 kubast2: Hmm
10:04 kubast2: Nothing else beside fan requests
10:04 kubast2: That comes after "reboot"
10:05 kubast2_: and the changed state
10:07 kubast2_: <6>[ 214.624862] nouveau 0000:01:00.0: volt: set 1137500uv: 0
10:07 kubast2_: <6>[ 215.141168] nouveau 0000:01:00.0: therm: FAN target request: 30%
10:08 kubast2_: http://dpaste.com/3RE7B3R partial kmsg ,full kmsg http://dpaste.com/14X60SB
10:17 karolherbst: kubast2_: are you on the machine through ssh?
10:17 karolherbst: try this: DISPLAY=:0 vblank_mode=1 glxgears
10:18 kubast2_: after 0f ok
10:18 karolherbst: I want to know if the card still works
10:18 karolherbst: and how much fps does glxgears print
10:18 kubast2_: wait
10:18 kubast2_: gotta install mesa-demos
10:18 karolherbst: do you have glxspheres installed?
10:19 kubast2_: it's in mesa-demos on arch
10:19 kubast2_: ok it works
10:20 kubast2_: gotta change 0f now
10:21 kubast2: Couldn't open display 0
10:21 kubast2: It worked
10:22 kubast2: Before I typed in 0f to pstates
10:22 kubast2: Wait
10:22 kubast2: Changed user
10:22 kubast2: Nothing on pc
10:22 kubast2: Nothing on ssh
10:23 kubast2: The same command "DISPLAY=:0 vblank_mode=1 glxgears"
10:23 kubast2: Previouslly displayed frames rendered in 5 sec
10:23 kubast2: And fps
10:23 kubast2: Now nothing
10:26 kubast2__: should I redo at check kmsg this time around ?
10:29 karolherbst: kubast2__: no
10:29 kubast2__: ok
10:30 kubast2: Ok so juicyssh executes command when I choose to close a session
10:30 kubast2: Good to know
10:57 karolherbst: kubast2: I ran out of ideas
10:58 karolherbst: kubast2: maybe you could mmiotrace the binary driver while it reclocks to the highest pstate
10:58 karolherbst: and then there is something different, but mhh
11:00 kubast2_: karolherbst, which repo I need to clone for mmiotrace?
11:01 kubast2_: envytool
11:01 imirkin: mmiotrace is built into the kernel
11:01 imirkin: you need demmio in the envytools repo to decode it, but not to create it
11:01 kubast2_: I have the repo cloned previouslly
11:02 kubast2_: compiled just incase nva
11:02 karolherbst: meh
11:02 karolherbst: why does the ubuntu page is 500ing now
11:02 karolherbst: *is...
11:03 kubast2_: no idea it works on mine pc
11:04 karolherbst: https://wiki.ubuntu.com/X/MMIOTracing ?
11:04 kubast2_: works
11:04 kubast2_: want me to wget this
11:04 karolherbst: no
11:04 karolherbst: it is for you
11:04 kubast2_: ok
11:04 karolherbst: so that you know what to do
11:14 kubast2: I don't have recovery mode ,turning off x login manager should do just fine?
11:15 karolherbst: nah, that you don't need to do
11:15 karolherbst: ohhh
11:15 karolherbst: wait
11:15 karolherbst: you have
11:15 karolherbst: kubast2: you have to enable the mmiotrace before loading the nvidia driver
11:15 karolherbst: that's a bit tricky to do
11:15 kubast2: What the ubuntu wiki tells me
11:16 kubast2: Is to turn off nvidia driver
11:16 kubast2: Then turn on mmio
11:16 kubast2: Do some 10s tests
11:16 kubast2: And check if the mmio log is avalible as a whole
11:18 kubast2: Ok not sure if it should still show output on monitor
11:18 kubast2: After turning off nvidia driver
11:19 kubast2: *I'm working in tty then?
11:20 karolherbst: kubast2: read again ;)
11:20 kubast2: Well I need to install grub on arch then
11:20 kubast2: Instead of systemd-boot ?
11:21 karolherbst: shouldn't matter actually
11:21 kubast2: Or create recovery entry
11:21 kubast2: Well the lsmod | grep nvidia
11:21 kubast2: Is now empty
11:21 kubast2: But tty is running and monitor output works
11:25 kubast2: Oh right
11:25 kubast2: It doesn't work now
11:25 kubast2: Since chainging to unstable arch repo
11:27 kubast2: * X doesn't work with nvidia driver
11:27 kubast2: because it tries to launch all the gpu drivers
11:28 pmoreau: It has always been trying every possible driver, and it still worked
11:28 kubast2: Hmm
11:29 kubast2: xinit: giving up
11:29 kubast2: There's xorg log
11:29 pmoreau: One thing that might cause troubles, is the new version of X which is not supported by your version of the NVIDIA driver
11:29 hakzsam: imirkin, hey, would you want to ACK those two patches for amd_perfmon on nv50? I would like to push them before the deadline :)
11:30 pmoreau: I had to rollback to X v1.17 for the NVIDIA drivers to work
11:30 imirkin: hakzsam: which 2 patches?
11:30 kubast2: How I can only roleback xorg?
11:30 imirkin: oh, the ones you just sent
11:30 hakzsam: yep
11:31 pmoreau: kubast2: `pacman -U /var/cache/pacman/pkg/{xorg-server-1.17.xxxxx,xorg-server-common-1.17.xxxx,etc.}`
11:31 imirkin: hakzsam: didn't the group stuff get removed? or was that something else?
11:31 hakzsam: imirkin, something else
11:32 hakzsam: imirkin, only group types have been removed
11:32 hakzsam: ie. CPU and GPU groups
11:32 pmoreau: kubast2: That's the idea. You need to do it for xorg-server, xorg-server-common, xf86-input-*, xf86-video-*, and probably xorg-server-devel
11:32 imirkin: hakzsam: ok
11:32 hakzsam: imirkin, now we don't care if it's a CPU or GPU groups
11:32 hakzsam: so, all groups are exposed
11:33 pmoreau: kubast2: For xf86-*, pick the 2nd latest version, and for xorg-server-*, the last 1.17.X
11:34 kubast2: Ok got it
11:34 pmoreau: And a reboot after the rollback seems necessary, as reloading isn't enough apparently
11:35 kubast2: Xfce4 starts so i
11:35 pmoreau: *reloading X
11:35 kubast2: Without reboot
11:35 kubast2: But I'm gonna reboot
11:39 kubast2: Hmm nvidia logo showed up
11:39 kubast2: I should start xinit commands after it?
11:41 pmoreau: You usually get the NVIDIA logo while X is starting, so I guess you're already in a environment with a running X server?
11:42 kubast2: Well I started xinit "sleep 10"
11:42 kubast2: And logo showed up after 10 secounds
11:42 kubast2: When I do it again after reboot
11:42 kubast2: It doesn't do that
11:43 pmoreau: What happens instead?
11:43 kubast2: Nothing
11:43 kubast2: It camed back to tty
11:43 kubast2: And when I did xinit again
11:43 kubast2: It fails
11:44 imirkin: kubast2: you're saying a ton of stuff, little of it makes sense to me... however note that nouveau can't display anything after blob loads. blob puts the display engine into a state we don't properly recover from.
11:44 imirkin: dunno if that helps you or not, but keep it in mind.
11:44 kubast2: Karol wanted mmiotrace
11:45 pmoreau: kubast2: Which error message do you get when it fails?
11:48 kubast2: Rebooted and it works now
11:49 kubast2: Launched glxgears thingy to force 0f/2 pstate/performance level
11:50 pmoreau: If you want to force the blob to go to the highest performance level, you can also set that from the NVIDIA X Server Settings.
11:59 hakzsam: imirkin, thanks
12:04 kubast2: Mouse movment should work out of the box ?
12:10 kubast2: My ~/.xinitrc is the same that the default one ,I added exec openbox-session ,because with twm it doesn't show anything
12:11 pmoreau: Did you rollbacked xf86-input-mouse?
12:11 kubast2: Yes
12:11 kubast2: Keyboard mouse
12:11 kubast2: xf86- all the things
12:11 pmoreau: Ok
12:13 karolherbst: mupuf: what is the state in ezbench for collecting power consumption ?
12:44 kubast2: I found something on arch bbs about xinit not launching keyboard ,I have a lot of things to prepare for school through ,at best I'll check if the change works and send mmiotrace on 24.11
12:46 frikinz: Hi. I'm experiencing some screen flickering. I'm actually not sure it comes from nouveau. How can I temporarily reboot without nouveau?
12:47 pmoreau: frikinz: Reboot with nouveau.modeset=0
12:48 frikinz: pmoreau: trying that thanks
12:50 imirkin: frikinz: what GPU, what type of monitor, and how is it connected?
12:54 frikinz: pmoreau: bingo
12:55 frikinz: VGA compatible controller: Intel Corporation Haswell-ULT Integrated Graphics Controller (rev 09)
12:56 frikinz: 3D controller: NVIDIA Corporation GF117M [GeForce 610M/710M/810M/820M / GT 620M/625M/630M/720M] (rev a1)
12:56 frikinz: 04:00.0 0302: 10de:1140 (rev a1)
12:58 frikinz: no, still flickers :)
12:59 imirkin: ok, so the nvidia gpu is actually unrelated to your issues right?
12:59 imirkin: usually '3d controller' is used when there are no outputs
13:00 imirkin: and GF117 doesn't have a display block at all
13:01 frikinz: imirkin: looks like not related.
13:02 frikinz: i915 then maybe
13:03 imirkin: yeah
13:03 imirkin: note that if you're using DVI, intel hw doesn't support dual-link
13:04 imirkin: however with an amalgamation of adapters it's possible to defeat its detection of the situation, but that doesn't make the hw magically able to support it
13:04 frikinz: I'm on a laptop using the integrated screen
13:04 frikinz: you're talking chineese to me :)
13:04 imirkin: ah. and said internal screen is flickering? you probably want #intel-gfx
13:04 frikinz: imirkin: thanks a lot
13:05 frikinz: that's the integrated GPU right? I thought only one of the two is used at a time <-- noob insude
13:05 imirkin: the situation i'm talking about can happen if you plug an hdmi <-> dvi adapter into a hdmi port and use a long dvi cable to hook into your monitor.
13:05 imirkin: frikinz: http://nouveau.freedesktop.org/wiki/Optimus/ -- read the soft mux section
13:05 imirkin: or rather... ignore what it says about hard muxes - you have a soft mux
13:06 frikinz: I see
14:24 frikinz: Well.. not so much. Haven't figured it out but at least I got a temp solution: reverting to kernel 3.16. Good old times..
14:25 karolherbst: frikinz: wanna bisect the kernel? :D
14:25 imirkin: frikinz: you should let the intel folk know
14:25 frikinz: karolherbst: thought about it :)
14:25 imirkin: frikinz: it's unlikely to fix itself on its own
14:26 frikinz: I've asked in intel-gfx
14:26 imirkin: did they say anything useful?
14:26 frikinz: imirkin: either got scared or spotted the noob. Got no answer :)
14:27 imirkin: file a bug at bugs.freedesktop.org
14:27 imirkin: DRM/i915 under... something. not sure. maybe there's a DRM top-level project.
14:40 frikinz: And.. the newest debian kernel (4.3) stops with an oops. that would have been too easy
14:49 frikinz: Someone has bisected already https://bugs.freedesktop.org/show_bug.cgi?id=91393#c25 , looks very similar. the dev reported today he's still investigating.
14:49 frikinz: thanks for the help guys
14:49 imirkin: you don't have a bdw
14:50 imirkin: you have a hsw
14:50 frikinz: oh
14:53 frikinz: I have hundreds of acronyms in my job but.. not these ones. Anyway....
14:53 imirkin: bdw = broadwell. hsw = haswell
14:54 frikinz: Ok
14:57 karolherbst: any idea what "fifo: SCHED_ERROR 08" is?
14:58 karolherbst: imirkin: fifo: PBDMA0: 00040000 [PBENTRY] ch 2 [00bf890000 WL2[3780]] subc 0 mthd 0000 data 00000000
14:58 karolherbst: any idea?
14:58 karolherbst: hit this a few seconds ago
14:58 imirkin: yeah. "shit's messed up, yo!"
14:58 karolherbst: :D
14:58 karolherbst: I thought that much
14:58 karolherbst: there is also
14:58 karolherbst: gr: TRAP ch 2 [00bf890000 WL2[3780]]
14:58 karolherbst: and gr: DISPATCH 80000002
14:59 karolherbst: and ILLEGAL_CLASS ch 2 [00bf890000 WL2[3780]] subc 0 class 000b mthd 0360 data 20164010
14:59 imirkin: yeah, some pushbuf got interrupted at an unfortunate moment and then the whole command stream got confused.
14:59 karolherbst: mhh okay
14:59 karolherbst: why can this happen?
15:00 karolherbst: this shouldn't be caused by reclocking, because the last reclock was some minutes ago
15:00 imirkin: nfc
15:01 karolherbst: k
15:02 karolherbst: imirkin: mhhh glxgears still works, glxspheres not
15:04 karolherbst: killall -SEGV kwin_x11
15:46 karolherbst: did anybody here used my pcie patches for their "main" machine?
15:48 koz_: karolherbst: Why, should we not have?
15:52 karolherbst: koz_: because they aren't mainlined yet
15:53 karolherbst: just wanted to know who is using them
15:55 koz_: Ah.
15:55 koz_: I'm currently gonna see if the latest Linux kernel's Nouveau stuff finally fixes my reclocking problem.
15:56 koz_: (meaning Linux 4.4)
15:57 karolherbst: koz_: what kind of reclocking problem?
16:00 koz_: karolherbst: Core clock doesn't move.
16:00 koz_: Memory reclocks fine, but core clock is stuck at like 1/3 normal.
16:00 koz_: This has been an issue for me for about 5+ kernel releases now.
16:01 imirkin: koz_: if you have a pwm voltage situation, make sure to also apply http://cgit.freedesktop.org/~darktama/nouveau/commit/?id=a539a9802af38236101de6ae8d9b827544de0f72
16:02 koz_: imirkin: Is this not included yet?
16:02 koz_: And yeah, it is a voltage issue.
16:02 imirkin: not in 4.4
16:02 imirkin: it will be... eventually
16:02 karolherbst: koz_: could I have your vbios?
16:02 imirkin: but ben didn't send a fixes pull
16:02 koz_: karolherbst: I think I reported this already. Let me find the issue..
16:06 koz_: https://bugs.freedesktop.org/show_bug.cgi?id=88890
16:07 karolherbst: I actually answered ther :O
16:07 koz_: Lol.
16:07 koz_: So maybe that patch could help.
16:07 karolherbst: the vbios will tell
16:08 karolherbst: nope
16:08 karolherbst: not PWM based
16:08 karolherbst: so the patch won't help
16:08 koz_: Alrighty.
16:08 karolherbst: imirkin: I thought pwm stuff is there for 4.4?
16:08 koz_: Continuing along previous plan, then. :P
16:08 imirkin: karolherbst: it is. just that one fix isn't... yet.
16:08 karolherbst: I see
16:09 karolherbst: well I have a fix for this voltage issue anyway
16:09 karolherbst: mupuf: third card
16:09 imirkin: is this the no-rammap issue?
16:09 karolherbst: koz_: you need this: https://github.com/karolherbst/nouveau/commit/a7c625c6be45bc790d47e58c3d403a3725fa6f43
16:09 karolherbst: imirkin: 0ed voltage table header
16:09 karolherbst: -- Mode GPIO, Base voltage 0 µV, voltage step 0 µV, acceptable range [0, 0] µV --
16:10 imirkin: ah right
16:10 koz_: karolherbst: Is this not in nouveau either?
16:10 karolherbst: koz_: no
16:10 karlmag: karolherbst: "thirds card".. is that how many graphics cards you have broken during this endevour? :-P
16:10 karolherbst: koz_: this is in a "just a guess" state
16:10 koz_: karolherbst: Ah.
16:10 koz_: So should I build your fork of nouveau then?
16:10 karolherbst: but this guess makes sense currently
16:11 karolherbst: koz_: https://github.com/karolherbst/nouveau/tree/nve7_vita_cell_voltage this branch please
16:11 karolherbst: not the default one
16:11 karolherbst: imirkin: my default one is in a highly "no_touchy" state now ;)
16:11 karolherbst: *high
16:11 koz_: karolherbst: So is it just the usual 'configure, make, sudo make install' dealio?
16:11 imirkin: koz_: cd drm; make
16:12 imirkin: it's an out-of-tree kernel module
16:12 koz_: imirkin: I've never actually built one of those before.
16:12 imirkin: you could do make modules_install with it
16:12 koz_: Nor installed.
16:12 imirkin: but it's same idea as with the kernel
16:12 imirkin: or you can just manually copy the nouveau.ko into whatever the right place is for your setup
16:13 koz_: Hmmm.
16:13 koz_: I guess I can try.
16:14 koz_: So I just remove my old nouveau and run sudo make modules_install?
16:14 koz_:really needs to be ELI5ed this.
16:14 imirkin: it'll overwrite it ;)
16:14 koz_: I'll remove it anyway, just in case.
16:15 imirkin: it'll build against your running kernel by default
16:15 koz_: That works for me.
16:15 imirkin: but that running kernel has to be compatible... you might need to upgrade. tbd :)
16:15 koz_: I'll just let this kernel finish building before I try that.
16:15 imirkin: i think right now it wants 4.4-rc1
16:16 koz_: Well, that's what I'm building.
16:16 imirkin: although you could revert some stuff and make it build against 4.3
16:17 karolherbst: imirkin: nah, my branches are compatible with the stuff I run and this is usually no rc kernel
16:19 karolherbst: koz_: if you run a 4.3 kernel currently, this is fine
16:19 karolherbst: just build the nouveau module and install it
16:19 koz_: karolherbst: I'm on 4.2.5 right now.
16:20 koz_: So it means a rebuild anyway - might as well take 4.4rc1 to improve my odds.
16:20 karolherbst: koz_: I never test anything against 4.4 :/
16:21 karolherbst: koz_: you would need this commit then: http://cgit.freedesktop.org/~darktama/nouveau/commit/?id=d55c3200d876032f6149c83045eea96c05472dc9
16:21 karolherbst: but I can create a branch for ya
16:21 koz_: karolherbst: If you would, that'd be great.
16:21 koz_: I'm a noob to all this.
16:22 koz_: My kernel will be building for a little white anyhow.
16:22 karolherbst: koz_: branch name is koz_
16:22 karolherbst: :D
16:24 koz_: karolherbst: THanks! Link?
16:24 koz_: (and I assume I'd just clone and build that?)
16:24 karolherbst: koz_: git clone https://github.com/karolherbst/nouveau.git -b koz_
16:24 karolherbst: ohh wait
16:25 karolherbst: add --depth 1
16:25 karolherbst: git clone https://github.com/karolherbst/nouveau.git -b koz_ --depth 1
16:25 koz_: OK.
16:25 karolherbst: this will clone it much faster
16:25 koz_: I'll just do that once my kernel compiles.
16:25 koz_: Also, has there been any Maxwell-related news?
16:26 imirkin: secure firmware load support was sent out for GM20B, but it is, allegedly, insufficient for desktop gpu's. also still no firmware.
16:27 koz_: imirkin: So basically 'nothing new'. :(
16:27 imirkin: "little new"
16:27 koz_: Well, we can only hope.
16:28 karolherbst: it's a bit sad somehow, because in the end, less has to be done on the nouveau side :(
16:29 koz_: karolherbst: It is, but unfortunately, the speed of this decision rests entirely on NVIDIA.
16:29 koz_: This gives me little hope.
16:31 imirkin: karolherbst: you say that with little regard to the 3d driver :)
16:31 imirkin: although i guess strictly speaking, yeah, "less", since we'd have to do all the other stuff _anyways_
16:32 koz_: Well, for my purposes, Keplers will do fine for a while.
16:32 karolherbst: imirkin: right
16:32 koz_: I still have about 4 upgrades to go first. :P
16:32 karolherbst: koz_: first gen maxwell is also nice
16:33 koz_: karolherbst: But it won't work with Nouveau...
16:33 karolherbst: koz_: why not?
16:33 imirkin: it will... just not very well
16:33 koz_: imirkin: Which cards are those?
16:33 karolherbst: koz_: gm10x
16:33 imirkin: GM107
16:33 koz_: I thought *all* Maxwells needed signed firmware.
16:33 imirkin: nah, only GM20x
16:33 koz_: Let me just check what cards those are.
16:34 imirkin: generally speaking, GTX 950+
16:34 imirkin: and GM107 is GTX 750 (Ti)
16:34 imirkin: but there are various variations, esp when getting into mobile chip names
16:34 karolherbst: 1gen maxwell for mobiles are all crappy cards
16:34 koz_: imirkin: I have a 840M in my laptop actually.
16:35 koz_: I've never gotten Nouveau working on it.
16:35 karolherbst: imirkin: the "maxwell" 860M has only aroun 65% compared to the kepler 860M :D
16:35 imirkin: what is that?
16:35 imirkin: 840M = GM108?
16:35 karolherbst: yes
16:35 koz_: GM107 according to the matrix.
16:35 karolherbst: koz_: mhh wiki says gm108
16:35 imirkin: koz_: gotta go according to lspci :)
16:35 koz_: I see.
16:36 koz_: GM108 indeed.
16:36 koz_: Someone should probably correct that. :P
16:36 imirkin: if there is an overriding logic to nvidia's names, it's certainly not apparent
16:36 karolherbst: :D
16:36 koz_: There isn't, I'm pretty sure.
16:36 imirkin: nah, single marketing model name maps to multiple chips
16:36 imirkin: even multiple gens
16:36 koz_: How 'not well' are we talking for GM107?
16:36 koz_: Like, does it not reclock at all or soemthing?
16:37 karolherbst: I was thinking about "upgrading" my gpu, but well, there aren't many choices :/
16:37 imirkin: koz_: there's no reclocking, but that's the least of your concerns
16:37 koz_: Ah.
16:37 koz_:will stick to Keplers.
16:37 karolherbst: kepler is best :)
16:38 karolherbst: sadly there are still too many issues for now
16:38 karolherbst: but we are getting there
16:38 koz_: Well, I just wish I could help, but I don't have the knowledge.
16:38 karolherbst: koz_: neither did I 3 months ago :p
16:38 karolherbst: or is it 4 months ago already?
16:38 karolherbst: no clue
16:39 koz_: karolherbst: I'm at the start of a PhD, which makes this kinda learning a bit hard. :P
16:39 karolherbst: D:
16:39 karolherbst: right :D
16:39 karolherbst: you gotta now what's more important for ya
16:39 karolherbst: *know
16:39 koz_: I'm at least glad I can test your changes.
16:40 karolherbst: imirkin: I really would like to see the vbios of this 780 ti card :/
16:40 koz_: What's the mightiest Kepler Nouveau supports right now?
16:40 karolherbst: koz_: mhhh, well mine kepler is a bit mighty already
16:40 koz_: (like, raw-numbers-wise)
16:40 karolherbst: GK106 should be in a good state
16:40 imirkin: karolherbst: good luck. he'll never give it to you.
16:40 karolherbst: :(
16:40 koz_: 650-Ti?
16:40 karolherbst: I know
16:41 karolherbst: koz_: nah, that's a slow one :p
16:41 imirkin: karolherbst: perhaps Tom^ is around and has a copy of his vbios sitting around somewhere?
16:41 karolherbst: koz_: currently I have a 770m
16:41 koz_: karolherbst: I meant a desktop card.
16:41 koz_: Ms are for laptops IIRC.
16:41 karolherbst: koz_: desktop is in a worse state sadly
16:41 karolherbst: there is something fishy with the memory reclocking thing
16:42 koz_: Well,then let me rephrase - which is the mightiest desktop Kepler Nouveau supports right now?
16:42 karolherbst: something display related or so
16:42 karolherbst: koz_: well the phoronix guy ran the 780 Ti
16:42 karolherbst: but he got core clocks issues
16:42 koz_: Same as me.
16:42 karolherbst: :O
16:42 koz_: Except I'm on a 650.
16:42 karolherbst: :(
16:42 karolherbst: I thought you had a 780 ti hidden somewhere
16:42 karolherbst: .(
16:42 koz_: I only fucking wish.
16:43 koz_: I don't have a spare computer price lying around. :P
16:43 koz_: I wonder how easy it would be for me to search out laptops with 770Ms...
16:43 karolherbst: koz_: not, these are usually beasts
16:43 karolherbst: and not worth the costs
16:44 karolherbst: except when you are too lazy to have two computers
16:44 koz_: I have two computers.
16:44 karolherbst: see
16:44 karolherbst: no need for a 770m
16:44 koz_: I just want my 650 reclocking properly.
16:44 karolherbst: yeah
16:44 koz_: That'd make me a happy bunny.
16:44 karolherbst: test my patch
16:44 karolherbst: :p
16:44 koz_: Will do - once the damn kernel builds. :P
16:45 koz_: If this doesn't work, I might just fork out the NZ$200 or so I need for a 760 and try again. :P
16:46 karolherbst: koz_: well with my card I am able to play stuff like bioshock infinite on full hd :p
16:46 karolherbst: with nouveau
16:46 koz_: I'm struggling to get Minetest to give me consistent 60fps on 1920 x 1080 on mine.
16:46 karolherbst: ah
16:46 karolherbst: nah
16:46 karolherbst: I don't hit 60 fps
16:46 karolherbst: it is playable alright
16:46 karolherbst: but not 60 fps
16:46 koz_: 30?
16:46 koz_: Even that'd be nice.
16:46 karolherbst: mhh round about
16:46 karolherbst: at ultra settings I mean
16:47 karolherbst: :D
16:47 koz_: Since as-current, I get drops to like, 7.
16:48 karolherbst: imirkin: there are 43 vbios of a 780 Ti on this collection side :/ insane
16:48 imirkin: hehe
16:49 karolherbst: maybe I find his
16:49 karolherbst: mhhh
16:49 karolherbst: strange
16:49 karolherbst: his gpu clocks are too high
16:49 koz_: Overlocking?
16:49 karolherbst: +50%?
16:50 karolherbst: okay, 40%
16:50 karolherbst: still
16:50 karolherbst: 780 Ti design is 876
16:50 koz_: Yeah, that's surprising.
16:50 karolherbst: and his has 1202MHz
16:50 karolherbst: that's a bit, mhh
16:50 karolherbst: strange
16:50 koz_: Yeah, they's some sick gains.
16:53 koz_: Damn kernels take ages to compile...
16:53 karolherbst: koz_: make -j8
16:53 karolherbst: :p
16:53 koz_: Yeah, even then.
16:53 imirkin: also don't build distro kernels
16:53 imirkin: their kernels have to work on every machine. yours just has to work on yours :)
16:54 karolherbst: imirkin: don't know his gpu looks a bit crazy, maybe that's some over OCed stuff nvidia gives to bechmarkers to get higher results, but sells crap in the end?
16:54 karolherbst: no idea
17:02 karolherbst: ohh nice
17:03 karolherbst: mupuf: got the vbios to be read by nvbios from the techpowerup site
17:04 karolherbst: I had to cut around 0x550 from the begining
17:06 karolherbst: mhhhh
17:06 karolherbst: this is strange
17:06 karolherbst: imirkin: the site says like 1085 MHz GPU clock, but the vbios says 2822 i it's highest pstate
17:07 karolherbst: even divided by 2 it is too high
17:07 koz_: Goddamn it kernel are you serious right now...
17:07 koz_: (build worked, install failed...)
17:08 karolherbst: .......
17:08 karolherbst: shit
17:08 koz_: Oh well, I'm going for a walk. Will report my findings as soon as I can.
17:08 karolherbst: those gk110 vbios are different
17:08 karolherbst: mehhhh
17:08 imirkin: fwiw i've never ever tried 'make install' on a kernel
17:08 imirkin: i always assume it'll end the universe :)
17:08 karolherbst: :D
17:09 karolherbst: imirkin: vdec freq = 3500 MHz and memclk freq = 988 MHz for 0f pstate
17:09 karolherbst: sounds about right :D
17:15 karolherbst: koz_: why don't you just install a 4.3 kernel from your distribution?
22:41 koz_: When I try to build nouveau from git, I get this: http://paste.rel4tion.org/111
22:41 koz_: What am I missing/doing wrong here?