02:16oygle: Video problem on Kubuntu running Nouveau display driver vers 126.96.36.199
02:16oygle: *-display description: VGA compatible controller product: G96 [GeForce 9400 GT] vendor: NVIDIA Corporation physical id: 0 bus info: pci@0000:01:00.0 version: a1 width: 64 bits clock: 33MHz capabilities: pm msi pciexpress vga_controller bus_master cap_list rom configuration: driver=nouveau latency=0 resources: irq:28 memory:fd000000-fdffffff memory:d0000000-dfffffff memory:fa000000-fbffffff ioport:bc00(size=128) memory:fe880000-fe8fff
02:16imirkin: what's the issue?
02:16oygle: The display for any 'window' activity is very poor, sometimes black, or parts of grey, and then finally some words appear. Trying to use the mouse results in nothing, video problems; eventually is does display. The main problem is video - VERY SLOW.
02:17imirkin: pastebin dmesg and xorg logs
02:17oygle: Today, others windows are not visible. I need to go to another firefox window, then back again, to actually sse thigs. Lots of flickering
02:20oygle: xorg.log at https://paste.ubuntu.com/23405303/
02:21imirkin: hmm.. that generally seems quite happy
02:22imirkin: when you say video is slow do you just mean all output, or playback of videos in some movie player application?
02:23imirkin: either way, dmesg should also be informative, please pastebin that
02:26oygle: dmesg is at https://paste.ubuntu.com/23405314/
02:27imirkin: no errors. no issues. no nothing. :(
02:29oygle: sorry, I should say 'display' not video. The computer is 8 yrs old and when I did a fresh install of Kubuntu 16.04.1, the display side of things was very slow. Now today, when scrolling in Firefox, lots of flickering and jumping. Plus many (child) windows do not display the full contents. Plus when a window is closed I can still see it for about 5 seconds. Could it all be 'resources' as this computer is old, but 4 gb ram and decent v
02:31imirkin: oh, it could be that all of a sudden things are looking more to GL for acceleration
02:31imirkin: and without reclocking the GPU's not up to it? dunno
02:31imirkin: that doesn't explain the issue though - esp misrendering
02:32oygle: am running exactly the same vers of Kubuntu on a laptop (new, dell) and no display problems.
02:32imirkin: but you're going to have to do a little more legwork to figure out what's wrong
02:34imirkin: try finding nouveau_dri.so and moving it out of the way (e.g. rename it to any other name)
02:34imirkin: and does that fix all your issues
02:40oygle: Hmm, ..definitely a big improvement. Not much flickering now.
02:40imirkin: now you're running without any hw GL impl
02:40imirkin: (it'll still end up using llvmpipe if something really wants GL, which is a software rasterizer)
02:44oygle: Okay thanks. Flickering and jumping back again (sigh), plus child windows don't appear sometimes.
02:45imirkin: the DDX should be pretty solid
02:45oygle: akonadiserver using 25% of cpu and have been having resource issues with KMail/akonadi - so maybe just running out of reosurces ?? I don't know.
02:46imirkin: yeah sorry, dunno
02:46imirkin: i don't use KDE
02:46imirkin: a number of people have also reported issues with KDE/QT5 + nouveau in recent times, although those tend to be issues with concurrency
02:48oygle: Okay thanks for your help. I assume it will be a waste of $$ to buy a newer video card. The mobo is asus.
02:49imirkin: you should be able to obtain a different PCIe GPU without issue
02:49imirkin: i recommend AMD over NVIDIA
02:50imirkin: it has much better support on linux
02:51oygle: Okay, I will see about an AMD video card for the asus - https://www.asus.com/Motherboards/P5Q_PRO/specifications/
02:53imirkin: pretty much any GPU should fit
02:57oygle: okay thanks for your help "imirkin" :)
03:01imirkin: feels like the quotes were in the wrong place ... maybe "help" ...
04:25imirkin: skeggsb: thoughts on https://bugs.freedesktop.org/show_bug.cgi?id=97620 ?
04:26imirkin: skeggsb: should we be masking various VBIOS reg writes based on GPC availability?
10:31hakzsam_: karolherbst: in nvc0_query_sw.c
10:38karolherbst: hakzsam_: yeah, that I know. I hacked something together, but the variable also contains odd things, so I ended up with garbage
13:33imirkin: gnurou: perhaps you know if we're supposed to skip some stuff while executing vbios? some vbios's write to GPC's that aren't there.
15:05imirkin_: gnurou: alternatively, should we just revert a6a0f67ca7aae2e6bec7ebf55d1e4853dc220816 ?
15:07imirkin_: hakzsam_: should you get emil to backport that IMAD fix as well?
15:07hakzsam_: he forgot to backport it I guess
15:07imirkin_: did you not cc stable?
15:08hakzsam_: I added the cc tag
15:08imirkin_: i'll ping him
15:15notme: is anyone there?
15:38RSpliet: notme: don't ask to ask, just ask
15:40notme: I am trying to configure a fresh install of kali linux 2016.1 - I have nvidia 1060 gpu and the default nouveau drivers only return 'unknown display' in settings
15:40notme: is there any way to use nouveau drivers and get my 1920x1080 output to work?
15:41imirkin_: notme: nouveau should be totally ineffective on your GPU - i don't think nouveau is driving your display in any way
15:43notme: ok. I am currently reinstalling and will try to get nvidia proprietary drivers working again. I am reinstalling because last time I got the proprietary drivers installed I could not boot properly. Is there a fix for this or does every machine need a unique fix?
15:43imirkin_: notme: talk to your distro suppliers. nouveau does absolutely nothing on your gpu.
15:44imirkin_: notme: and if you're looking for good open-source support, buy AMD
15:49hakzsam_: RSpliet: I don't remember which version of blob you use for tracing compute programs, can you tell me again?
15:50hakzsam_: I downgraded to 367, but it doesn't work for me...
15:52pmoreau: 367 or 364?
15:52hakzsam_: probably 364
15:52hakzsam_: but it needs linux<4.6
15:52hakzsam_: what a pain
15:58hakzsam_: okay, works fine with 364 at least
16:26hakzsam_: pmoreau: ^
16:26pmoreau: hakzsam_: Great! I’ll have a look!
16:26hakzsam_: similar to kepler, except warpid which uses one more bit
16:27hakzsam_: now, I will fix up my kernel and hope :)
16:28pmoreau: They really need that many bits for SM.id? o.O
16:28imirkin_: GM200 :)
16:28pmoreau: I don’t remember seeing that many SMs on the Titan X…
16:28hakzsam_: well, they used as many bits for kepler as well
16:29imirkin_: but more seriously, just coz those bits are extracted doesn't mean they're ever set...
16:29pmoreau: sure, but I would have expected warpid to have more bits than smid
19:13karolherbst: mupuf: I don't get what you have against the way I did the power budget stuff. There is just one 0xff check in the iccsense subdev and I just use the parsed values in hwmon. Just reading out the entries will result in interpreting issues later. Also the max and crit power value might not be just a simple table lookup. Anyway, we cache the max value for the voltage the same way. I would like to have those values be const, but
19:13karolherbst: meh... there are two sane approaches to this: have a function returning those values or store them in the subdev, but I guess you prefer the function thing?
19:14karolherbst: the we would have nvkm_iccsense_power_cap_get_max and nvkm_iccsense_power_cap_get_crit
19:57nuovolnx: Excuse me, I have installed nvidia propetary driver on my debian 8.6 64bit because i have got a problem with my nvidia gt 320
19:58nuovolnx: can I use a open source driver? The my video card is supported by nouveau?
19:58imirkin_: nuovolnx: should be... relatively supported.
19:59nuovolnx: imirkin_, with 3d?
19:59imirkin_: sort of
19:59imirkin_: expect hangs.
19:59imirkin_: works better for some people than others
19:59imirkin_: no real way to tell without trying
19:59mupuf: karolherbst: I do not feel too strongly about introducing this cache
19:59mupuf: I just find it ... inelegant
20:00mupuf: and contrary to other parts of nouveau
20:01nuovolnx: now I have this problem
20:01nuovolnx: I have installed nvidia driver from nvidia site. driver.run
20:02nuovolnx: now when i run nvidia-settings as root, it works fine
20:02imirkin_: you won't get much help here for nvidia blob. try #nvidia or their support forums.
20:02nuovolnx: instead when i run it as user it don't works fine
20:03towo`: installing things by bypassing the package-management is a stupid idea
20:04nuovolnx: towo`, I don't expert of linux, I have heard that amd card is supported automatically by linux
20:04towo`: sure, amd has real good opensource support
20:05towo`: but debian provides the blob as debian packages
20:05towo`: and one should use that instead of clueless starting some binary installer
20:06nuovolnx: as you told you I'm not expert of linux
20:06towo`: then you should learn tha basics
20:07nuovolnx: I tried to install with apt-get from nonfree
20:07towo`: but again, not a topic of this channel
20:07nuovolnx: but for me it is difficolt to configure the headers
20:08towo`: there is nothing to configure
20:08nuovolnx: ok thank you so much, goodbye
20:08towo`: the headers has to be installed for the running kernel, nothing more
20:09nuovolnx: I don't now the headers. I read header from guide howto
20:09nuovolnx: for me is difficult to install a driver
20:09nuovolnx: in linux
20:10nuovolnx: I will try on #nvidia channel
20:11karolherbst: mupuf: I see
20:11karolherbst: mupuf: for me it doesn't feel right to interpret the vbios table entries in multiple places, so for that there should be at least a function
20:12mupuf: yes, absolutely
20:13karolherbst: painful part is, with every version the entries got like 12 bytes bigger...
20:13karolherbst: they are insanle huge on maxwell
20:14karolherbst: the entry size on your gm206 is 0x2e
20:15karolherbst: 6 looks like your cap entry though
20:15karolherbst: mupuf: remember the max power cap of your gm206 in nvidia-smi? 130W?
20:15mupuf: nope, I dont :s
20:15karolherbst: would be byte 0xa in the header
20:15mupuf: yeah, there are likely more thresholds and more behaviour
20:16karolherbst: it would make sense, if they would contain anything besides 0s
20:16karolherbst: it's like the three power values, 0xf is 0x0c and byte 0x29 is like 0x01
20:17karolherbst: what is currently plugged into reator?
20:18karolherbst: yeah, it will be 130W for your gm206
20:18karolherbst: and crit is 150W
20:18karolherbst: and header 0xa is the field for that entry
20:19karolherbst: moved from 0x9 to 0xa, so yeah (ver 0x20 vs 0x30
20:22karolherbst: mhh, skeggsb gm206 has 150W and 216W according to this...
20:23karolherbst: Sophira: mind checking something for me on your gm204 with the propritary driver?
20:30Sophira: karolherbst: I'm unfortunately in the middle of something right now :/ I'd also need to set the proprietary driver up as I'm only using nouveau at the moment. I can do that, but not right now, sorry :/
20:32karolherbst: Sophira: k, no worries then, maybe I find somebody else who is already running nvidia
20:32karolherbst: Lekensteyn: I think you have a gm204 optimus setup?
20:33vitis: Hello i was wondering if anyone want some tests on Fermi codename NVC1 i have optimus laptop with nvidia gt635m
20:33karolherbst: uhh I know a good channel for that kind of question I have :D
20:33karolherbst: vitis: ask RSpliet
20:34karolherbst: mhh, or maybe not, my question actually requires that I have the vbios...
20:34karolherbst: Yoshimo: you also have a gm204, right?
20:35Yoshimo: 980, 560TI and a 285 indeed
20:35vitis: oh and i also have other laptop with 940m GM108
20:35karolherbst: Yoshimo: running nvidia on the 980?
20:35Yoshimo: unfortunately yes, for the time being
20:36karolherbst: vitis: has to be gm20x, because of the newer stuff in the vbios, it is really rare for a gm108 to have what I look for
20:36karolherbst: Yoshimo: perfect
20:36karolherbst: I need some nvidia-smi info
20:37Yoshimo: karolherbst: you did not have any new ideas about pmu on the gm2xx series so far, did you?
20:38Lekensteyn: karolherbst: I think so
20:38karolherbst: Yoshimo: nvidia-smi -q -d POWER
20:38karolherbst: Lekensteyn: same for you, but then under optirun ;)
20:39Lekensteyn: I've no blob atm, would have to reboot and retry for that
20:39karolherbst: uhh, what kind of system is that :O I can switch between nouveau and nvidia in 5 seconds! :D
20:40Yoshimo: brb, have to reboot
20:40Lekensteyn: lockdep :)
20:40Lekensteyn: brb travelling, let me know if I can grab more details at once
20:40karolherbst: just that
20:47Yoshimo: karolherbst: http://pastebin.com/MHxduCvW
20:55karolherbst: Yoshimo: awesome, thanks. Exactly what I expected
20:55Yoshimo: what are you looking for?
20:55karolherbst: power limits
22:08karolherbst: okay, now I have to find that silly entry fot the ver 0x10 entries
22:10karolherbst: mupuf: ever saw a table being always empty for a specific version?
22:11karolherbst: nvm, found some fermis
22:13karolherbst: lovely, how I like this...
22:13karolherbst: I can really hear it already: let's stop that bs we did up to fermi and start makeing _sane_ vbios tables with kepler
22:28Lekensteyn: karolherbst: do you still need my nvidia-smi output?
22:38karolherbst: Lekensteyn: yes
22:39karolherbst: Sophira: if you find some time, it would be awesome to get the same from you "nvidia-smi -q -d POWER", isn't time critical, so whenever you got time is fine
22:40karolherbst: Lekensteyn: yours is especially interesting, cause that byte isn't set I am itnerested in
22:43Lekensteyn: karolherbst: ok, let me send one patch and then I'll reboot
23:25Lekensteyn: karolherbst: is running under X important? I get N/A or Not Found for all except Power Draw
23:27karolherbst: Lekensteyn: expected, kind of
23:27karolherbst: that byte isn't set on yours
23:27Lekensteyn: this is driver version 375.10, I have only nvidia kernel module loaded
23:27karolherbst: sad, I kind of hoped nvidia would have a fall back for it...
23:27karolherbst: obviously, it hasn't
23:28Lekensteyn: in case you are interested, here is the full nvidia-smi -q output: ix.io/1Bfv
23:29Lekensteyn: is it worth trying to change a setting in the BIOS? (MSHYBRID -> Discrete)
23:33Lekensteyn: karolherbst: what is that byte?