00:01 cyberpear: is nouveau-firmware from the rpmfusion-nonfree-tainted repo related to this project? -- is it likely to be useful to me trying to run my Pascal gp107 card w/ nouveau?
00:05 gnarface: cyberpear: only maybe. https://nouveau.freedesktop.org/wiki/VideoAcceleration/
00:05 gnarface: (check here for info)
00:07 gnarface: oh actually, sorry, just looked and that card is too new, according to this
00:08 gnarface: theoretically it might help you if you have one of the cards they've managed to get it working for, but even then it's only for hardware video acceleration
00:08 gnarface: so it won't have an impact on opengl performance or other non-video desktop applications
00:08 cyberpear: thanks...
00:09 cyberpear: indeed, looks like it's only useful for older cards
00:09 gnarface: yea sorry it's just bad news. i'm not sure if nouveau works AT ALL with that card
00:09 gnarface: nvidia started requiring encrypted firmware at some point and that pretty much brought all progress on newer hardware to a dead halt
00:10 cyberpear: it works (sometimes) w/ the runpm_fixes patches applied
00:10 gnarface: interesting
00:10 Lyude: yes but don't worry, they're going to give us the firmware any day now
00:10 gnarface: lol
00:10 cyberpear: "any day now" (TM)
00:10 Lyude: ;)
01:00 imirkin: cyberpear: nouveau-firmware is likely a very ancient repo of nv4x firmware
01:01 imirkin: gnarface: no encrypted firmware anywhere nvidia
01:02 gnarface: imirkin: please don't invalidate my statement based on improper terminology without filling in the blank vacated by that nullification of my statement with the proper statement
01:03 gnarface: using proper terminology
01:03 gnarface: it's not just rude to me, it's unhelpful for the other readers
01:03 imirkin: it's similarly unhelpful to misinform
01:04 gnarface: it wasn't a deliberate attempt to misinform. you however are leaving the conversation in a much more misleading state
01:04 imirkin: i know i'm holding back the tide with a broom, but ... https://xkcd.com/386/
01:04 imirkin: i never suggested it was deliberate
01:05 gnarface: would it have been accurate if i had said encrypted vbios instead of firmware?
01:05 imirkin: no statement with the word "encrypted" would be accurate
01:05 imirkin: the vbios, firmware, etc just aren't encrypted
01:06 gnarface: interesting
01:06 imirkin: [as it relates to nvidia hardware, that is...]
01:06 imirkin: you might have been thinking of "signed", which is an entirely different and unrelated thing
01:06 gnarface: there we go
01:06 gnarface: except it's not
01:06 imirkin: except it is.
01:07 gnarface: nah you're splitting hairs now
01:07 gnarface: https://www.overclock.net/forum/69-nvidia/1633917-why-did-nvidia-lock-down-encrypt-vbios-pascal-cards-3.html
01:07 gnarface: and the whole internet talks about it the same way i did
01:07 imirkin: encrypted = unreadable
01:07 imirkin: see xkcd comic above.
01:07 gnarface: right, so cryptographically signed just not encrypted. fine. but don't shoot my statement down without making that clarification. that's bull shit.
01:08 imirkin: how is that clarification... the statement was entirely wrong
01:08 gnarface: that's not the first time you've behaved that way towards me, either.
01:08 imirkin: yeah, it's not the first time you misinform users coming in here
01:09 gnarface: right but the way you counterdicted my statement leaves them thinking there is no problem with the 900 and 1000 cards related to nvidia's actions with the firmware. do you actually not see how your statement is even MORE misleading?
01:09 gnarface: without the full correction, you are just confusing people more, not helping.
01:10 imirkin: all i know is that a bunch of users come in here, get wrong info, and leave
01:10 gnarface: if you're trying to convince me you've made an effort here to correct that situation, you've failed utterly.
01:10 imirkin: noted
08:33 alkisg: Hi guys. A program (the window manager marco) crashes xorg on nouveau, and running this works around the issue:
08:33 alkisg: xrandr -s other-resolution; xrandr -s original-resolution; marco
08:33 alkisg: Is it possible that the problem is caused by uninitialized variables in nouveau?
08:35 alkisg: I filed a bug report for marco there (https://github.com/mate-desktop/marco/issues/548) but since just toggling resolution fixes it, I'm thinking marco is just the symptom and the core of the issue might be in nouveau instead...
08:35 alkisg: This started happening when marco added Xpresent support
08:35 karolherbst: ahhh "Date: Sun, 31 Dec 2017 20:12:01 UTC"... downstream...
08:38 karolherbst: alkisg: I blame macro unless somebody convinces me otherwise
08:41 alkisg: karolherbst: but marco is restarted, it doesn't have saved state between the resolution changes...
08:41 alkisg: To rephrase: starting marco => crashes. Starting xorg => changing resolution => putting back resolution => starting marco => works
08:41 alkisg: Marco doesn't have saved state in this scenario, it's some state outside marco that works around the issue...
08:42 karolherbst: and how does the stacktrace look like?
08:42 karolherbst: but if only macro crashes, it's a bug inside macro
08:42 alkisg: Xorg crashes
08:42 alkisg: Not marco
08:42 karolherbst: okay
08:42 alkisg: Let me upload it...
08:42 alkisg: (the crash)
08:43 karolherbst: yeah, having the crash should be useful... just that Xorgs stacktraces are far from usefull in the general case :(
08:44 alkisg: This causes grief in all schools here that use nouveau (we're not using the nvidia driver), so I'm able to spend a lot of time in providing feedback, if anyone cares to help. Moment, reproducing + uploading crash...
08:45 karolherbst: hopefully it will be a trivial fix... but yeah.. without the crash it's only wild guessing and I don't even think I have super old cards...
08:45 karolherbst: 8800 is the oldest I've got here
08:45 alkisg: In recent cards, it doesn't crash, but it's causing artifacts like very badly drawn windows with wrong calculations, when the mouse is moved etc
08:45 karolherbst: or was it a 8600? doesn't matter
08:45 karolherbst: mhhh weird
08:45 karolherbst: alkisg: do you know what's the newest card being affected?
08:45 alkisg: pageflip off or xrandr change fixes all cases
08:45 karolherbst: I might be able to reproduce
08:46 alkisg: I got an "EN210" from a school that was affected, that's GT210,isn't it?
08:46 karolherbst: depends on what nouveau reports
08:46 karolherbst: but yeah, I have a G86 which is 1st gen tesla
08:46 karolherbst: and if tesla is affected I might be able to take a look
08:46 alkisg: I'll need to open a pc and plug it in; let me upload the crash first...
08:46 alkisg: Great, ty
08:47 alkisg: I can also give you remote, if it'll help any
08:48 alkisg: http://alkisg.mysch.gr/temp/_usr_lib_xorg_Xorg.0.crash
08:51 alkisg: TNT2 => always crashes, old cards => crash once per hour or something, new cards => artifacts; I haven't seen any *not* affected cards yet
08:52 alkisg:installs the gt210 card...
08:59 karolherbst: alkisg: btw, your IPv6 on the domain is broken
09:00 alkisg: Yeah sad story, the greek school network :/
09:03 karolherbst: uff
09:03 alkisg: Window manager warning: Log level 16: XPresent is not compatible with your current system configuration.
09:03 karolherbst: I think this crash report is useless :/ Do you think you can get me anything less troubling to deal with? like the Xorg log which should contain the stacktrace or something?
09:03 alkisg: 01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GT218 [GeForce 210] [10de:0a65] (rev a2)
09:04 alkisg: Sure, uploading xorg.log...
09:06 alkisg: karolherbst: https://termbin.com/z0uf
09:11 karolherbst: mhh.. at least it looks like an error in the nouveau ddx...
09:11 karolherbst: maybe I will be able to see the same issue here as well
13:08 codedmart: I am looking at a thinkpad p1. It comes with a nVidia Quadro T1000. Is that supported by nouveau?
13:09 karolherbst: codedmart: that's turing, isn't it?
13:09 karolherbst: codedmart: I don't know how that works with reverse prime and stuff, but we have kernel support for it, but no OpenGL yet
13:09 karolherbst: but
13:09 karolherbst: runpm should work
13:10 codedmart: Yes it is turing
13:11 codedmart: Hmm... I am not very knowledgable with graphics stuff, but with no OpenGL cause problems for me? I don't care about gaming or much stuff for grahics. I am mostly interested in the cpu/ram power I get with this laptop.
13:11 karolherbst: codedmart: your desktop is driven by your Intel GPU anyway
13:11 karolherbst: you usually need the nvidia GPU for external displays
13:12 codedmart: I probably won't even use the nVidia card unless required for external display.
13:12 karolherbst: yeah...
13:12 karolherbst: I would check if external displays just work
13:12 karolherbst: if so, just ignore it.. you should get proper battery lifetime though as the nvidia GPU should get turned off by nouveau
13:12 codedmart: They told me they do, but not sure how much I can trust a lenovo sales rep.
13:12 karolherbst: yeah.. dunno
13:13 karolherbst: codedmart: the nvidia GPU is at pci address 1:00.0, right?
13:13 codedmart: Not sure I don't have it yet. I would assume so.
13:13 karolherbst: ohh
13:13 karolherbst: yeah.. then it's all guessing at that point anyway
13:13 codedmart: If I do need the nvidia for external display is no opengl an issue?
13:14 karolherbst: no
13:14 karolherbst: but hardware acceleration might be
13:15 karolherbst: and for that we require firmware from nvidia we don't have yet
13:15 karolherbst: but it's in the work
13:15 karolherbst: might be ready next year or so....
13:15 karolherbst: dunno
13:15 karolherbst: maybe earlier
13:15 karolherbst: who knows
13:15 codedmart: OK fair enough for me. Thanks for all teh info :).
13:15 karolherbst: its not like nvidia was ever reliable if it comes to that
13:16 codedmart: I do know that dell has a precision 5540 that comes with the same card and ubuntu preinstalled. Not sure what drivers it is using though.
13:17 karolherbst: it's probably easier to get a laptop without a second GPU
13:18 codedmart: karolherbst: Recommendations? I really like thinkpads, but want as much power and ram I can get. Especially at least 32gb of ram or more.
13:18 karolherbst: no recommendations from my side
13:21 wrl: codedmart: there's the P1
13:21 wrl: oh you already brought that up, hah
13:22 codedmart: wrl: Yeah, but that has the second gpu.
13:24 karolherbst: alkisg: okay.. so I can't reproduce any weirdness with marco on my g86
13:25 karolherbst: *g84
13:53 alkisg: karolherbst: which marco did you try with? This started with marco 1.18.2, and was working with 1.18.1
13:53 karolherbst: whatever is newest in fedora30
13:53 alkisg: It should be more recent than that then.... Hrm...
13:54 karolherbst: 1.22
13:54 karolherbst: 1.22.2 actually
13:55 alkisg: karolherbst: I'll pinpoint the exact function that fails, and ask again here. Thank you very much for your effort. I can give you remote access to the unit that always crashes if it helps, but I guess it'll be easier after I pinpoint the exact function...
13:55 alkisg: I compiled marco without XPresent support, and it worked fine
13:55 alkisg: So it's something in its xpresent code that triggers the problem
13:56 alkisg: I'll also check with fedora30, maybe they default to composite=no, which doesn't have the issue
13:56 karolherbst: alkisg: maybe you can check if updated desktop fixes it?
13:56 karolherbst: I don't really want to track down already fixed bugs
13:56 alkisg: Sure, I can check with fedora30 in that unit, I'll do so tomorrow
13:57 alkisg: At least from the marco side, the code's still there
13:59 alkisg: The weirdest thing is that the person that added that code in marco, said he copied it from xfwm4, ...which works fine :/
14:03 karolherbst: yeah.. it sounds like some weirdo issue
15:02 imirkin: alkisg: if you can get symbols on that stack trace, i can definitely investigate
15:03 imirkin: whats "marco" btw?
15:03 imirkin: fwiw i'm not aware of any existing crashes in our EXA support. sounds like it happens across generations too =/
15:04 diogenes_: imirkin, wm same as compiz.
15:04 imirkin: (i *am* aware of some bugs)
15:04 imirkin: aha, so a compositor
15:04 diogenes_: yes
15:04 imirkin: is it likely to be easy for me to just run it without buying into the mate kool-aid?
15:04 imirkin: i.e. does it have dependencies outside of, say, gtk or whatever?
15:05 diogenes_: of course, should be as easy as marco --replace
15:05 imirkin: may i ask what "--replace" does?
15:06 diogenes_: it replaces your current compositor, for instance if i'm on xfce and im using xfwm4 and i want to use compiz instead, i do: compiz --replace
15:06 imirkin: what if i don't have a current compositor?
15:06 diogenes_: then theoretically marco will become the running one.
15:07 imirkin: (i'm still unclear on what a compositor does, except eat up resources and slow down drawing)
15:07 karolherbst: imirkin: eliminate tearing :p
15:07 diogenes_: it uses openGL backend to prevent tearing and nice things on desktop.
15:07 diogenes_: like shadows, transparency etc.
15:11 imirkin: in regular desktop usage, tearing is a non-issue
15:14 imirkin: anyways. i'll try to repro at some point. however getting a stack trace with symbols would expedite that considerably
15:14 imirkin: the nouveau ddx is fairly failure-tolerant, so i'm surprised. probably missed a spot.
15:15 imirkin: although that failure is on a NV5, which is probably not as well tested as the later variants.
15:15 imirkin: (but it sounded like it was a cross-gen issue)
15:16 imirkin: (so more likely some common helper is getting an unexpected thing)
15:16 imirkin: good to know someone actually uses nouveau with a NV5
15:20 karolherbst: imirkin: I already tried running it on a g84 and it worked fine, but you have much older hardware :)
15:21 imirkin: i also have a somewhat different userspace setup
15:21 imirkin: speaking of ... did you try on top of the nouveau ddx? :)
15:21 karolherbst: imirkin: also.. loading the nvidia firmware is kind of borked... could be that we ended up removing that support by accident or so.. or the instructions are out of date or whatever
15:21 imirkin: filenames are different now
15:22 imirkin: although i thought we re-added support for the older ones?
15:22 imirkin: i forget
15:23 alkisg: imirkin: if you happen to have an ancient tnt2, it 100% segfaults there, while in other cards it's more rare, so it takes more time to reproduce,
15:23 alkisg: or if remote access can help, I can give you ssh/vnc etc to one box with tnt2,
15:23 alkisg: otherwise, I'll try to learn how to add debug symbols etc and send them,sure
15:23 imirkin: alkisg: i do have a NV5
15:23 imirkin: the trick may be that i have a hard time getting marco up and running. we'll see.
15:24 imirkin: unfortunately i just use my main machine for development
15:24 alkisg: imirkin: I do not know the models, but in my tests, mx400 and 4000 only crashed once per hour or so
15:24 alkisg: Vanta and tnt2 crashed immediately
15:24 imirkin: NV5 = tnt2
15:24 alkisg: Great
15:24 imirkin: (the PCI kind, no AGP)
15:24 alkisg:tests with agp; not sure if it matters
15:25 imirkin: no AGP motherboard here
15:25 imirkin: (well, technically the PowerMac7,3 has one, but the power supply no longer works. plus BE + TNT2 = can't work)
15:25 alkisg: imirkin, I can also set up any distro/version you may want and give you remote there, if it helps
15:25 imirkin: alkisg: let's wait on that until i fail to repro
15:26 alkisg: Greaty, ty
15:27 imirkin: (note that i do this in my spare time, so don't expect miracles in terms of turnaround time)
15:28 karolherbst: imirkin: actually.. it works. noueau just throws some errors
15:28 karolherbst: and then falls back to the legacy stuff
15:28 imirkin: karolherbst: right, i think that's expected
15:29 karolherbst: yeah... just confusing :D
15:29 karolherbst: anyway.. that reclocking issue is a memory corruption issue
15:29 karolherbst: check the screenshot and dmesg extract
15:29 karolherbst: we get things like "illegal instr encoding"
15:29 karolherbst: and nvidias firmware doesn't help
15:29 imirkin: ok
15:29 karolherbst: it's a super weird issue
15:29 imirkin: was a quick easy thing to check :)
15:29 karolherbst: we even made the SEQ scripts doing the exact same thing
15:30 karolherbst: but 0x7 -> 0xf: broken, 0x7 -> 0xa -> 0xf: works
15:30 imirkin: oh
15:30 imirkin: that sounds familiar
15:31 imirkin: iirc skeggsb said that sometimes the blob did a 2-stage reclock
15:33 karolherbst: we do as well
15:33 karolherbst: for gddr5 you have to
15:33 karolherbst: _but_
15:33 karolherbst: I just remember something.. that's one of the GPU with two different memory widths
15:33 karolherbst: most of the memory is connected with 128 bit, some only with 64bit
15:33 karolherbst: maybe that's causing some weirdo side effects
15:34 imirkin: i thought that was a first on GTX 960
15:34 imirkin: (well, and the G80 has a weird memory layout...)
15:35 imirkin: or GTX 970 or whatever
15:35 karolherbst: wiki says "128+64" :)
15:35 karolherbst: for the 660
15:35 karolherbst: and I am sure it's one of them
15:38 karolherbst: mhh, the 970 was where it was extreme
15:38 karolherbst: 224 + 32
15:41 imirkin: hm, i'm probably thinking of something else -- how the parts are laid out? gtx 970 added something which required our memory support to be reworked
15:41 karolherbst: there were even fermis with that
15:41 karolherbst: "1024 MB RAM on 192-bit bus assemble with 4 × (128 MB) + 2 × (256 MB)."
15:42 karolherbst: imirkin: yeah... wondering if we need something similiar for those older chips as well...
15:42 karolherbst: dunno
15:42 karolherbst: or if that's transparent or something
15:45 imirkin: yeah, GF108 had something funny too
15:45 imirkin: we detected half the ram for a while
15:46 karolherbst: mhh, wiki states GF116 and GF114
15:46 karolherbst: but could be that there are more of those
15:47 imirkin: ok. well GF108 definitely had something funny in its memory
15:47 imirkin: perhaps a different thing than what the wiki is talking about
15:47 imirkin: some kind of part asymmetry
15:48 karolherbst: mhh.. well, the thing is, if you move up from 0xa to 0xf it works, so that's something which keeps me wondering... it's probably something in the reclocking code, but... if the SEQ scripts are identical to nvidia, maybe there is something weird we have to do
15:48 karolherbst: or maybe we have to do some of their funky waits
15:48 karolherbst: or something
16:08 adko: Hey all -- I'm trying to switch from using the nvidia proprietary driver to nouveau and i get a black screen (actually monitor times out / turns off) at boot unless I run `nomodeset` -- what might I try or read?
16:08 adko: arch linux btw.
16:10 phillipsjk: If your card is very new, it may not be supported.
16:11 adko: it's not that new.. but maybe? GTX 1060
16:11 adko: i see 2d is WIP on the feature matrix
16:12 imirkin: adko: how is the monitor connected?
16:12 imirkin: and what linux kernel are you running?
16:13 adko: 5.2.14-arch2-1-ARCH and DVI
16:13 imirkin: huh, surprising
16:15 adko: i vaguely remember nouveau being the default when I first installed this system.. and I had to jump through some hoops to get nvidia working.. so I probably jacked something up.
16:15 adko: (but now I wanna go back)
16:15 karolherbst: maybe something breaks on the modesetting thing
16:16 diogenes_: adko, then probably nouveau is blacklisted.
16:16 karolherbst: adko: are you able to boot with nouveau blacklisted?
16:16 karolherbst: ssh into it and then get the dmesg log after doing "modprobe nouveau"?
16:16 karolherbst: diogenes_: that makes no sense on a desktop system
16:16 imirkin: diogenes_: if it were, screen wouldn't go black
16:16 karolherbst: adko: or is that a laptop and it's about external displays?
16:17 adko: diogenes_: it seems like nouveua is only blacklisted when I have the nvidia driver installed.
16:17 diogenes_: ok i see.
16:17 adko: karolherbst: it's a desktop
16:17 karolherbst: adko: okay... then blacklist nouveau so it doesn't get loaded
16:17 karolherbst: then you have efifb and get your display
16:17 karolherbst: ssh into it, modprobe nouveau
16:17 karolherbst: and check dmesg
16:17 karolherbst: that might tell you what's up
16:17 adko: ah yeah hadn't thought about ssh'ing on and checking dmesg when it's black
16:18 karolherbst: well... sometimes the system hangs so you can't connect anymore
16:18 karolherbst: but if it works without blacklisting, that's fine
16:18 adko: karolherbst: I can tell the system is not hung.
16:18 karolherbst: ahh, okay
16:18 karolherbst: yeah, thenn ssh :)
16:18 karolherbst: anyway, gtg
16:18 adko: ok be back in a bit. thanks
16:20 nemo: Hey folks... I just installled a debian stable (actually devuan stable but whatever) - I'm sure that sends chills through the hearts of the nvidia and intel driver teams...
16:21 nemo: I was mostly wondering if there's a handy database you guys might have of device strings so I could see if this card of mine *would* be supported if I was on the latest bleeding edge
16:24 nemo: I guess I should start by JFDDGing nouveau device database or something ☺
16:24 phillipsjk: Nemo not quite by device string: https://nouveau.freedesktop.org/wiki/FeatureMatrix/
16:25 nemo: ow
16:25 imirkin: nemo: what GPU do you have?
16:25 imirkin: and what kernel
16:25 nemo: I guess the problem is lspci isn't even giving me a device name right now - which probably implies old kernel? "NVIDIA Corporation Device 1f02 (rev a1)
16:25 nemo: "
16:25 imirkin: that's based on your pci.ids
16:26 imirkin: 1f02 TU106 [GeForce RTX 2070]
16:26 imirkin: heh
16:26 nemo: 4.9.0 kernel
16:26 imirkin: good luck =]
16:26 imirkin: you have an expensive space heater.
16:26 nemo: hm. 4.9.88+deb9u1 (2018-05-07)
16:26 nemo: whoo boy
16:26 nemo: imirkin: well. surely the kernel-tainting-binary-blob would work right
16:26 nemo: or I guess I could switch to unstable *sigh*
16:27 nemo: problem with debian is if you're on unstable suddenly all problems are your fault
16:27 adko: not sure what I should be looking for in dmesg
16:27 adko: but I don't see anything jumping out at me...
16:28 imirkin: nemo: yeah, definitely blob should be fine
16:28 imirkin: nemo: turing modesetting is supported in the bleeding edge
16:28 nemo: turing?
16:28 imirkin: but accel isn't, and won't be in the foreseeable future
16:28 imirkin: nemo: turing is the name of the generation of GPU you have
16:28 nemo: ah...
16:28 nemo: ok. well. good to know before I went down path of upgrading everything
16:28 nemo: imirkin: this machine is rather linux hostile
16:29 nemo: imirkin: but, eh, take what company is willing to offer me
16:29 adko: (hopefully relevant) dmesg: https://pastebin.com/eT25AePZ
16:29 nemo: turn off safe boot, turn off the raid scsi which isn't even useful with their config, unbreak the default windows when doing that in case I need it for returns/dell support/our support …
16:40 HdkR: Interesting, a dell that is linux hostile? Usually they are pretty good aside from the Nvidia GPU stuck in it
16:41 nemo: HdkR: https://www.dell.com/community/XPS/HELP-Install-Ubuntu-Dual-Boot-switch-RAID-On-to-AHCI/td-p/5752914
16:42 nemo: HdkR: that's the one I ran into, aside from the usual trusted computing issues
16:42 imirkin: it's super-dumb, since it's set that way to force windows to pick the intel driver rather than the windows generic driver
16:42 nemo: HdkR: the drives were completely invisible, although why, no idea.
16:43 imirkin: coz they pci id was set to something non-standard rather than the AHCI one
16:43 HdkR: Oh yea, swapping that mode is always fun. Windows gets pissy while Linux doesn't usually care :P
17:17 fseit: Hi, anyone there know if with nouveau the gt 1030 will be able to run fullhd resolution and if will be able to play fullhd video?
17:18 imirkin: if your cpu can decode it, sure
17:23 fseit: It's an opteron 4284 it should be able, right now i'm using a gt710 it works but it struggle and inside a VM the video and audio go out of sync, i've tryed with a more powerfull amd card with prorpietary driver and the videos inside the VM runs pretty well, but since i want to use free software only with no closed driver and firmware i want to use a more powerfull nvidia, and the gt 1030 seems fine, but i just don't know if it work. With the amd card
17:23 fseit: without closed firmware i could not even start X
17:24 fseit: That's why my question, i have to buy a the gt 1030 and i would be sure that this card work at full hd without proprietary stuff
17:26 HdkR: gt1030 requires proprietary firmware though?
17:28 HdkR: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/tree/nvidia/gp108/gr
17:28 HdkR: More specifically https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/commit/nvidia/gp108/gr?id=2567e092339cd3403d697dc2e0967c31b7acb989
17:33 fseit: So if i will buy it i will be in the same situation as the amd?
17:34 HdkR: Even worse since you don't get power management
17:35 alkisg: imirkin, karolherbst: with my limited knowledge, I put "g_warning()s" in the marco code, and I traced it down to an XPresentPixmap() call:
17:35 alkisg: https://github.com/mate-desktop/marco/blob/master/src/compositor/compositor-xrender.c#L1294
17:35 alkisg: Xorg crashes at that point. Any hints on how to proceed?
17:35 alkisg: Note that if I just switch resolution to 800x600 and then back to 1024x768, everything works; would it make sense to printf the values there and compare?
17:45 imirkin: alkisg: yeah, it sounds like something that tickles an internal bug or failure of some sort
17:46 imirkin: really need to see where the failure is happening
17:46 imirkin: stuff like X calls is pretty far removed from the driver
17:47 imirkin: otoh ... XPresentPixmap ... hm. that sounds like a flip.
17:47 alkisg: imirkin: I can put some code in marco and disable its use of Xpresent; but since it affects newer cards too (although without crashing), would it be useful to hunt this down more?
17:47 imirkin: alkisg: do you see an error in dmesg of any sort?
17:47 imirkin: whatever the bug is, we should def fix it
17:47 alkisg: [ 1124.177880] traps: marco[5120] trap int3 ip:b70c39d0 sp:bfb90950 error:0 in libglib-2.0.so.0.5600.4[b7075000+12e000]
17:47 alkisg: [ 2660.863157] perf: interrupt took too long (2536 > 2500), lowering kernel.perf_event_max_sample_rate to 78750
17:47 imirkin: however looking at the application isn't going to make life easy
17:48 imirkin: gotta look at the ddx, since that's what's dying
17:48 alkisg: Too deep waters for me :/
17:48 alkisg: Those are all I got in dmesg, and that marco is old one, maybe from when I was trying to compile/run too new versions with wrong dependencies
17:56 fseit: https://www.phoronix.com/scan.php?page=news_item&px=NVIDIA-GP108-Firmware
17:56 fseit: reading there seems i cant use 3d accelleration without firmware but it should work or do i understood badly?
17:57 imirkin: fseit: 3d accel sure. you were talking about about video playback
17:58 imirkin: fseit: also there's no reclocking, so gpu is stuck at lowest memory / core speeds
18:00 fseit: yes is it right, but i whould like to know at least if without closed firmware the system will boot inside the DE, and i hope for video the accelleration will be done by the cpu, as i understood about HdkR wrote is the system do not even work without these blobs
18:01 imirkin: modesetting should work without any blobs
18:01 HdkR: It'll bring up a display, which is what your GT710 is already doing
18:01 HdkR: So something "more powerful" doesn't make any sense here for just display output
18:01 HdkR: You could use those USB 2.0 to DP adapters even
18:02 fseit: do you think will be able to bring my display at 1080p?
18:02 imirkin: if the GT710 is a GK208, that should defitely be better than a GT1030
18:02 fseit: why?
18:02 imirkin: coz kepler support is way better
18:03 imirkin: you can get video decoding accel there too, there are instructions for retrieving firmware for that
18:03 fseit: that's a bad news i was hoping to have someting better
18:04 imirkin: some GT710's are GF108's, which are probably not as good
18:04 fseit: anyway, thanks for the reply
18:04 fseit: mine is gk208
18:04 imirkin: (this is why we tend not to speak in terms of marketing names)
18:04 imirkin: ok
18:04 imirkin: what do you wish worked better?
18:11 fseit: video and audio play on movies, now it struggles a bit, but i'm not enough tech to do something more than buy a card and i don't want to use closed firmware so my hopes where about more powerfull and newer card
18:15 HdkR: Get an older Kepler based card or a more powerful CPU to overcome software decoding limits?
18:15 HdkR: Depends on if the older Kepler supports the video formats you desire
18:22 phillipsjk: When I bought a Geforce 6200 based on advice here, it was so old it came with a coupon for a better (or at least newer) card. (Was happy with my purchase because nouveau moslty worked on it -- think I found a few bugs)
18:25 karolherbst: I hope that advice was like 10 years ago :p
18:30 phillipsjk: Closer to 5, but PCI bus support was also a requirement.
18:31 phillipsjk: Well, that or half-height AGP, but those were not really avialable other than the card I already had.
18:33 phillipsjk: Had to replace the machine when the caps finally stopped working.
19:49 imirkin: ah yeah, the PCI bus requirement kinda limits you in what you can get
19:49 imirkin: allegedly there were some tesla-era GPUs with PCIe -> PCI bolted onto them, but i've never seen one in the wild
20:34 gruetzkopf: is external PCI-PCIe an option? (plus lowprofile card)
20:35 imirkin: to plug PCI into a PCIe slot? probably. i don't think there was much in terms of external PCI slots though.
20:35 imirkin: unless you count PCMCIA