13:21 loa: hello! what is last nvidia video card which will have dynamic frequency under nouveau?
13:26 pmoreau: In a perfect world, or in the current world?
13:28 mlankhorst: skeggsb: why u spkng hngarian? mkng t hrd t fllw wt hpns n nv50_disp
14:22 soreau: In a perfect world, nvidia doesn't exist
14:22 soreau:hides
14:36 loa: pmoreau, wanna buy used gpu and be able to use it with nouveau for games.
14:37 loa: so looks like it is real world.
14:38 pmoreau: Well, currently you can’t reclock Maxwell v2 and newer, because we can’t control the fan without NVIDIA’s binaries. They were supposed to release those, but haven’t yet.
14:39 pmoreau: Fermi reclocking is being worked on, but it is not usable yet.
14:39 pmoreau: Most Tesla and Kepler should work fine.
14:40 pmoreau: An AMD GPU might be a better idea if you want to use your GPU with open-source drivers.
14:52 imirkin: loa: a GTX 780 Ti is the most powerful GPU that nouveau can reclock at the moment. the GTX 750 Ti came out later, but is less powerful. nouveau can still reclock it.
14:52 imirkin: loa: none of those have dynamic frequency changing logic at the moment, but may eventually
14:52 imirkin: loa: as pmoreau said, if you're looking for OSS support, stick to AMD (or Intel)
14:55 loa: ok, thx, got this.
14:58 mooch: i just got a gtx 750 ti lol
14:59 mooch: can't use it until my broken desk gets replaced tho
15:17 karolherbst: mooch: you will need 4.10 anyway, so
15:22 mooch: why?
15:23 mooch: i was just gonna use the blob cus VULKAN
15:27 mooch: that computer also has windows on it so yeah
15:46 imirkin_: mooch: that'd be the case if you were looking to use nouveau
15:46 mooch: ah, okay
15:46 mooch: what's new in 4.10 for nouveau?
15:47 imirkin_: karol's voltage change & co stuff landed
15:47 imirkin_: including, but not limited to, hooking up reclocking for gm107
15:47 mooch: oh neat!
15:47 mooch: yeah, that'd be useful
15:48 NanoSector: anything I can test for gk107 (GT750M)? I'm the guy whose card would reclock the core but hang when memory reclocked
15:49 imirkin_: i guess karol is familiar with your situation?
15:49 NanoSector: yeah i worked with him to provide logs and stuff
15:51 NanoSector: i'm not sure if anything has changed for kepler in the meantime
16:02 imirkin_: doubtful
16:09 karolherbst: NanoSector: I guess I already asked for a mmiotrace five times?
16:09 NanoSector: yeah :P
16:09 karolherbst: k
16:10 karolherbst: maybe I just didn't take a look into it yet
16:10 karolherbst: could be something silly
16:10 NanoSector: I think we poked around and found out that reclocking the core worked but the memory locked it up
16:10 NanoSector: as it worked when disabling memory reclcoking
16:10 NanoSector: *reclocking
16:11 mooch: holy fuck is mwk gonna make nv04_graph work up to g80???
16:11 mooch: nv4:g80 dam
16:15 pmoreau: I think there is no stopping him now: he's on his way! :-D
16:44 mwk: mooch: the general idea is to have a unified testsuite for NV1-NV4x
16:45 mwk: right now I have simple mthd tests working on NV4/NV5 and most of scan+state tests working on NV1x
16:46 mwk: and I should have the first NV10 methods tests running in a moment
16:46 mooch: ah
16:46 mooch: mwk: maybe you should start working on pfifo tests now
16:47 mwk: I'd love to throw in NV20 into the mix as well, but I'm out of machines with AGP slots
16:47 mwk: also, NV1/NV3 tests should be eventually merged into the unified testsuite as well, but that's not exactly high-priority
16:48 mwk: pfifo, eh
16:48 mwk: sounds good
20:07 NanoSector: Lekensteyn, I'll run some tests with pcie_port_pm=off and 4.7 once I get out of the shower in a bit
20:08 NanoSector: sorry for not looking at it earlier, didn't have the time or motivation :x
20:08 Lekensteyn: NanoSector: np, but note that pcie_port_pm=off does not have any effect on 4.7 (only 4.8)
20:09 NanoSector: yeah i mean i'll test with it on 4.8 and then on 4.7 without it
20:09 NanoSector: :)
20:10 NanoSector: also yay on PRIME vertical sync being added to Xorg 1.19, even though I use wayland mostly
20:21 karolherbst: NanoSector: are you able to recompile nouveau fast?
20:21 NanoSector: compiling a kernel takes me about 2 hours, compiling a standalone nouveau.ko takes me 2 minutes
20:21 NanoSector: or less
20:21 NanoSector:still hasn't showered
20:21 karolherbst: well there is something you could try
20:22 NanoSector: fire at me :)
20:22 karolherbst: have to think first
20:23 NanoSector: hehe okay then I'll go shower in the meantime
20:29 Lekensteyn: karolherbst: have something to test?
20:42 karolherbst: NanoSector: actually a mmiotrace of nouveau would be good
20:59 NanoSector: karolherbst, https://drive.google.com/open?id=0B2F3P3d66S-1aWt1S0hDc0xkbFU that's from the first of august
20:59 NanoSector: but i don't remember if that was nvidia or nouveau
21:00 karolherbst: should be nvidia
21:01 karolherbst: it is the one I already have
21:01 NanoSector: yeah okay
21:05 NanoSector: woo, laptop hung completely when reclocking
21:05 NanoSector: :)
21:06 NanoSector: ah journalctl picked up on an error
21:07 NanoSector: karolherbst, does this help? http://hastebin.com/yatukoxebe.xml
21:07 karolherbst: NanoSector: uhh right :/
21:08 NanoSector: karolherbst, http://hastebin.com/ijonaqahav.go that's all the mmiotrace that was captured
21:08 karolherbst: yeah, that won't help
21:08 karolherbst: there is a bug I can't track down for whatever reason
21:09 NanoSector: which is?
21:09 karolherbst: a bug within the trace
21:09 NanoSector: oh, so stuff I don't understand :P
21:09 karolherbst: you could revert cfa52c0cfa4d727aa3e457bf29aeff296c528a08 in the kernel for now
21:10 NanoSector: well nouveau tends to freeze my entire system anyway at times
21:10 NanoSector: when reclocking that is
21:11 NanoSector: it didn't freeze the second i enabled mmiotracing, i enabled 0a pstate and a second later it hung
21:12 karolherbst: okay
21:12 karolherbst: should be something usefull in the trace then
21:12 NanoSector: yeah the file i sent you was all that was captured
21:13 karolherbst: mhh, would be fun if it is a silly race condition
21:13 karolherbst: uhhh
21:13 karolherbst: you have to enable the tracer before loading nouveau
21:13 NanoSector: oh
21:13 NanoSector: lol
21:13 NanoSector: so i have to blacklist nouveau, reboot, enable the tracer, load nouveau, then reclock?
21:13 karolherbst: yeah
21:13 karolherbst: mhhh
21:14 karolherbst: actually why the hell not
21:14 karolherbst: it could be a race condition all along
21:14 NanoSector: is that good or bad news? :x
21:15 karolherbst: don't know
21:15 karolherbst: have to waste another weekend on this
21:16 NanoSector: oh you don't have to look into it for me :P like i said before i'm in no hurry or anything
21:17 NanoSector: just had some spare time, figured something new might've come up
21:23 NanoSector: Lekensteyn, pcie_port_pm=off does not make it work
21:23 NanoSector: it still hangs on resume
22:11 Lekensteyn: NanoSector: huh, and 4.8.6 works fine?
22:12 NanoSector: Lekensteyn, no, it does not
22:15 Lekensteyn: NanoSector: here you reported success with an older version, that one worked fine for you? https://bugs.freedesktop.org/show_bug.cgi?id=98398#c5
22:16 NanoSector: Lekensteyn, oh I mean the suspend/resume issue, pcie_port_pm=off does make power management work
22:16 NanoSector: but the card not turning off issue was fixed with your patch
22:17 Lekensteyn: oh, so it is a different issue that already existed before?
22:17 NanoSector: yesh
22:22 Lekensteyn: it is complicated..
22:26 Lekensteyn: I think that ACPICA bug 1333 is the root cause for your problem, acpi_osi="!Windows 2013" should be a workaround until that one is fixed
22:26 Lekensteyn: NanoSector: have you tested that one already?
22:27 NanoSector: nope, will add that, thanks
23:36 NanoSector: Lekensteyn, yep, adding that acpi_osi thing to my kernel command line works wonders