14:27agneli: guys I have this issue: https://paste.debian.net/1192782/
14:27agneli: it happens while running 0ad
14:28agneli: results are such that game works in menus perfectly
14:28agneli: but when I launch it it comes to a halt
14:29agneli: for let say 15-20 minutes then it is perfectly palayable for 2-3 minutes
14:29agneli: then the story repeats...
14:29agneli: I have no idea what to do...
14:29agneli: same game on the very same machine works perfectly...
14:30karolherbst: mhh, sounds like a mesa issue
14:30karolherbst: what GPU is that with?
14:30agneli: quadro 160M
14:30agneli: 01:00.0 VGA compatible controller: NVIDIA Corporation G98M [Quadro NVS 160M] (rev a1)
14:31karolherbst: maybe imirkin knows what's up?
14:31imirkin: probably runs out of vram
14:31imirkin: submits fail
14:31imirkin: or something
14:32agneli: that is the game: https://play0ad.com
14:32agneli: i'd say pretty demanding for such an old (~10yrs) hw
14:32agneli: 256MB VRAM
14:33karolherbst: agneli: is there some benchmark mode or something one could do to trigger it?
14:33karolherbst: not really interesting in playing 0ad :D
14:33karolherbst: but I do have a similiar GPU
14:33agneli: I used to run on uningine on this machine and it was all ok
14:33agneli: but that was long ago
14:34agneli: surely much older versions of the software
14:34agneli: shall I download and run it again?
14:34agneli: or you mean the benchmark mode in the game?
14:35agneli: that game is already a benchmark - max 7fps this machine is able to do
14:37imirkin: agneli: you may be able to reclock
14:37imirkin: if you want to crash faster ;)
14:38agneli: karolherbst: I just got that - I will see if I can reproduce that wiht some other software
14:38imirkin: agneli: look at /sys/kernel/debug/dri/0/pstate
14:38agneli: imirkin: it does not crash
14:38agneli: it just stops... and then works again
14:38agneli: and then stops
14:39imirkin: yeah, that's semi-expected then - probably just tries to use a LOT of vram
14:39imirkin: given the "ttm buffer eviction failed" messages
14:39agneli: but as said on windose on very same machine works perfectly
14:39imirkin: or with linux and nvidia blob drivers, presumably
14:39imirkin: i'm not saying that a better driver can't manage this better
14:40imirkin: i'm just saying this makes sense in the context of nouveau :)
14:40agneli: that is the option I do not want to use... :( but yes I can check that...
14:40imirkin: nah, you don't need to do that
14:40imirkin: i'm just saying ... nvidia driver manages memory better
14:40RSpliet: Any improvement if you reduce texture quality?
14:40agneli: but that would mean no 0ad for me?
14:40imirkin: we have very harsh policies about what's in vram and what's not
14:40agneli: RSpliet: i reduced everythign possible already
14:41imirkin: it'd work better if we were more flexible
14:41imirkin: e.g. allowing some stuff to be in gart if it doesn't fit
14:41imirkin: allowing the kernel driver to place stuff however it feels like
14:41imirkin: but that's currently not super-supported
14:41imirkin: (iirc it bails of both the VRAM + GART flags are set on a placement)
14:43agneli: reducing resoolution will not help I guess?
14:43imirkin: it may
14:43imirkin: the fbo's will be smaller
14:44RSpliet: from my very limited understanding, the majority of VRAM tends to be occupied by textures. Although this may not be the case for a 2.5D game like this
14:45imirkin: RSpliet: fbo's also take up room though :)
14:45imirkin: esp with intermediate renders, etc
14:48agneli: Guys, it seems there is a bug with 0ad related to my distro, I think I need to install some other linux and check
14:49agneli: I am on gentoo
14:49agneli: I will report back once I have more data
14:50agneli: I will get another distro, I will compile from original source, try to reproduce with soem other software, etc
14:50imirkin: what could distro *possibly* have to do with this?
14:51agneli: compilation option?
14:53imirkin: lol ok
14:53imirkin: (no, that's not it)
14:53imirkin: but do whatever you like, obviously
16:33agneli: imirkin: and what shall I do in your view, please?
16:56imirkin: i don't think there's a solution here
17:02agneli: older version ;)
17:03agneli: however they claim 256MB should be OK
17:03HdkR: Wonder how often they actually test that claim
17:04imirkin: agneli: yeah, with a better driver
17:04imirkin: i mean, you could also improve nouveau
17:36agneli: imirkin: sure, can then cry together :)