00:00karolherbst: ahh, I see
00:00imirkin: if compression isn't supported, it should just disable the compression bit and move on with life
00:00imirkin: instead of returning a EINVAL
00:02selk: imirkin: i would appreciate if you can suggested me a card :)
00:02selk: i already have amd hardware
00:03selk: and the gt 730 is giving me troubles
00:04imirkin: not gonna do better than gt730
00:08HdkR: ick, GT730 has had fermi and kepler versions
00:09HdkR: and the fermi version suuuucks
00:09selk: I try to boot it and suddenly a green background appears, freezing...
00:10imirkin: that's ... less-than-ideal
00:10imirkin: nouveau's bad, but usually not THAT bad
00:11selk: i have other card a gtx 970 which works with nouveau
00:11selk: the problem with the gt 730 from evga, is on the initial boot
00:16gnarface: grr evga
00:16gnarface: you mean it's green at the bios POST screen?
00:17gnarface: i had TWO evga cards start overheating on me after about 6 months to a year of age
00:18selk: I try to boot the kernel image, when it is loading, it freezes, white text under green background
00:18gnarface: after the first one fried itself, i learned my lesson, and removed the heatsync from the card, cleaned off the black sticky tar crap they used instead of thermal paste, then re-applied arctic silver (V or VII, i don't know)
00:18selk: (very early)
00:19gnarface: the second card immediately went back to sane temperature ranges, seemingly before any permanent damage
00:19selk: oh, interesting
00:19gnarface: actually it was several degrees colder than stock, as i recall
00:19gnarface: (the first card was not salvageable. the ram got fried i think)
00:21gnarface: if you never see it happen before the boot loader, then it might not be a hardware issue though
00:21gnarface: that's a good way to tell; if even the bios screen is corrupted, that's got nothing to do with nouveau
00:22gnarface: if it only ever goes wonky after the kernel loads though, you know it's probably the driver
00:22selk: yep, i was thinking in obtain another card, and make sure that it works with nouveau
00:22selk: also, i am exploring radeon as an option
00:22selk: the kernel that i am using is 4.18.9
00:22imirkin: selk: kepler generation is probably best-supported
00:26selk: imirkin: a 740 maybe?
00:27imirkin: any 7xx should be fine
00:27imirkin: as long as it boots :)
00:27gnarface: didn't someone come in here the other day with something weird like a 710 or something like that that was not fine?
00:28gnarface: or did that get resolved? i forget
00:29selk: 710 is available as gddr5
00:30selk: i should try to bought from msi, asus, or ... ?
00:31imirkin: every time i've touched msi, it was total crap
00:33selk: it looks good: https://h-node.org/videocards/view/en/1972/NVIDIA-Corporation-GK208--GeForce-GT-710B---rev-a1-/1/1/NVIDIA-Corporation/undef/undef/undef/video-card-works/gt%20710
00:33selk: I use linux-libre...
00:33karolherbst: MSI looks like "cheap premium", seriously. Who actually thought that their brand is fine? :D
00:34gnarface: they have good initial quality to value
00:34gnarface: i go for asus because of durability
00:34karolherbst: doesn't matter if anytime I see MSI I think "uhhhmmm, how about no"
00:34selk: hehe, thank you guys
00:34gnarface: i don't know why i've heard the opposite about their stuff complained about alot around here.... ASUS is good, their stuff has been lucky for me i guess or something
00:35karolherbst: I bought an ASUS mb for my work machine
00:35gnarface: as far as WORST longevity of all time, i'd say EVGA but only because BFG-Tech is out of business now
00:35karolherbst: but this mb was black.. so maybe I choose because of that (and it gave me what I wanted)
00:35karolherbst: red mbs are the worst
00:35gnarface: only on the sample data i have though, EVGA cards might still be a good buy if you're happy to disassemble the fan housing and re-apply the thermal paste properly
00:35karolherbst: and green ones look old
00:36gnarface: there's always this margin of vendor discretion when it comes to factory overclocking and EVGA seems to be the ones still pushing that limit the hardest
00:37karolherbst: doesn't change much anyhow
00:37karolherbst: so you get like 5% higher clocks, wow
00:37gnarface: yea, and it costs longevity and it costs heat and it costs power efficiency AND they're skimping on the thermal paste on top of all that
00:38gnarface: so basically you buy EVGA if you want to win benchmarking contests more than you care about having to replace the card in a week
00:38gnarface: i'd go for somethig from ASUS with 3 fans on it any day
00:39karolherbst: I don't want to buy stupid looking stuff
00:39karolherbst: and most "pseudo high-end" stuff looks just stupid
00:39karolherbst: which more or less eliminates all overpriced products :D
00:40karolherbst: in the end it doesn't really matter anyway
00:43gnarface: eh, i don't mind a whole lot if they make it look stupid but internal lighting is pushing it
03:40imirkin: skeggsb: would appreciate a look at https://bugs.freedesktop.org/show_bug.cgi?id=108500 -- i think something went wrong in one of the VMM rewrites
04:29imirkin: skeggsb: either we need to flip things over to use mcp77_mmu (which has a vmm without compression) or we need to not mark the memory as stolen or we need to allow compression on stolen pages...
14:15pendingchaos: imirkin: ping on https://patchwork.freedesktop.org/patch/254788/ ?
14:16imirkin: pendingchaos: r-b
14:16pendingchaos: I'll be pushing it then
19:19selk: is the "quadro p1000" supported by nouveau?
19:27joepublic: I just came to say thank you for all your hard work and for what you are doing. Trying to build a libre system and just ordered an NvGTX660 to replace a radeon.
19:31selk: joepublic: are you using linux-libre?
19:34joepublic: selk - no, not with a radeon I'm not.
19:34joepublic: trying to get there.
19:35joepublic: I have two machines here with radeons running debian, and my goal is to replace the radeons and get them running trisquel with linux-libre.
19:36joepublic: I can use the Radeons in other machines that don't need accelleration by using them with the vesa driver (I think)
19:36karolherbst: selk: it might or might not work, but Pascal should be more or less supported
19:38joepublic: Speaking of Pascal I am thinking about getting a Jetson TX2, but I suspect that 3d performance will be that great.
19:41karolherbst: yeah, without the PMU firmware it isn't that great
19:50selk: karolherbst: i was checking the quadros for durability terms, but now i am in the doubt ...
19:51karolherbst: do you need those for business?
19:51karolherbst: if not, quadros are overpriced
19:51selk: yes, expensive
19:52selk: p600 is more accessible than the p1000, but i don't know if will work
19:52selk: both are based on pascal
19:54karolherbst: well, you won't get much performance out of those
20:17selk: probably I should go for the gt 710, gddr5, pci-express 2.0
21:05Centurion_Dan: I have a GF106 (0c3880a1) on my laptop and I'm seeing complaints about missing firmware in the kernel log..
21:05Centurion_Dan: should I be extracting it?
21:28Centurion_Dan: the firmware that is, or is there a new way of obtaining firmware??
21:55coderobe: Hi, I'm getting lockups in nouveau on a GTX 1060 (6G vram model) https://shr.codero.be/MellifluousMagentaSwallow2.txt
21:55coderobe: This happens as soon as what seems like the first gpu access, starting with a 'nouveau 0000:01:00.0: Refused to change power state, currently in D3'
21:55coderobe: That's on 4.18.9
21:56coderobe: Any ideas?
22:01joepublic: Although I am not knowledgeable enough to help you, I am curious whi the manufacturer of the card is.
22:02coderobe: joepublic: it's builtin on a laptop
22:03joepublic: So not a card. Got it. I am slow sometimes and apologize for my unhelpful contributions. :)
22:24selk: coderobe: D3?
23:09Centurion_Dan: coderobe: what model laptop, and whats the output of lspci -vvv for that card
23:10coderobe: Centurion_Dan: lspci already triggers the hang and completely locks up
23:10Centurion_Dan: (not that I can help much, but it's basic information that helps to identify which gpu is at issue).
23:12Centurion_Dan: check dmesg for identifying information from early boot where nouveau is loaded
23:13coderobe: NVIDIA GP106 (136000a1)
23:13Centurion_Dan: what laptop make and model
23:13coderobe: it's a completely unbranded no-name laptop
23:13Centurion_Dan: do older kernels work ok
23:14coderobe: well, i suppose i have only tested 4.x kernels
23:15karolherbst: coderobe: yeah, runpm is broken
23:15karolherbst: coderobe: nouveau.runpm=0
23:15karolherbst: or don't use nouveau and enable runpm through sysfs
23:15karolherbst: it's some signed firmware/script magic and we don't really know what's wrong
23:16karolherbst: or well, no way to fix that correctly as nvidia might has to help out with that
23:16coderobe: hm, i see. runpm=0 effectively keeps the gpu online though, right?
23:25karolherbst: coderobe: yes
23:25coderobe: ah, thanks for the info!