21:30gst568923: hi guys
21:32gst568923: I have a GF119 [GeForce GT 705]
21:32gst568923: and I would enable the VDPAU video acceleration
21:33gst568923: I have run "dmesg | grep -i chipset" but the output is blank
21:36gnarface: gst568923: try `vdpauinfo`
21:37gnarface: should work, i think
21:42gst568923: How can I check the correct version for the video engine?
21:43gst568923: I found this https://nouveau.freedesktop.org/wiki/VideoDecoding/
21:43gst568923: my video card is NV..?
21:44gnarface: oh, you can get that from lspci probably
21:45gnarface: try: lspci|grep nvi -i
21:45gnarface: or maybe this would be better: lspci|grep 'vga\|video' -i
21:46gst568923: return me
21:46gst568923: return me
21:46gst568923: 01:00.0 VGA compatible controller: NVIDIA Corporation GF119 [GeForce GT 705] (rev a1)01:00.1 Audio device: NVIDIA Corporation GF119 HDMI Audio Controller (rev a1)
21:46gnarface: i believe GF119 is the same thing as NV119
21:47gnarface: i don't understand the reason for changing it to GF from NV
21:47gnarface: so you may actually be VP6
21:47gnarface: that might explain why my 650 is showing more vdpau features than you
21:48gnarface: (but i'm also using the proprietary driver on it)
21:49RSpliet: gnarface: GF -> "G" (Graphics? I think...) Fermi. GK->Kepler, GM -> Maxwell
21:50gnarface: ah, interesting.
21:51gst568923: https://nouveau.freedesktop.org/wiki/CodeNames/#NVE0 in my case is NVE7 or NV108?
21:53gnarface: good question
21:53gnarface: i don't know
21:54gnarface: gst568923: maybe it is actually NVC0?
21:54gnarface: if GF means fermi...
21:55gnarface: that would mean GF119 has to be NVD9 on this chart
21:55gnarface: NVD9 (GF119) GeForce 410M, 510 (?), GT (520, 520M, 520MX, 610), 610M
21:55gnarface: Quadro NVS 4200M
21:55gnarface: i guess Geforce GT 705 must be covered by the (?)
21:56gnarface: this table is incomplete
21:57gnarface: gst568923: that's in a laptop, right?
21:57gst568923: no desktop
21:58gst568923: in this table the series 700 is indicated as VP5: NVD0-NV108 (GeForce 600, 700 series, a few low 500s; corresponds to VDPAU feature set D) https://nouveau.freedesktop.org/wiki/VideoAcceleration/
21:58gnarface: yea, the more i look at this table, the more sure i am that yours is falling under NVC0 family (Fermi) - NVD9 (GF119)
21:58gnarface: something weird about the number 705 means it was rare
21:59gnarface: i inferred from that it must be some mobile OEM device, but i guess they put these in desktops too
21:59gnarface: when they use a weird number sometimes they're doing that to relabel a stripped-down version of a card from another series
22:00gst568923: is the Dell desktop
22:01gst568923: in this link https://www.techpowerup.com/gpu-specs/geforce-gt-705-oem.c2578 architecture is indicated "Fermi"
22:01gnarface: yea, they're known for these type of shenanigans.
22:02gnarface: so it's a Fermi chip that they relabeled in the 700 range but barely, so they could trick people into thinking they're getting a Kepler era gpu while retaining plausible deniability in case it were to ever come up in cout
22:03gnarface: you're right this raises some questions about the completeness of these charts...
22:03gst568923: but VDPAU is indicated feature set D
22:03gst568923: VP4.2: NVC0-NVCF (GeForce 400, 500 series; corresponds to VDPAU feature set C)
22:03gst568923: VP5: NVD0-NV108 (GeForce 600, 700 series, a few low 500s; corresponds to VDPAU feature set D)
22:04gnarface: well i can guarantee that it won't actually DO that without the firmware.
22:04gnarface: i can't guarantee getting the firmware will fix it, i can just guarantee that it will definitely not work without it
22:05gst568923: I have downloaded NVIDIA-Linux-x86_64-390.87.run for extracting the firmware
22:05gst568923: but this https://github.com/imirkin/re-vp2 is only for "340.32" version
22:06gnarface: yea. just so you know i've gone through all of this with another card and come up dry. my best advice is to just download both and if one doesn't work, try the other.
22:07gnarface: mine was a VP2 class card with known issues doing this
22:07gnarface: so you may have better luck
22:10RSpliet: gnarface: https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units is actually more complete in some senses
22:11gnarface: ah, good link RSpliet. take a look at this gst568923 ^^
22:12RSpliet: Also, note how NVIDIA cheated their way out of their codename convention by introducing a new one for Turing (TU1xx)
22:12gnarface: gst568923: i have a second piece of advice in case you get this working; augment the stock cooler if you want it to last
22:12gst568923: good link, in this link I read "The GeForce GT 705 (OEM) is a rebranded GeForce GT 610, which itself is a rebranded GeForce GT 520"
22:12RSpliet: They almost clashed with Tesla (GT) :-P
22:12gst568923: what does it means?
22:12gnarface: it means Dell rips people off
22:12gnarface: they're lying cheats
22:12gnarface: simple as that
22:13gnarface: and they've colluded with the hardware vendors to help with the lies
22:13gnarface: it would be illegal if law makers were tech savvy
22:13RSpliet: gnarface: well, that's a bit harsh. It means NVIDIA didn't invest in creating a new low end GPU, but took an old one and placed it in their new naming/coding at it's expected performance point
22:14RSpliet: See how it went from 5*20* (low-end, but could be worse) to 6*10* (the worst) to 7*05* (worse than the worst)
22:14gnarface: at it's expected *price point* you mean
22:14gnarface: and the stock cooler is designed to cook the card early if you try to do any real work with it
22:15gnarface: and they use like DDR2 ram in them or something
22:15RSpliet: Yeah, it's an absolute shit GPU, but that's what you get for getting a card whose last two digits are 05 :-P
22:15gst568923: than the codename is NVD9 (GF119) GeForce 410M, 510 (?), GT (520, 520M, 520MX, 610), 610MQuadro NVS 4200M
22:16gnarface: gst568923: yea, i think so too
22:16gnarface: pretty sure
22:16RSpliet: That's the one
22:17RSpliet: gnarface: NVIDIA is unlikely to continue that game at the same level, Intel pretty much killed their low-end business
22:17gst568923: and than https://nouveau.freedesktop.org/wiki/VideoDecoding/
22:18gst568923: NVD9 YES (VP5)
22:19gnarface: yea, should be VP5 but i think that still implies that firmware is required for this
22:19RSpliet: gst568923: yep, but we can't redistribute the firmware files, NVIDIA does not distribute them individually and nobody has reimplemented the firmwares in the open source domain, so you'll have to use that firmware cutting tool
22:19gst568923: this script is only for VP2 engine? https://github.com/imirkin/re-vp2
22:20gnarface: i think there's no risk to just copying ALL the firmware files from a full install other than the extra 100MB of wasted disk space
22:20RSpliet: There's a step-by-step manual on that wiki
22:21RSpliet: which is remarkably easy to follow if you're on arch or gentoo ;-) but otherwise shouldn't be a lot of trouble either
22:23gst568923: https://nouveau.freedesktop.org/wiki/NVC0_Firmware/ VP5 not have userspace firmware
22:23gst568923: I have Debian
22:23RSpliet: I *just* have you a link with a very easy manual
22:25RSpliet: Do as the romans do, don't worry about perhaps-slightly outdated documentation.
22:29gst568923: For Extracting kernel video firmware on VP3/4/5: Before should I run mplayer with vdpau?
22:30gnarface: no, no point in that
22:30gnarface: you might want to run it to see how much cpu it burns on a reference video if you haven't already
22:30gnarface: but i guess i assumed you were already familiar with the current performance profile and that's why you're trying to get vdpau to work in the first place...
22:31gnarface: you can still play video without vdpau, it just uses more CPU doing it
22:32gst568923: gnarface yes but in this discussion I would run mplayer with VDPAU (more GPU, less CPU usage)
22:32gnarface: gst568923: without the firmware, mplayer will just fail over to software and ignore your vdpau settings
22:33gnarface: you'll see no performance difference
22:33gst568923: I would extract firmware for VP5
22:33gnarface: yes, but i'm not sure it's even different firmware
22:33gnarface: i think it's all the same firwmares...
22:33gst568923: f**ck nvidia property code
22:34gst568923: gnarface I read this: An effort is underway to fully reverse the underlying engines and create open-source firmware to provide out-of-the-box video decoding, please join us in #nouveau on irc.freenode.net if you're interested in helping.
22:34gst568923: who is the open source firmware
22:35gnarface: it's a work in progress
22:36gnarface: i have no idea what progress, but i suspect not much.
22:36gnarface: it's in the realm of "would require divine intervention to make any useful progress"
22:36gst568923: I hope
22:37gst568923: gnarface let's continue the discussion tomorrow because here in Italy is late
22:37gst568923: good night ;-)
22:37gnarface: gst568923: well i'm not an official member of the team but i fully encourage you to come back here to engage them later
22:37gnarface: good night
22:43RSpliet: gnarface: if divine intervention is what it takes to find someone with time and skills to throw together the required assembly...
22:43RSpliet: I think a good chunk of REing has been done already
22:43gnarface: i was under the impression that what remains doesn't work because getting it working would require random arrangement of electrons
22:44gnarface: or some fantastic amount of lucky guessing
22:44RSpliet: At the end of the day, the video decoding engines are just falcon cores with added instructions
22:44RSpliet: So most of the ISA is known afaik
22:45RSpliet: The annoying bits fo'sho, like flow control and scalar instructions ;-)
22:45RSpliet: pardon, control flow
22:46RSpliet: It would take someone with knowledge of video codecs, who is comfortable in assembly
22:49gnarface: maybe you're right but that doesn't account for why the G92s still freeze when attempting to decode h264
22:50gnarface: i'm told it once worked
22:50RSpliet: Ah, but that's not the firmware, as the blob doesn't freeze afaik. Buffer mgmt, broken communication between driver and firmware, incomplete initialisation, a nissing hw workaround... lots of options
22:51gnarface: well i'm betting on incomplete initialization
22:51gnarface: but the current version of the blobs don't even support those cards anymore
22:51gnarface: so the last anyone has seen it working in the wild was several years back
22:53mooch2: hey, i've got a question. doom 2016 on the switch uses a shader instruction called nodep, but what does it actually do?
22:55gnarface: i bet someone idling in here knows the answer to that but it's not me, sorry
22:55gnarface: how well does the new Doom run on that thing anyway?
22:55gnarface: how do the graphics compare to the PC port?
23:11mooch2: gnarface, all it runs are the videos with that stubbed
23:11mooch2: also, it turns out it's called "liveOnly" in nouveau and is basically just a hint instruction
23:11mooch2: meaning yuzu can safely ignore it
23:12mooch2: well, a hint bit in the texture instructions
23:12mooch2: but yeah
23:13mooch2: shit wrong window, sorry