13:11 debianusr81: Well, I have now "re-seated" the graphics card, very carefully, but there is no difference. Same nonsense logged in dmesg and same nonsensical lag when making a selection rectangle on the KDE Plasma desktop. :(
13:11 debianusr81: This kind of thing is just...
13:13 MrCooper: it's most likely some kind of HW issue
13:14 MrCooper: the PCIe link to the graphics card is unable to reliably transmit data
13:14 MrCooper: you can try playing with some PCIe related settings in the BIOS setup
13:17 debianusr81: Will have to try that next, I guess... :(
13:17 debianusr81: But since it's this computer, my only one, I can't do it while IRCing.
13:18 debianusr81: Anything in particular I should be looking for and change it to?
13:38 HdkR: PCIe slot version
13:38 HdkR: Drop it down versions until hopefully the errors stop 4.0 -> 3.0 -> 2.0
13:56 debianusr81: That actually sounds like something which could be it, because the PCI Express slot in the motherboard is 3.0 but the card is 4.0.
13:57 debianusr81: But I was told that it's backwards-compatible, so it never occurred to me to set anything in the BIOS settings.
13:57 HdkR: It's backwards compatible, the system will auto-negotiate the connection to 3.0
13:57 HdkR: So forcing 2.0 will be interesting
13:58 debianusr81: Wait... since the mobo is 3.0, I guess it won't even have an option for 4.0... so 2.0... hrm... that's worse than what the hardware supports. :(
13:59 debianusr81: Any other option for me to try if that doesn't help?
13:59 debianusr81: (It's so cumbersome to get back in here.)
14:00 HdkR: Trying another slot on the motherboard if at all possible
14:01 debianusr81: The only other 16x slot is much slower "electrically" according to the mobo's manual, so using it would probably only make it worse...
14:02 debianusr81: (It's 4x electrically but 16x mechanically.)
14:02 HdkR: Busted connection versus "slow" connection, I think slow would still win
14:02 debianusr81: No visual signs of damage on it, BTW, but of course that probably says little.
14:02 debianusr81: I guess I will try that next, if the BIOS change doesn't work...
14:04 debianusr81: And I guess you have a point. It's not like I'm maxing out the real performance of the card. All I want is to be able to use my desktop environment without lag and stuff.
14:04 MrCooper: you won't notice slightly slower PCIe link speed in general, only with specific use cases which transfer large amounts of data across the link
14:04 dwfreed: (like video games)
14:04 debianusr81: It's also impossible for me to know what the difference is in speed between PCI Express versions and between the "x"es.
14:05 MrCooper: dwfreed: even most games won't be significantly affected
14:05 dwfreed: the speed of a single lane basically doubled between version 2, 3, and 4; x16 means there are 16 lanes
14:05 dwfreed: but the amount of throughput available even in a PCIe 2.0 x1 is a lot
14:05 MrCooper: doubling per lane is still just doubling with the same number of lanes
14:06 debianusr81: Yeah, but they make a difference between "mechanical" and "electrical" compatibility (the current slot is 16x for both but the other one is 16x mechanically and 4x electrically).
14:07 debianusr81: Alright, I might as well try it now...
14:41 debianusr81: First of all, I restarted the machine and set all the PCIE slots to "Gen 2" (were set to the maximum "Gen 3"). No difference. Same sluggish selection rectangle.
14:42 debianusr81: Then I tried changing the graphics card into the other mechanical 16x slot, but something physically prevented the left part of the card to slide in, no matter what I tried and no matter how much pressure I applied.
14:42 debianusr81: I eventually ended up putting it back into the first 16x slot and now I'm back to square one. :/
14:42 debianusr81: I even removed the cable underneath (front audio) which I initially suspected was blocking the left part of the card, but this was not the case.
14:45 _ds_: Maybe put the old card back in to see if the PCI errors persist?
14:46 debianusr81: You mean the NVIDIA card? I had extreme problems with that, which I always attributed to NVIDIA not working well on Linux according to numerous people.
14:47 debianusr81: Moving objects around on the desktop was extremely sluggish and weird with that card.
14:47 debianusr81: The sole reason for buying this AMD card was to be able to use my computer normally on Linux.
14:48 dwfreed: that the issues have happened with two cards suggest that it's not the cards
14:51 debianusr81: What do you mean? Isn't it known that NVIDIA cards work horribly on Linux?
14:51 _ds_: Checking kernel logs from before the new card was installed may suffice.
14:51 debianusr81: I made a fresh installation of Debian after the new card was installed because at first, there was no difference on the desktop at all, so no such logs.
14:52 dwfreed: nvidia card support has gotten a lot better
14:52 MrCooper: it does sound like some motherboard HW issue could explain all your symptoms
14:52 dwfreed: nouveau makes great progress, and the nvidia proprietary driver is generally good as well
14:53 dwfreed: currently typing this on a 10 year old computer with an ancient nvidia card, using the proprietary driver that works for it (which is old and deprecated, but still functions with modern kernels!)
14:53 dwfreed: I have no issues with my computer
15:12 debianusr81: Well, I certainly do. >:( <:(
15:12 debianusr81: Might very well be semi-broken hardware.
15:14 debianusr81: But again, how can something work at all if it's broken? Seems like it wouldn't be able to feed a picture at all, or do any kind of work.
15:15 debianusr81: But it does render pictures on the screen without artifacts or flicker, just in an extremely sluggish manner for certain things, and seemingly "pausing" every so often briefly.
15:16 debianusr81: If I grab a Dolphin window and drag it around, it's not as smooth as one would reasonable expect it to be. It "stutters".
15:17 debianusr81: The mouse cursor itself is perfectly responsive, though.
15:17 dwfreed: things aren't always so black and white
15:17 MrCooper: the messages in dmesg mean the PCIe link is able to recover from at least some of the transmission errors, that presumably involves retrying the transmission though, which might tank performance
15:18 debianusr81: The "smooth lag" when making a selection rectangle is the most annoying part IMO.
15:18 debianusr81: Hmm.
15:18 MrCooper: in other words, I wouldn't worry about performance as long as there are PCIe transmission errors
15:19 MrCooper: the HW isn't working as intended
15:19 debianusr81: Well, I would buy a new computer right now if money weren't an issue, even though this one is not yet 5 years old.
15:19 debianusr81: Seems like the motherboard is broken indeed. :(
15:19 dwfreed: replacing a motherboard is usually cheaper
15:20 debianusr81: Yeah, but then it has to be compatible with all the existing parts and soon enough, I've replaced all the parts.
15:20 dwfreed: it most likely is
15:21 dwfreed: the largest compatibility conflict with motherboards is with CPUs, and in some rare cases, RAM (usually when we're on the cusp of switching to a new RAM standard)
15:21 dwfreed: so you just buy the right motherboard for your CPU (and possibly your RAM), and everything else generally Just Works
15:22 dwfreed: (and of course the right size that fits in your case, though if it's got 2 x16 slots, it's probably ATX)
15:42 debianusr81: I agree that it's sound advice, but at the same time, I know my "luck" with hardware... Things have a tendency to get... complicated. I'm leaning more into the idea of buying a second machine and keep this one for backup, as I really need a minimum of two usable units as a minimum safety precaution anyway... Still, sure would've been nice to know for sure that it was the mobo...
15:44 dagelf: Anybody keen to come help out building a python interface to speak directly to AMD drviers? https://www.twitch.tv/georgehotz
15:45 dagelf: I know there are some AMD veterans in here who has trodden this path..
19:13 Venemo: dagelf: what do you mean by directly to AMD drivers?
19:14 Venemo: debianusr81: considering it's a 5 year old motherboard, it shouldn't be prohitively expesive to get another one of its generation or maybe just borrow one to do your tests
19:22 dwfreed: Venemo: debianusr81 is gone
22:32 sgm: https://paste.debian.net/plain/1312641 - can regular user try to debug this further, or it's enough info to fill a bug report?
22:34 sgm: It was stable for hour or two of game play, then from what I see GPU just crashed
23:30 Venemo: sgm: There is no useful info in there, but perhaps if you gave more info (eg. what game it is, what GPU you have etc.) then it'd be enough for a bug report.
23:54 _ds_:sees “Tainted: P OE” in that log text