pathetic on-board GPU performance

[n0body_special]

Hi all
My graphics card just died and shall be honorably remembered.
I switched to my onboard graphics card, which after an inxi command I found out is an "Intel HD Graphics 530", whatever the heck that is. I know onboard graphics chipsets are nothing near comparable to normal graphics cards, but performance of my onboard chipset is really nothing more than ABSOLUTELY PATHETIC.
The mouse pointer is flickering, keyboard strikes delay about 1 second, and CPU usage is stuck 80%-90% when just moving my mouse pointer around. The simplest application takes 10-30s to open. I know that on-board chipset performance is bad, but this bad??? Computer is practically unusable for any productive work.
Driver manager says don't worry dude, your computer does not need any additional drivers but I sure as hell know the driver manager is wrong, because my system really really needs a driver, because, how to put it, 30 years ago computers used to be much faster when simply moving the mouse pointer around, than my pc is right now.

Any ideas?
My inxi -Fxz says:

Code: Select all

System:
  Kernel: 5.4.0-72-generic x86_64 bits: 64 compiler: gcc v: 9.3.0 
  Desktop: Cinnamon 4.8.6 Distro: Linux Mint 20.1 Ulyssa 
  base: Ubuntu 20.04 focal 
Machine:
  Type: Desktop Mobo: MSI model: Z170-A PRO (MS-7971) v: 1.0 
  serial: <filter> UEFI: American Megatrends v: 1.D0 date: 08/30/2016 
CPU:
  Topology: Quad Core model: Intel Core i5-6600K bits: 64 type: MCP 
  arch: Skylake-S rev: 3 L2 cache: 6144 KiB 
  flags: avx avx2 lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx 
  bogomips: 27999 
  Speed: 3659 MHz min/max: 800/3900 MHz Core speeds (MHz): 1: 3601 2: 3600 
  3: 3600 4: 3600 
Graphics:
  Device-1: Intel HD Graphics 530 vendor: Micro-Star MSI driver: i915 
  v: kernel bus ID: 00:02.0 
  Display: x11 server: X.Org 1.20.9 driver: modesetting unloaded: fbdev,vesa 
  resolution: 3840x2160~60Hz, 1680x1050~60Hz 
  OpenGL: renderer: Mesa Intel HD Graphics 530 (SKL GT2) v: 4.6 Mesa 20.2.6 
  direct render: Yes 
Audio:
  Device-1: Intel 100 Series/C230 Series Family HD Audio 
  vendor: Micro-Star MSI driver: snd_hda_intel v: kernel bus ID: 00:1f.3 
  Device-2: Arkmicro USB2.0 PC CAMERA type: USB 
  driver: snd-usb-audio,uvcvideo bus ID: 1-10:4 
  Sound Server: ALSA v: k5.4.0-72-generic 
Network:
  Device-1: Realtek RTL8111/8168/8411 PCI Express Gigabit Ethernet 
  vendor: Micro-Star MSI driver: r8169 v: kernel port: e000 bus ID: 02:00.0 
  IF: enp2s0 state: down mac: <filter> 
  Device-2: Ralink RT2870/RT3070 Wireless Adapter type: USB 
  driver: rt2800usb bus ID: 1-11:5 
  IF: wlx00c0ca58e854 state: up mac: <filter> 
Drives:
  Local Storage: total: 6.11 TiB used: 2.99 TiB (48.9%) 
  ID-1: /dev/sda vendor: Mushkin model: MKNSSDTR480GB size: 447.13 GiB 
  ID-2: /dev/sdb vendor: Seagate model: ST6000DM003-2CY186 size: 5.46 TiB 
  ID-3: /dev/sdc vendor: Patriot model: Burst size: 223.57 GiB 
Partition:
  ID-1: / size: 16.81 GiB used: 6.13 GiB (36.5%) fs: ext4 dev: /dev/sdc1 
  ID-2: /home size: 174.69 GiB used: 40.72 GiB (23.3%) fs: ext4 
  dev: /dev/sdc3 
  ID-3: /usr size: 27.37 GiB used: 7.72 GiB (28.2%) fs: ext4 dev: /dev/sdc2 
Sensors:
  System Temperatures: cpu: 41.0 C mobo: 29.8 C 
  Fan Speeds (RPM): N/A 
Info:
  Processes: 260 Uptime: 30m Memory: 7.67 GiB used: 3.45 GiB (45.0%) 
  Init: systemd runlevel: 5 Compilers: gcc: 9.3.0 Shell: bash v: 5.0.17 
  inxi: 3.0.38

It can't be THAT slow...

[rene]

It can't be THAT slow...

Indeed it can and is not. You are using the correct (and fairly good) "modesetting" driver for your 530 and are correctly configured for 3D acceleration. You are driving one 4K and one smaller additional display it seems making for a fair number of pixels to push out the door but it should undoubtedly handle even that fine and certainly in regular desktop rather than 3D situations.

As such it seems you will be experiencing something at least in essence unrelated to your graphics hardware's capabilities as such. I.e., is some bit of left-over software that expects to find the now-deceased former graphics hardware acting up or some such? Does top show anything of that nature?

[n0body_special]

Thanks for helping rene

I tried shutting down as many programs as possible. Nothing changes. I can't find anything strange. Ofc other than the fact that when moving a window around, Xorg and Cinnamon cpu usage go off the charts. The other thing is that the mouse pointer flickers on the main 4k display connected on displayport, but doesn't flicker on the secondary display which is on DVI.

My system at rest:


My system while dragging around a window:

[senjoz]

How does your system behave, if you boot in live session from installation media? If you disconnect the 4k monitor, is there any difference?

Have you any idea why the graphics card died? Is it really dead? Maybe something is wrong with the motherboard.

Regards, Jože

[ajgreeny]

Have you uninstalled the nvidia driver you were using?

I'm not sure why it should be necessary to do so but can't think of anything else to suggest. My Intel 620HD performs very well with no delays so I agree there is probably some other underlying problem not specifically related to your integrated graphics.

[rene]

Yes, I suppose that testing how things behave from a Live-system session and/or with one or both the displays disconnected from boot onwards is indeed one of the better things to try. Certainly that kind of CPU usage does not count as expected.

If with both displays connected the symptoms are the same on the Live system and although not generally recommended, I guess you can still try switching to the "intel" Xorg driver from the current "modesetting" one. To do so create as root if needed the directory /etc/x11/xorg.conf.d/ and a file /etc/X11/xorg.conf.d/intel.conf as

Code: Select all

Section "Device"
	Identifier	"Device0"
	Driver	"intel"
EndSection

and log out and back in or reboot. inxi -Gx should after that show "intel" rather than "modesetting" and, maybe, your problem/bug is helped.

There's another report of this here: viewtopic.php?t=323952. That poster also needed a specific bugfix to keep the "intel" driver from crashing but let's for now hope you don't. Certainly it's the case "modesetting" should normally be preferred and "intel" is a basically dead driver, but it seems you in your multiple-display case may still need it if "modesetting" is buggy.

[EDIT] More information on the MESA_LOADER_DRIVER_OVERRIDE thing if you in fact still need that as well to keep things from crashing: https://linuxreviews.org/Intel_Iris. But first try without.

[n0body_special]

I discovered the source of the problem (but no solution to it).

Normally I use the command: xrandr --output HDMI-2 --pos -1680x-0 --scale 2x2 ,which causes the small monitor to use normal (1x) UI scaling while the larger one stays on double DPI as I have chosen. It also moves the small monitor 1680 pixels (as much as its horizontal resolution) to the left of the large one. I have put this command to autostart with the system at every boot. This setting has always worked fine for me for a long time, that is, until I removed the GPU.

I discovered that If I disconnect either of the two monitors and then reconnect it, the above setting is reset and the problem is resolved. The flickering is gone, everything is fast, but the UI scaling is wrong.
If I give the command again (xrandr --output HDMI-2 --pos -1680x-0 --scale 2x2) then the problem comes back, the system lags like crazy and the mouse pointer is flickering again.

So I guess the xrandr command messes something up. I don't know much about this command, I did enormous online research and spent many hours of trial and error, to even end up with this command that worked for me.

All this because Mint does not allow for separate UI scaling in each screen in a dual screen setup. You have to either pick normal-1x or double DPI for both screens. This effectively forces you to use same size screens, because normal 1x DPI is useless for a 4k screen, as everything is too tiny, and double DPI is useless for a 1680x1050 monitor, as everything is gigantic and you cant even fit a single window.

Note: in windows everything works fine, you can't even tell the GPU is gone... I hate windows, I don't want to return to them just for this problem

[rene]

That command does not seem to want to make a huge amount of sense to me. You scale 2x2 so I take it that HDMI-2 is the 4K display. But then why do you position it at -1680x-0? I take it because you position the 1680 display to the right of the 4K one but this seems to put the origin of the X screen at 1680/3840 of the 4K screen. Do very much compensate for me not using dual-display and as such probably having utterly misunderstood all of that, but how does

Code: Select all

xrandr --output HDMI-2 --primary --pos 0x0 --scale 2x2 --output DVI-?? --pos 3840x0

fare, of course with the proper DVI-?? for the 1680 screen?

If that completely does not work you'll need to otherwise ignore me and hope for someone with a clue to come along: I don't as said in fact use dual-display on Linux and never have in combination with HiDPI scaling.

[n0body_special]

It's weird but 2x2 actually means half the dpi. Thus, from a display of hi-dpi (2x), it means normal (1x).
If you "man xrandr" and go to "--scale" you will see:

Code: Select all

--scale xxy
              Changes the dimensions of the output picture. Values superior to
              1 will lead to a compressed screen (screen dimension bigger than
              the dimension of the output mode), and values below 1 leads to a
              zoom  in  on the output. This option is actually a shortcut ver‐
              sion of the --transform option.

I find it strange myself, but I didn't create xrandr
The secondary screen it to the left of the primary, and has a res of 1680, so a --pos -1680x0 makes its desktop end where the primary one starts. The result is this:


@senjoz: the card is hardware damaged. I'm pretty sure because when booting, some huge green vertical stripes appear on both displays and booting freezes. The card was an ATI Radeon R9 back from 2015. It was time for it to die.
@ajgreeny: as my card was an ATI, I never installed any nvidia drivers.
@rene: switching from modesetting to intel caused the display to fill with little colorful polygon shapes (but otherwise operational -no freezing). I returned to modesetting, it was worth a try however, thanks.

Until I discover something else, I am returning to double dpi on both screens

[rene]

I had noticed you had above said the secondary was on DVI (although you had also said that the primary was on DP) but I take it then that should've been HDMI. Okay, if your 4K display is the primary display it makes sense to keep its origin at (0,0) so makes sense to me then and at least lays to rest a foggy notion I had that with the odd origin X ended up doing fractional scaling internally, thereby accounting for the huge CPU load.

All I've left to suggest is a repeat of the suggestion to try with the old "intel" rather than "modesetting" driver.

[EDIT] Never mind; see you already did; missed it on the first read. I'll withdraw: no idea.