Another sad Nvidia Linux Mint Cinnamon

[rapid brook]

It's my first day on Linux, and its been a little painful. I have a prebuilt pc with a Gigabyte Motherboard and 3060ti, it previously ran windows, the SSD Linux is currently installed on is fresh. I disabled secure boot and used the Driver Manager to install the drivers. 580 (recommended) and 570 both result in a "no signal" upon boot. I hit esc while the logo was still going to get into what I assume is a safe mode, then had success with 535. I decided to test things out with a game on steam (which really narrows down problems) Age of Darkness, I think it was an Unreal game, worked with proton 10, but I had no sound. With proton experimental i got an error about my gpu not supporting shader level 5 and... d3d11 or something along those lines, forgot to screen shot that. I was hoping somebody could point me in the right direction. As far as I can tell the 535 driver installed correctly, I do hear system sounds, and as far as I can tell it works. But I am pretty far out of my comfort zone atm. Any help/guidance would be appreciated. I am an aspiring game dev, so having the gpu function correctly with shaders/vulkan is pretty important.

[empty flicker]

does 550 exist?

[lethal jolt]

which kernel are you running?

[empty flicker]

(uname -r)

although most likely 6.14

[lethal jolt]

I'm running my old GTX 970 on 580 driver with 6.8.0 kernel. No issues.

his card is much newer

[lusty flame]

Do you dualboot?

[lethal jolt]

no, just linux

[lusty flame]

I meant @rapid brook

[lethal jolt]

oh sorry. I'll be quiet 🙂

[rapid brook]

Hey, Not dual booting, I just removed the SSD with windows and havent put it in yet, preferably i never will 🤣

when i booted today... immediately after the grub i went straight to pure black screen, and was forced to power down and try again, it wasnt until i booted with the usb and hit restart there.. that somehow this worked this time? idk

[lusty flame]

Have you disabled fast startup in windows?

[rapid brook]

I dont remember if i disabled that when i got the pc in the first place... thats in the bios right?

[lusty flame]

No, that's in windows

[rapid brook]

oh, in that case... the only windows on this pc atm is the boot info in the bios. windows has never touched the 1 sdd in the pc right now

going through dmesg, i found 2 reds

make that 3

[lusty flame]

What does

inxi -SGxxx

say?

And

glxinfo -B

[lusty flame]

Looks ok. What was the problem again?

[rapid brook]

well... when i just turned it on today... it boots to a black screen... where i have not choice but power down

hitting esc just shows the grub briefly before black screen. I just started trying linux yesterday... so I likely just dont know hot to properly access safe modes or debugs idk...

[lusty flame]

But you're clearly booted now

[rapid brook]

yeah... i dont know why... this time worked and the 2-3 boots before didnt

the only difference this time... was i booted from usb. then restarted

[lusty flame]

Is this a desktop or laptop?

[rapid brook]

desktop

[lusty flame]

hmm.... Show output of

lsblk -f

and

cat /etc/fstab

[lusty flame]

Do you have a "fast boot" option in bios?

[rapid brook]

dunno... let me go look

It's disabled

so... yeah... just booted again no problem... seems intermittent 😞

[rapid brook]

OH! found some crash logs from my boots right before talking to you!

[lusty flame]

What are windows 10 features, and show me secure boot options

[rapid brook]

secure boot is disabled, that screen doesnt show that until you "click" on it

the crashes always start here

[rapid brook]

ill jump into the bios again...

[lusty flame]

Does it boot ok now?

[lusty flame]

Use "reset to setup mode", just to be sure and save settings

[rapid brook]

I can't click or move to the ones with the yellow blocks, including reset to setup mode

[lusty flame]

Some bioses require that you set a password first. Just remember it so you can disable it after

[rapid brook]

So... If I set secure boot to custom, I can restore factory keys and enter audit mode, reset to setup still gets skipped

Then is I switch back to standard

I've entered no each time and it shows standard again

[lusty flame]

In the security tab, there should be an option to set a password

[rapid brook]

Set an admin password?

[lusty flame]

most likely

[rapid brook]

Hmmm still can't get to the reset

Do you think this is the source of the headache?

[lusty flame]

Can you access key management then?

[rapid brook]

Yeah

[lusty flame]

What's there?

[lusty flame]

What are those yellow squares?

Write down the password you've set, btw

[rapid brook]

Yeah got it written

Setting secure boot to disabled didn't stop this all from firing? I feel pretty uneasy about changing anything here

[lusty flame]

maybe you need to save and restart - if you enter with the password maybe they'll go away

[rapid brook]

Ok so setting the password made before bios activate require the password

No changes in the bios

I cleared it back out

The bios is one of the Y versions, that I don't think I can flash to a more current from the manufacturer, it was prebuilt

Secure boot it set to disabled, and it remembers the last setting when it boots, so I can switch as needed.

[lusty flame]

check in mint with

mokutil --sb-state

[lusty flame]

Good. So do you still get a black screen on boot?

[rapid brook]

i think i have booted 4-5 times in a row with no problem? same was true last night though...

https://github.com/nulls0x0/mint-setup/blob/master/mint-22_2-nvidia-gaming-setup-guide.md is this a better way to install the nvidia drivers?

[lusty flame]

No, you install nvidia drivers from driver manager

[rapid brook]

i see... then my current result is the best it will get?

[lusty flame]

Rtx cards should handle driver 580

[rapid brook]

it should, but 570 and 580 make it unusable. after the reboot from installing the driver, i get no signal.

its had to do much of anything with no visual

[lusty flame]

can you screenshot the driver manager window?

[rapid brook]

is there a way to restore the point before boot? in the event that trying those drivers break things again?

[lusty flame]

Yeah, it recommends 580-open and you try regular one

sudo apt install nvidia-driver-580

[rapid brook]

so... do it in the terminal vs the driver manager?

[lusty flame]

via terminal

[lethal jolt]

I don´t have the open drivers, is that related to the kernel version in use?

[lusty flame]

What they call -open is only partially open, and can be wonky on some systems

[rapid brook]

5 sanity checks. I hope it didnt roll and 1s 🤣

its done what it was doing. I kinda expected to be prompted to reboot. also... whats up with the nonexistent thing?

[lusty flame]

what does

dkms status

say?

[lusty flame]

looks ok. try restarting

[rapid brook]

ok

[lusty flame]

try reconnecting hdmi cable

[rapid brook]

i hit the power button on the computer, and it shut down a few moments later... thats the only thing that seems to do anything i can tell there

booted again, his esc when the linux logo popped up and now i am back on

doesnt looks like it populated a crash log this time

[lusty flame]

so.... when exactly does it hit the black screen?

[rapid brook]

i see the gigabyte, then the linux logo. it happens before desktop appears and before password prompt

when it just happened, and i reboot afterwards, i always hit esc during the logo

things go really fast there... but i always see some nvidia stuff in red... orange this last time

[lusty flame]

Ok are you on the desktop now?

[rapid brook]

yeah

[lusty flame]

what does

dmesg | grep -i nvidia

says?

[rapid brook]

sorry.. what key do i press to get that I?

[lusty flame]

shift + \

[lusty flame]

huh.... can't read EDID

Do you have another hdmi cable?

[rapid brook]

its.. probably because i am using a tv as a monitor... and the tv does suck

[lusty flame]

tv as a monitor?

[rapid brook]

i get that a lot 🤣

[lusty flame]

Does your tv have a Pc mode or game mode?

[rapid brook]

it has always worked before... and it does detect the native 4k resolution. I changed it to 1080p because things were too small

[lusty flame]

I meant settings on the tv

[rapid brook]

i dont really mess with those... the tv just randomly freezes, and i occasionally have to factory reset it. its like they designed it with a memory leak or something... 10/10 never buying an element tv again

on the other hand... i have dealt with it enough that i can recognize when it is the source of the headache vs a gaming console or pc

[lusty flame]

What does

xrandr

say?

That TV is likely causing trouble, not drivers or mint

[rapid brook]

i dont think a tv could cause crash logs

[lusty flame]

It's late here, I'll continue tomorrow

[rapid brook]

I appreciate the help!

this was definitely an improvement though! got sound in a game (using playing a game in steam as a baseline to see if things will work out for me)

[rapid brook]

update* tested 5 steam games... all of them work! I think the gpu trouble is good now! that last install worked!

[rapid brook]

This is probably true now. I havent found any issues outside of the tv now, no more crashing. If it doesn't pick up a signal right the first time, i get no signal now, but typing the password brings the screen back when the desktop opens up, with mixed results

I really appreciate all of your help! You are awesome!

[lusty flame]

My bet is that you're occasionally get a black screen / no signal because of wonky display (tv) EDID detection. Do you want to work on that?

[rapid brook]

oh wow... you'd help with that?

it seems to be intermittent. i ran the terminal command that revealed the edid issue earlier, it is not saying it couldnt get them right now

[lusty flame]

We can try and see if it gets better

[rapid brook]

I set it to game mode... But I am pretty sure this "smart TV" just thinks that means darker 🤣

The only thing i can think of is to update the firmware and pray the manufacturer finally cared enough to fix things in the software

I am 90% sure the cable is not an issue, I bought one specifically for it shortly after getting it

even in windows it occasionally just went black screen, but switching inputs back and forth always fixed the problem, it seems the linux driver may be slightly more picky. But another thing this TV does is become unresponsive (which i never notice until i try to change the volume or turn it off, also... this is when it performs the best 🤣 , until i have to unplug it)

and it had the same behavior with consoles as well. I just went through all the settings and theres no passthrough on the OG firmware

[lusty flame]

We can try capturing the EDID on a "good boot", then force mint to use it regardless of TV being flaky about it. Hopefully that'll fix most of the no signal outcomes

[rapid brook]

oh dude you are awesome!

[lusty flame]

After I had breafast then

[lusty flame]

Ok, let's see what port is your tv connected to. In the terminal run

for p in /sys/class/drm/*/status; do echo -n "${p%/status}: "; cat "$p"; done | grep -i hdmi

@rapid brook

[rapid brook]

alright

[lusty flame]

HDMI-A-1 then. Ok, next type these.

sudo apt install read-edid edid-decode -y
sudo get-edid > ~/element-tv.bin
edid-decode ~/element-tv.bin

[rapid brook]

alright.. dont see anything that failed...

[lusty flame]

What does last command say?

[rapid brook]

oh ... i should have done those 1 at a time

[lusty flame]

preferably, but it seems it worked

Next type these

sudo mkdir /lib/firmware/edid
sudo cp ~/element-tv.bin /lib/firmware/edid
ls /lib/firmware/edid

What does last command say?

[lusty flame]

ok. next type

sudo xed /etc/default/grub

[lusty flame]

Yeah, you need to edit 1 line so that it'll look like this:

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash drm.edid_firmware=HDMI-A-1:edid/elevent-tv.bin nvidia-drm.modeset=1"

[lusty flame]

yeah. save the file and close the editor, then run

sudo update-grub
sudo update-initramfs -u -k all

[rapid brook]

the terminal populated this when i hit save

[lusty flame]

that's fine

[rapid brook]

ok... so close it and continue

[lusty flame]

yeah

[lusty flame]

ok. now restart it a few times, see how it goes

[rapid brook]

kk

ok.... rebooted 3 times with 3 different results

1st time, went no signal, i typed the password then it came up perfect

rebooted second time, went no signal, typed password, then resolution was set to 4k and nothing had x on it, and the bar at the bottom did not populate, also couldnt launch anything or shutdown via keyboard commands, had to hard power off, a quick press of the power button did nothing, had to hold it.

3rd time... just worked like it should 🤷‍♂️

[lusty flame]

And 535 driver booted fine every time?

[rapid brook]

i didnt check, but the instal of the 580 you walked me through in the terminal seems to have worked reliably

is there a way to slow down the grub when you push esc on the linux logo during boot?

[lusty flame]

not really, why?

[lethal jolt]

you can see the boot messages in a log somewhere

[rapid brook]

just thought it would be easier to catch stuff. its usually a .5 second

oh

[lethal jolt]

/var/log and then boot.log I think

[lusty flame]

dmesg command shows the full sequence

but it's long

[lethal jolt]

install Gnome-logs for a graphical interface

it shows the errors on one screen

[lusty flame]

Anyway, let's try something. type

sudo xed /etc/default/grub

and add video=HDMI-A-1:e to the end of the line we edited earlier

[rapid brook]

thats from the gnome log

like so?

[lusty flame]

that's interesting.... close the file, don't save and run

dmesg | grep -i edid

[lusty flame]

What does

ls -l /lib/firmware/edid/element-tv.bin

say?

[lusty flame]

left side seems cut off

[rapid brook]

thats where it starts

[lusty flame]

Ok, let's try this

sudo mv /lib/firmware/edid/element-tv.bin /lib/firmware/element-tv.bin
sudo rmdir /lib/firmware/edid
ls -l /lib/firmware/element-tv.bin

@rapid brook

[lusty flame]

Ok, next type

sudo xed /etc/default/grub

and change the line to

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash drm.edid_firmware=HDMI-A-1:element-tv.bin nvidia-drm.modeset=1"

[lusty flame]

good. Save and close, then run

sudo update-grub
sudo update-initramfs -u -k all

[lusty flame]

ok, restart and run

dmesg | grep -i edid

[rapid brook]

heres one before reboot

wish me luck

[lusty flame]

dmesg | grep -i edid

AFTER reboot

good luck

[rapid brook]

that looks conserably different

[lusty flame]

No errors this time. Ok restart a few times, see how it goes

[rapid brook]

Ah can't really see

I used the welcome thing to reach the display, which is populating blanks

Is there a way to shutdown without cutting power?

[lusty flame]

try ctrl+alt+t

if the terminal comes up, type reboot

[rapid brook]

Currently synced to display DFP-0

Couldn't get the terminal up

[lusty flame]

You don't get black screens anymore, but the brightness is off?

[rapid brook]

This time small res, no bar can't get terminal

3 fails in a row, should I load a snapshot?

[lusty flame]

how do you bring these programs up then?

[rapid brook]

I keep the welcome going

It's often the only thing I can interact with when it boots wrong

[lusty flame]

try ctrl+alt+esc

[rapid brook]

That... Made the desktop background appear?

Oh and hid the things on the desktop

[lusty flame]

can you get the menu?

[rapid brook]

I just cut the power

4 fails

[lusty flame]

Ok, that doesn't work it seems

can you press ctrl+alt+f3 ?

[rapid brook]

That... Made the screen black, the lost signal

[lusty flame]

hmm

ctrl+alt+f7 then

@rapid brook

[rapid brook]

heres the logs, finally got an ok boot

gonna go for a smoke

[lusty flame]

Ok, let's undo what we did then. In the terminal type

sudo xed /etc/default/grub

change the line back to original GRUB_CMDLINE_LINUX_DEFAULT="quiet splash"

save the file, close the editor and run

sudo update-grub

[rapid brook]

should i go ahead and reboot?

[lusty flame]

yeah. if it goes black screen, just type your password and hit enter

[rapid brook]

Loaded wrong again, 4k res and no background/bottom bar

Ctrl Alt t does not bring up terminal

[lusty flame]

you mentioned having to reset the tv now and then

[rapid brook]

Yeah that's usually when a no signal shows up, switching inputs fixes it on windows, just power off the TV and unplugged it, then turned it back on

No change on the Linux side

[lusty flame]

Well I undid the changes I made, so I guess get a normal monitor / tv. I'm out of ideas

[rapid brook]

Yeah...

heres the logs from the 2 failed boots

[rapid brook]

I just did a restore to right after the gpu update

also found that by adding terminal as an icon on the desktop, you can access it before the system fully freezes

[grizzled brook]

this could also be a potential GPU seating issue

since it also happens in Windows it's pointing to a hardware related issue, either GPU or monitor, in case GPU that would be potentially not seating correctly resulting in decreased signal quality/instability

[rapid brook]

ive been investigating a bit more. Maniak is right about the tv being part of the problem. I should assume worst case scenario on handshakes. I think the boot problems are cause by race conditions that have been identified on the 580 driver on linux

I thought i was on a "good boot" right now because everything looks right... but the pcei looks like it defaulted to gen 1 on this boot, i might try setting it to default to gen 3 in the bios

[rapid brook]

in this case i just switched from windows a few days ago, where it performed just fine. I doubt it is hardware or seating related.

[grizzled brook]

try the gen 3 in BIOS

[rapid brook]

VBIOS Version : 94.04.6B.00.F0

i just learned about this too... looks like my vbios on the gpu is not new enough for 580 driver

https://docs.nvidia.com/datacenter/tesla/tesla-release-notes-580-82-07/index.html