Hi Exil,

you can run that command. It will sow you youyr grafic card and driver in use with.

lspci -nnk | egrep -A3 -i "3D|VGA"

You could try glxgears, although it's little use because it uses such a tiny sub-set of OpenGL. It may not even be around anymore.

Check your kernel driver with and your X driver module in /var/log/Xorg.0.log.

Interesting... All I can say is that I would have expected temperatures to be somewhat higher.
This could mean that your cooling setup is **very good**, the system is misreporting temperatures or your system power profile is massively underclocking your CPU.

Your Intel Core i5-8400 has a Processor Base Frequency of 2.80 GHz which should rise to 4.00 GHz (Max Turbo Frequency) on demand.
It is possible that some firmware setting is incorrectly throttling your CPU (unlikely, but I "never say never").
I'll have a look at the mobo manual to see if there are any clues there...

Another possibility is that your current kernel does not have full Coffee Lake support (I'm unsure without looking at the current kernel spec).
I do know that Sky Lake support has improved with the later stable kernels for my Sky Lake CPU.
This would be easy for you to check.. You can add the latest stable kernel easily by using a program called UKUU. This will add a chosen kernel without removing any others. If you want any help here, just ask. I use this one on my Mint installations.

You appear to be using the modesetting drivers.. (or were, as shown in your inix output listing). This is the "recommended" default that "should" be perfectly fine. All I can say here is that it is not 100% for my GPU (Intel UHD 620) on Mint. This is why I'm running a lower resolution in Mint for now. I haven't had the time to investigate this fully yet.

Try temporarily changing your monitor resolution to 1152x864. This is a wide screen resolution. Re-run your tests and see if your problems still exist.
This should at least give some indication if the problem is some sort of kernel/xorg interaction or if it lies elsewhere.

I'll have a look at the mobo manual and see what's what..

Bodge99

I don't see where the OP is from, but I've done hardware. If he is sitting in a fridge at zero degrees, he would need a total (package + thermal paste + heatsink) of under 0.5 degrees C per watt to stay at 30 degrees of junction temperature.

Heatsinks at 0.5 degrees are on sale, but not small or neat ones. I've heard no mention of liquid nitrogen, so for a 6 core i5 + GPU to be running at ≤30 degrees, there has to be no load, no screen, 5 cores off and a force 9 gale blowing through the cabinet. I've a twin core i3 here, no load worthy of the name (<10% CPU in top) and I'm showing 52ºC with lm_sensors. Mine is also a lower power cpu (35W vs 65W).

Install lm_sensors if it's not there. Run 'sensors-detect' and accept defaults. Then run 'sensors.' Post the output.

No I am not sitting at 0 degrees

When I bought this pc I was a bit worried about overheating because I had that problem on my previous pc. The CPU heatsink would get clogged up by dust and the computer would overheat. So this pc I wanted at better solution and I wanted it to be more quiet too. As I know nothing about hardware (or software for that matter) I went and got advice and bought a Panteks Eclipse p400 to take care of the noise and for cooling Cooler Master Hyper 212X and Cooler Master MWE Bronze 650 for power supply. So to me 30ºC sounded plausible, but from what you say business_kid, that may not be that plausible.

I ran the sensors-detect and sensors and this is the output:

coretemp-isa-0000
Adapter: ISA adapter
Package id 0: +25.0°C (high = +82.0°C, crit = +100.0°C)
Core 0: +23.0°C (high = +82.0°C, crit = +100.0°C)
Core 1: +24.0°C (high = +82.0°C, crit = +100.0°C)
Core 2: +22.0°C (high = +82.0°C, crit = +100.0°C)
Core 3: +24.0°C (high = +82.0°C, crit = +100.0°C)
Core 4: +24.0°C (high = +82.0°C, crit = +100.0°C)
Core 5: +22.0°C (high = +82.0°C, crit = +100.0°C)


Anyway I must say that we have come a very long way from were we started and the system works much better

OK. I'll accept that sensors output. You have an extremely efficient cooling setup, no load, and you're not consuming the 65W claimed for your cpu. In fact, unless that goes up when you work it, your fault lies there. I ran some car racing on youtube and cpu temperatures jumped 12 degrees. Glxgears shows 60Hz (full or small screen) and glxinfo |grep render has the magic words in it FYI, here XFCE offers 4 VTs in X. I opened 2 browsers in 2 of them got 2 different fullscreen HD videos of car racing going on the race bit (Rallying is harder work than F1 for the gpu), and started swapping between them quickly for a few minutes; No problem getting 70 degrees. How does yours do when you make it work?

I loaded a Youtube video with a car racing game and then I loaded two different slot games https://yggdrasilgaming.com/games/trolls_bridge/#tryit and a random game from unibet.com. I have to run games like this for work and running two at a time in different browsers and on different screens would cause the system to freeze before. Now it is handling it quite ok.

The glxinfo |grep render output is:

direct rendering: Yes
GLX_MESA_multithread_makecurrent, GLX_MESA_query_renderer,
GLX_MESA_multithread_makecurrent, GLX_MESA_query_renderer,
Extended renderer info (GLX_MESA_query_renderer):
OpenGL renderer string: Mesa DRI Intel(R) HD Graphics (Coffeelake 3x8 GT2)
GL_ARB_conditional_render_inverted, GL_ARB_conservative_depth,
GL_NV_conditional_render, GL_NV_depth_clamp, GL_NV_packed_depth_stencil,
GL_ARB_conditional_render_inverted, GL_ARB_conservative_depth,
GL_MESA_window_pos, GL_NV_blend_square, GL_NV_conditional_render,
GL_OES_element_index_uint, GL_OES_fbo_render_mipmap,

And the temperatures were:
coretemp-isa-0000
Adapter: ISA adapter
Package id 0: +51.0°C (high = +82.0°C, crit = +100.0°C)
Core 0: +51.0°C (high = +82.0°C, crit = +100.0°C)
Core 1: +48.0°C (high = +82.0°C, crit = +100.0°C)
Core 2: +50.0°C (high = +82.0°C, crit = +100.0°C)
Core 3: +49.0°C (high = +82.0°C, crit = +100.0°C)
Core 4: +46.0°C (high = +82.0°C, crit = +100.0°C)
Core 5: +44.0°C (high = +82.0°C, crit = +100.0°C)

Things that do happen from time to time however can be seen on the two attached video clips

https://youtu.be/9fVkACiKJGg
https://youtu.be/P4OF3r-zIzQ you will see the slot game behind the terminal window

Have you tried running these videos with a lower screen resolution?? Just to see if these errors still appear...

Bodge99

I've a laptop, so things run hotter anyhow (small fan, heat sink, low cooling power etc.) but things on your machine do appear normal, except for your problem.

Any time firefox freezes on me, there's a script running. Things digging for information 'gold' or hacks for windows systems. Firefox eventually times out and asks do I want to stop it - and I always answer yes. Do you keep a minimized terminal open? I would, and try to use it. Something like for example, should get you out of it.

bodge99 I did the test with the resolution set to 1152x864 and that went very smoothly indeed so I guess that would be the final solution. It is just such a horrible resolution to work with I wish there was another way.

business_kid I have not been able to do the sudo pkill before, because the computer was not responsive at all. After the adjustments we have made I may be able to do so.

I am however happy with how the computer works now and I am very grateful to you both for the help . So thank you both!

Hi, I just meant testing performance/stability at a lower resolution, not as a permanent solution. Obviously you want to be able to use your hardware to its fullest potential!!

It would appear that you have a similar problem to me.. I'm determined to get to the bottom of this and I have a few ideas on how to proceed. Time will tell!

If I find a solution then I'll post here.

Bodge99

OK. You might get the 'sudo pkill' option to work now.
If you think it works, don't fix it. If you want it better, read on.

If 1152x864 works, and 1920x1080 doesn't, your system's video is marginal. 1920x1080 is 2,073,600 dots per frame, 1152x864 is 995,328 dots per frame. Video only rewrites only those parts of the screen that have changed, and if you have a lot of stuff going on onscreen, there's a lot of rewriting to do. Someone reading the news is easy. But the dash cam of a rally car as it goes off the road and turns over is a lot of rewriting. If your WM supports VTs in X, you can switch screen in X by pressing Ctrl_F1, Ctrl_F2, etc. A good test is to get videos going on 2 or 3 and swap between them (requiring a full rewrite).

Things you can do: Check all drivers: Lower the refresh to the minimum you find acceptable. On the bigger screen size, try reducing the window size gradually to cope with the lack in speed. Your screen might be 1920x1080, but the bit that changes is smaller. Also, your sizes are skewed: 1920x1080 is 16:9 aspect ratio; 1152x864 is 4:3 aspect ratio is a squarer screen (= 16:12 when you think of it). 16:9 sizes are 1600x900; 1440x810; 1280x720. Also consider upgrading the video hardware - whatever form that takes. Intel GPUs suck, as a rule. I believe 8th Gen cpus have AMD GPUs, if you have one of those. But I am not the man to consult on leading edge or bleeding edge hardware. Trailing edge - that's my area.

I think I will try to live with the system as it is for now, and if it becomes unbearable I will continue

Hi,
I don't know why I wrote 1152x864... of course this is a 4:3 resolution.. my bad!

I'm assuming that my particular problem is similar to Exil's (similar hardware, common hardware design ancestry..).
I get severe random screen flickering, ghosting partial images from earlier screens and random partial screen blanking. All with a static desktop (no moving images). Only at 1920x1080 in Mint.

It would appear to be some interaction between the kernel and Xorg (I'm suspicious of CPU/GPU throttling effects) as these problems only occur in Mint. [EDIT: I'm hearing rumours that a component of systemd my be involved...]

Slackware and Devuan are perfectly fine on the same hardware using modesetting drivers at 1920x1080 (same kernel version).

The video hardware is not marginal at 1920x1080.. It is designed and certified for this resolution. I agree that Intel GPUs are "not the best". Later versions are much improved and are perfectly adequate for most "non-gamers".

I'm determined to get to the bottom of this.. I want to find the exact cause of the problem without relying on workarounds.
Intel graphics should be "rock-solid" with a modern distro.

Bodge99

We were using those numbers for years - that's why a 4:3 size popped into your head.

Partial old images is because when the screen has changed, you're far too slow to rewrite the whole thing, so it rewrites a portion (say 1/12th) and, along with rewrites, it may take a long time to get the whole screen up.

It's slow video. Besides the i915 kernel module, Intel GPUs may benefit from

1. xf86-video-intel dri drivers
2. vaapi drivers, a vdpau replacement afaict (IIRC i965_drv.so)
3. Vulkan sdk - some thing for managing multiple GPUs

Have you got all those? I've recently been through video players, and only a correctly compiled vlc was using vaapi. The rest seemed to look for vdpau, then fall back to sucky graphics. If your Mint is in a VM, of course, you're dead in the water, because it's all passed to the main system via vboxvideo unless it's a separate monitor for VMs.