Gaming crashes KDE

Overclocking is a per Silicon thing. So you have to find the sweet-spot for yourself.

Start with 50Mhz, run test play games, if stable increase for 50Mhz more, and so forth until you will not start to crash. If it crashes decrease 50Mhz and increase instead for 10 and repeat the process, if crashes go back one step and crease by 5, repeat, crashes go back by 5 increase by 1.

Just so you know OC GPU is lengthy process, many times its not even worth it, cause it depends on-the Silicon lottery.

As @Seimus already wrote, it’s about finding the sweet spot.

However, to quickly test whether this is even the issue, I would go really low, like 25% below default and then test whether the crashes stop. If they don’t, we’re on the wrong track.

At this point I have it down to 1900. I’ve just been decreasing it a little bit more each time it crashes and each time it seems to do better and better. So far haven’t had a crash at 1900 so that might be what it likes if not I’ll go down to either 1850 or 1800 but it does seem like this is fixing my issue.

Thanks for the suggestion

Maybe it fixes your crashing, but by under-clocking it you take performance from it.

Each GPU should be stable at its reference default/max boost clock without actually touching it.

If you GPU stabilizes when you under-clock it, this points to a bigger issue.

2 Likes

I thought of this as kind of first step in narrowing it down, because I actually had a Radeon card last year that could not run stable on stock settings either.

It might be worth playing with both memory and GPU limits. In my experience memory causes much less performance decrease, barely noticeable in many cases. So if that does the trick it would be better than limiting the GPU.

Next step would be to run a benchmark and watch things like temperature, which could be related to the issue, but doesn’t have to be.

And if I remember correctly, it was also possible to overvolt in Adrenaline. Within reason, that could be used to stabilize the card on higher clocks. It’s important to watch the temps though when doing that because they could get higher than stock.

EDIT: 1 minute later it turns out, the card still crashes at 1800, so forget what I said…

1 Like

Alright, so update… I have under clocked it clear down to 1800 and it is still crashing… it might be a bit better but not completely fixed. I will have to take a look at some of the other troubleshooted steps that have been requested to see if I can narrow it down even further or supply details that help find the cause… Not sure if I’ll be able to get a photo of Htop that is relevant since it black screens and then when the screen initially comes back up the image is from a minute or 2 ago and it rarely freezes then blacks out.

From what I can tell, temps look normal and Im not seeing any spikes but I might not be looking in the right spot to find temps

I would have to agree with some others in this thread that it looks like it may be some issue with the VRAM, and possibly faulty hardware. First step I would suggest is testing VRAM with memtest_vulkan-git that is in the AUR.

To install it, open up Konsole (or terminal of your choice) and use this command:

paru -S memtest_vulkan-git

(remember to press Q to exit out of the review of AUR packages)

memtest_vulkan-git doesn’t require any special set up to run. Just run it and let the test run for at least 6 minutes. You can stop the test with Ctrl+C keys. More information about it here:

4 Likes

Here are the results of the VRAM test:

An interesting thing i just remembered, I first started out using the KDE lite version of Garuda and I don’t remember having any issues with gaming using that version and I had the same PC setup, I think the issues have only started after staring to use KDE Dragonized so that I wouldn’t have to install as many things in the initial setup.

Please do not post pictures of the terminal. They are not searchable and may be difficult to see on different devices. Please always paste terminal output directly here and format it with a code block. Thanks. :slight_smile:

1 Like

I completely forgot about that. If I run it again I will copy and paste.

Unfortunately these next photos have to be uploaded as phone photos as I can’t get them from the PC as they are from when it freezes and resets.


Is the power usage consistently at 50W? Have you limited something somewhere? What kind of power supply do you have?

This seems to be the case, I haven’t limited it myself; according to LACT, Power usage limit is 289/332. I have a Modular 1000W PSU with 3 separate PCIe 8 pin cables going to it.

Edit: the difference between the 2 photos on settings is that the first image had auto fan control which only turned the fans on shortly before freezing, 2md image has manual 50% fan control

I this reset only took like 3-5 minutes to happen this time:

I will do the memory test again and will paste the results in a code block this time.

What load did you run when look at the graphs? The power usage can be explained by low utilization of the GPU Core.

Try to run the most demanding GPU bound game you have and check if Power usage goes up with GPU Utilization and Core Clock.

2 Likes

I was playing ATM 10 with shades active and 64/128 PureBDCraft Resource pack.

Hmmm.. Im not sure what the most demanding game I have is… I would think either Call Of Duty or Boarderlands 2…

Here is the VRAM test again

https://github.com/GpuZelenograd/memtest_vulkan v0.5.0 by GpuZelenograd
To finish testing use Ctrl+C

1: Bus=0x10:00 DevId=0x73BF   16GB AMD Radeon RX 6900 XT (RADV NAVI21)
2: Bus=0x00:00 DevId=0x0000   32GB llvmpipe (LLVM 19.1.7, 256 bits)
Override index to test:1
Standard 5-minute test of 1: Bus=0x10:00 DevId=0x73BF   16GB AMD Radeon RX 6900 XT (RADV NAVI21)
1 iteration. Passed  0.0211 seconds  written:    3.0GB 446.9GB/sec        checked:    6.0GB 416.8GB/sec
49 iteration. Passed  1.0188 seconds  written:  144.0GB 445.4GB/sec        checked:  288.0GB 414.1GB/sec
285 iteration. Passed  5.0009 seconds  written:  708.0GB 444.9GB/sec        checked: 1416.0GB 415.3GB/sec
1698 iteration. Passed 30.0065 seconds  written: 4239.0GB 443.6GB/sec        checked: 8478.0GB 414.6GB/sec
3118 iteration. Passed 30.0133 seconds  written: 4260.0GB 446.2GB/sec        checked: 8520.0GB 416.3GB/sec
4537 iteration. Passed 30.0069 seconds  written: 4257.0GB 445.9GB/sec        checked: 8514.0GB 416.1GB/sec
5941 iteration. Passed 30.0125 seconds  written: 4212.0GB 439.9GB/sec        checked: 8424.0GB 412.2GB/sec
7335 iteration. Passed 30.0010 seconds  written: 4182.0GB 436.1GB/sec        checked: 8364.0GB 409.8GB/sec
8746 iteration. Passed 30.0211 seconds  written: 4233.0GB 442.5GB/sec        checked: 8466.0GB 413.9GB/sec
10146 iteration. Passed 30.0102 seconds  written: 4200.0GB 438.1GB/sec        checked: 8400.0GB 411.3GB/sec
11554 iteration. Passed 30.0114 seconds  written: 4224.0GB 441.4GB/sec        checked: 8448.0GB 413.3GB/sec
12968 iteration. Passed 30.0009 seconds  written: 4242.0GB 444.1GB/sec        checked: 8484.0GB 414.9GB/sec
Standard 5-minute test PASSed! Just press Ctrl+C unless you plan long test run.
Extended endless test started; testing more than 2 hours is usually unneeded
use Ctrl+C to stop it when you decide it's enough
15761 iteration. Passed 30.0117 seconds  written: 4227.0GB 442.0GB/sec        checked: 8454.0GB 413.4GB/sec
17176 iteration. Passed 30.0008 seconds  written: 4245.0GB 444.8GB/sec        checked: 8490.0GB 415.0GB/sec
18588 iteration. Passed 30.0063 seconds  written: 4236.0GB 443.1GB/sec        checked: 8472.0GB 414.3GB/sec
19979 iteration. Passed 30.0057 seconds  written: 4173.0GB 435.2GB/sec        checked: 8346.0GB 408.8GB/sec
21389 iteration. Passed 30.0130 seconds  written: 4230.0GB 442.2GB/sec        checked: 8460.0GB 413.8GB/sec
22808 iteration. Passed 30.0184 seconds  written: 4257.0GB 445.7GB/sec        checked: 8514.0GB 416.0GB/sec
24206 iteration. Passed 30.0128 seconds  written: 4194.0GB 437.8GB/sec        checked: 8388.0GB 410.5GB/sec
^C
memtest_vulkan: no any errors, testing PASSed.
press any key to continue...

Also two questions:

  1. Do you have RE-BAR enabled in bios
  2. How is your power connector to the GPU connected, are you doing daisy chain?
1 Like

No? I’m not sure though. I think this is the page with it:

In text form

Advanced\PCI Configuration 
Please go to Boot page to disable CSM
*Greyed out* Above 4G decoding : disabled
*Greyed out* Re-Size BAR Support : Disabled
*Selectable* SR-IOV Support : Disabled

Ok RE-BAR is disabled, which is good cause it can sometimes cause weird problems.

And what About the power connectors on GPU, are you daisy chaining?

No, each one is separate directly to the PSU

Check whether it goes over 50W or is limited there.

2 Likes

I saw the power get up to 145W, so doesn’t appear to be a power limit.. I played Ark Survival Evolved for 15-25 minitues with no crashing just runnign around. IDK if I should play longer to see if it will crash.