[SOLVED] Crazy crash when using 2 RAM modules (32 GB) - lines across the screen and repeating sound

[kubajz 22]

Hi, I've been facing this issue for a while now, searched a lot of forums, tried many things but I still don't know what is going on here. Basically, I bought 2 new Kingston 16 GB RAM cards and put them in my laptop, replacing single 8 GB card. But since then my computer randomly crashes when I use Firefox. And in a very strange way indeed. I thought that this problem only occurs when I watch videos on Youtube, but today it happened when I was browsing through download history. I recorded a video some time ago of what exactly is happening. Turn on your sound for the 'best' experience: https://www.youtube.com/watch?v=meeulBNrUHY I immediately tested both RAMs using memtest86, but no problem found. Then I removed one of the modules and the problem went away. After about a week of testing by regular use of the computer, I put the module in the other slot to test if that isn't damaged. Nothing. I did the same thing with the other module and everything worked fine. Each module was in each slot for at least a week, actually I've been running with only one for months and no crashes happened. Only when I use both of them at the same time. I need 32 GB because I often run photogrammetry software which uses a ton of memory. This can run all day and I've never had a single crash. I've noticed some error messages when the system is booting, which I believe weren't there or at least not as many of them when I used only one module.

This led me to think that it might be related to audio (nvidia hdmi, which I'm not using) and I disabled the audio device in /etc/udev/rules.d and while the messages didn't go away, which I assume is because the device gets loaded before I disconnect it, I can no longer see it in the output of lspci -vv. But the system still crashes. This is what the output of journalctl looks like.
Click to enlarge.

I can't see anything suspitious. I even installed linux-crashdump which should create a log when something bad happens, but it doesn't in my case. The computer crashes about once or twice a day and it's very unregular. But only when I use Firefox (I've not tried other browsers yet). I wasn't able to find anything on the internet which would describe what I'm experiencing.
Sorry for the long post, but just to summarize:

> Computer crashes only when I connect both modules and use Firefox.
> Never crashed when only one module was in use. Doesn't matter which one in which slot.
> Photogrammetry often runs for many hours and is very resource demanding, but it never crashes, even when I use both modules.
> Color of the streaks isn't always the same. I had green and red as well.

[rene]

I expect a hard 32G limit of your hardwware is with 32G of RAM causing your GPU's framebuffer to overlap RAM which although it supposedly should work (check your system manual) does in fact not. First of all rummage through your BIOS and flip settings for IOMMU/VT-d/AMD-Vi and/or "64-bit I/O address decoding" or anything of the sort. If it needs Linux to fix, the word to google for is MTRR...

[kubajz 22]

Thank you so much for your reply. According to the manufacturer, my computer (acer nitro 5) has a limit of 32 GB, so I thought it would work fine. The BIOS is a joke, but I believe I saw some VTD setting there. I'll investigate and post an update.

[rene]

Acer Nitro 5 seems to exist as a series of systems with very different specs. As such you'll want to post the output of inxi -Fxz for a closer look (although not, currently, by me; am off to bed)

[manyroads]

@kubajz 22 You might check to see if your BIOS actually sees your memory, if you have not already. I think your machine allows you access to your BIOS via F2. Look to see if the memory count is correct 32GB.... If that is off by a bunch, you may have the cards incorrectly seated (take them out & put them back with your power off). Reboot & recheck to see if they are okay. You may also want to check to make certain you have up-to-date (most current) bios software. That's always a joy to update (not!). 9 times out 10 the above fixes this sort of problem for me.

If it doesn't work for you... please do what @rene suggested.

[kubajz 22]

Yes, the reported memory size is correct. Modules have been reseated so many times that I have lost the count. I will post the output of the suggested command when I wake up. I think I live in a similar timezone to rene's..

[kubajz 22]

The output of inxi -Fxz is:

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: Laptop System: Acer product: Nitro AN515-54 v: V1.23 serial: <filter>
Mobo: CFL model: Octavia_CFS v: V1.23 serial: <filter> UEFI: Insyde v: 1.23
date: 10/25/2019
Battery: ID-1: BAT1 charge: 44.4 Wh condition: 44.4/57.5 Wh (77%) model: LG PABAS0241231
status: Full
CPU: Topology: Quad Core model: Intel Core i5-9300H bits: 64 type: MT MCP arch: Kaby Lake
rev: A L2 cache: 8192 KiB
flags: avx avx2 lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx bogomips: 38400
Speed: 4000 MHz min/max: 800/4100 MHz Core speeds (MHz): 1: 4001 2: 4000 3: 4029
4: 4092 5: 4002 6: 4091 7: 4100 8: 4073
Graphics: Device-1: Intel UHD Graphics 630 vendor: Acer Incorporated ALI driver: i915 v: kernel
bus ID: 00:02.0
Device-2: NVIDIA TU117M [GeForce GTX 1650 Mobile / Max-Q]
vendor: Acer Incorporated ALI driver: nvidia v: 460.56 bus ID: 01:00.0
Display: x11 server: X.Org 1.20.9 driver: modesetting,nvidia
unloaded: fbdev,nouveau,vesa resolution: 1920x1080~120Hz
OpenGL: renderer: GeForce GTX 1650/PCIe/SSE2 v: 4.6.0 NVIDIA 460.56
direct render: Yes
Audio: Device-1: Intel Cannon Lake PCH cAVS vendor: Acer Incorporated ALI
driver: snd_hda_intel v: kernel bus ID: 00:1f.3
Device-2: NVIDIA driver: snd_hda_intel v: kernel bus ID: 01:00.1
Sound Server: ALSA v: k5.4.0-72-generic
Network: Device-1: Realtek RTL8111/8168/8411 PCI Express Gigabit Ethernet
vendor: Acer Incorporated ALI driver: r8169 v: kernel port: 3000 bus ID: 07:00.0
IF: enp7s0 state: down mac: <filter>
Device-2: Intel Wi-Fi 6 AX200 driver: iwlwifi v: kernel port: 3000 bus ID: 08:00.0
IF: wlp8s0 state: up mac: <filter>
Drives: Local Storage: total: 709.83 GiB used: 260.97 GiB (36.8%)
ID-1: /dev/nvme0n1 vendor: Micron model: 2200 MTFDHBA512TCK size: 476.94 GiB
ID-2: /dev/sda vendor: Western Digital model: WD2500BEVT-24A23T0 size: 232.89 GiB
Partition: ID-1: / size: 467.96 GiB used: 260.96 GiB (55.8%) fs: ext4 dev: /dev/nvme0n1p2
Sensors: System Temperatures: cpu: 63.0 C mobo: N/A gpu: nvidia temp: 38 C
Fan Speeds (RPM): N/A
Info: Processes: 299 Uptime: N/A Memory: 31.02 GiB used: 955.5 MiB (3.0%) Init: systemd
runlevel: 5 Compilers: gcc: 9.3.0 clang: 10.0.0-4ubuntu1 Shell: bash v: 5.0.17
inxi: 3.0.38

I was looking for 64 address decoding in BIOS but I didn't find anything. This is a part the output of lspci -v (the GPU):

01:00.0 VGA compatible controller: NVIDIA Corporation TU117M [GeForce GTX 1650 Mobile / Max-Q] (rev a1) (prog-if 00 [VGA controller])
Subsystem: Acer Incorporated [ALI] TU117M [GeForce GTX 1650 Mobile / Max-Q]
Flags: bus master, fast devsel, latency 0, IRQ 150
Memory at a3000000 (32-bit, non-prefetchable) [size=16M]
Memory at 90000000 (64-bit, prefetchable) [size=256M]
Memory at a0000000 (64-bit, prefetchable) [size=32M]
I/O ports at 4000 [size=128]
Expansion ROM at a4080000 [virtual] [disabled] [size=512K]
Capabilities: <access denied>
Kernel driver in use: nvidia
Kernel modules: nvidiafb, nouveau, nvidia_drm, nvidia

[rene]

If it were the specific issue I was vaguely recalling/thinking of I imagine I'd quickly find more reports giving those full specs --- but that ain't happening. As such, let us retreat to more generic steps. Most promisingly, I at

https://www.acer.com/ac/en/US/content/s ... t/7964?b=1

see that's there's a much newer BIOS available: 1.33 from 2021-03-02 vs. your 1.23 of 2019-10-25. I'd definitely update that, and not even just as a very generic step: if more people were in fact experiencing this chances that a newer BIOS does something are fair.

The "manual" seems to be as to BIOS options as big of a joke as what you said said settings themselves were so I can't comment further on that/those. I would've potentially commented to try and lower RAM frequency and/or flip a current XMP settings, or..., even if there seemed a little directly of interest setting available.

If the new BIOS and/or settings don't help I'd make very sure you have the latest binary Nvidia driver installed; I'm not an Nvidia user so which that is I couldn't tell you but presumedly the Mint "Driver Manager" would be helpful.

If nada still, I'd try switching to the open source "nouveau" driver for your 1650 even if only to test. Note though; although I'm sure you'll have gotten that machine to use its graphics to the fullest "nouveau" isn't unaccelerated or anything like that; could perform the same for your purposes.

Or, what I'd in fact undoubtedly try first is see if this happens on Windows with the newest Nvidia driver. If so that says quite a bit (although including that you may not be able to get it solved).

After that we'd be at suggestions to try with a "nomodeset" boot which is going to end up useless since even if that works for you, being without a graphics driver at all will not work for you.

Anycase. First of all trying that newer BIOS seems like a very good idea. Hope you still have the Windows install around to do so from; in true Acer fashion the README for the BIOS helpfully instructs only to "Click EH50F133.exe under Winodows mode". Yes, I'm sure you'd never have figured that much out yourself --- assuming you'd have been the kind of person capable of understanding what was meant by "Winodows mode" in the first place of course.

[kubajz 22]

I was just trying to find a way to install the new BIOS under Mint, but from what I've read it doesn't sound too promising. This laptop came with some sort of linux, which I immediately replaced with Mint. I have a HDD with win7 though. I can try to put it in and run the update from there. What do you think about running the executable using Wine? Can it destroy something if it fails?
Regarding the graphic drivers - I tried to switch to the intel integrated GPU, but it didn't help, although the streaks looked a little different. It actually started to crash even when the computer was doing nothing, just displaying the desktop. Switched back to nvidia proprietary driver which should be the most recent. I don't think it will help, but I can try to use the nouveau.
One thing that came to my mind is to plug in the original 8GB RAM module along with one of the 16GB ones for total memory of 24 GB which is below the limit. Although 24 GB is not sufficient for me (actually 32 can barely cut it and even then the system often has to swap).

[rene]

What do you think about running the executable using Wine? Can it destroy something if it fails?

If it works at all it probably works --- but it probably doesn't work at all and even if does I'd in fact very much advise against it anyway even when the bit before the hyphens is technically probably true. If a BIOS flash fails it has potential to fully brick your system so when judged in the context of (the probability theory concept of) expected value I'd quickly shy away from Wine. Some of that will be due to me shying away from Wine quickly period, but still, don't do that. On a system as new as that having a small Windows 10 partition around for events precisely as that would be something I'd advice anyway. Although it seems then you don't have a Windows 10 license? You probably do even if you yourself didn't receive it with Windows 10: a new Windows 10 install would probably be automatically activated when online. Even if not note you can always run a Windows 10 copy unactivated legally for 30 days or so. Long enough at least to flash a BIOS. And yes, maybe that Windows 7 install works.

It's very relevant that the same happens with the Intel integrated graphics as that basically nullifies all of my notions of the issue up to now except the generic BIOS-update one. In the context of my own first reply in this thread I would expect, or would HAVE expected prior to that Intel information, that you'd "solve" the issue. Note though that it would be a sucky solution even if 24MB were enough: two unequal DIMMs would force the machine into quite suboptimal single channel mode. I'd do my utmost to avoid that, and then first of all update the BIOS. Some way.

[kubajz 22]

@rene In your first post you were talking about MTRR. What did you mean by that?

BTW: I'm still trying to run the Windows, but it didn't work for the first time so I'm currently installing a version specifically for UEFI on a USB stick. If that fails again, I'll probably try to rip the installation DVD and install Windows onto a spare HDD.
I also tried running a live version of Mint, but it crashed as well. Still I think that the lines I see on the screen are just a part of RAM that isn't image data. I had an idea of opening a bunch of images, minimizing them and when it crashes I should technically see parts of the images on the screen. If that hypothesis is correct.

[rene]

The Memory-Type Range Registers are CPU-internal configuration registers that control how the CPU caches accesses to the address regions they describe. You can imagine that e.g. DMA or I/O-memory as mapped in from some piece of to the CPU external hardware, such as the GPU, has to generally be treated differently than regular RAM: former has the potential to change out from under the CPU so that it needs to take care to not wrongly cache values in its CPU-internal caches; same basically for the write direction: I/O-memory should generally not be written back at leisure such as RAM can be, but be written through the caches immediately since the external hardware needs to see the chance ASAP and in some particular order.

I when reading your post had a recollection of reading about a not so old Linux bug where in cases of overlapping actual RAM and I/O memory --- something for which there would be no alternative with 32G of RAM if your system had a hard limit of 32G physical even for I/O --- incorrect MTRR settings got configured, but even then I failed to find what I was in fact seemingly remembering. Hence the lame google advise...

I've moreover also not found that issue again when I could search more specifically using your posted specs so it might be better to ignore me there. cat /proc/mtrr lists your MTRRs but frankly your hardware is quite significantly too new for me to then also authoritatively comment further on; most knowledge I have at that level sort of expired at the shift from 32-bit to 64-bit already.

Do I take that you haven't yet managed to update your BIOS? Even this MTRR thing is the kind of thing that could supposedly be solved with a newer BIOS if Linux is now fed wrong memory-range information by the BIOS for example. I'd be quite keen on not even trying to look at anything else before that update.

[kubajz 22]

I was trying to run the Windows but it just didn't boot (UEFI system), even though it boots on my old laptop (BIOS). I'm currently creating bootable USB specifically for UEFI using WinToUSB. It takes a long time. If everything fails though I'll probably go the Wine route.. I was looking at the output of the /proc/mtrr already but I frankly don't know what does that output mean. Could I somehow determine from it if the memory has been actually mapped wrongly?

[rene]

Could I somehow determine from it if the memory has been actually mapped wrongly?

Probably not if there's an actual bug; something as obviously comparable as that would supposedly not have managed to survive a single kernel release candidate.

The format of /proc/mtrr is fairly obvious; sets of base, size regions with a cacheability attribute. In the case of this old system for example, with an uncacheable 512M address region overlaying my 256M (Intel integrated) graphics framebuffer mapping region:

Code: Select all

rene@hp8k:~$ lspci -s 00:02.0 -v
00:02.0 VGA compatible controller: Intel Corporation 4 Series Chipset Integrated Graphics Controller (rev 03) (prog-if 00 [VGA controller])
	Subsystem: Hewlett-Packard Company 4 Series Chipset Integrated Graphics Controller
	Flags: bus master, fast devsel, latency 0, IRQ 16
	Memory at f0000000 (64-bit, non-prefetchable) [size=4M]
	Memory at e0000000 (64-bit, prefetchable) [size=256M]
	I/O ports at 1230 [size=8]
	[virtual] Expansion ROM at 000c0000 [disabled] [size=128K]
	Capabilities: <access denied>
	Kernel driver in use: i915
	Kernel modules: i915

rene@hp8k:~$ cat /proc/mtrr 
reg00: base=0x000000000 (    0MB), size= 4096MB, count=1: write-back
reg01: base=0x0dd800000 ( 3544MB), size=    8MB, count=1: uncachable
reg02: base=0x0de000000 ( 3552MB), size=   32MB, count=1: uncachable
reg03: base=0x0e0000000 ( 3584MB), size=  512MB, count=1: uncachable
reg04: base=0x100000000 ( 4096MB), size= 4096MB, count=1: write-back

I'd as said have doubts that I'd necessarily correctly interpret results for your much newer hardware --- and would also as said not really expect the issue to necessarily be as visible as odd cacheability attributes. And chances of all this being relevant at all seemed much higher before the information that the issue's also present with the integrated graphics anyway.

[SMG]

If everything fails though I'll probably go the Wine route..

I also would not recommend doing that to update UEFI.

Have you tried running with the 5.8 kernel available in Update Manager? That would probably be a better fit for your 9th gen Intel processor than the 5.4 kernel.

Open Update Manager. Select View > Linux Kernels and click Continue. Make sure 5.8 is selected on the left panel and then click the top-most option on the right panel. An "Install" button will appear. Install the kernel and then reboot for it to become active.

[kubajz 22]

Have you tried running with the 5.8 kernel available in Update Manager? That would probably be a better fit for your 9th gen Intel processor than the 5.4 kernel.

T̶h̶a̶t̶ ̶t̶o̶t̶a̶l̶l̶y̶ ̶f̶i̶x̶e̶d̶ ̶i̶t̶.̶ N̶o̶ ̶w̶a̶y̶.̶ It's been playing videos for hours and everything is ok. S̶e̶e̶m̶s̶ ̶t̶h̶a̶t̶ ̶i̶t̶ ̶w̶a̶s̶ ̶a̶ ̶l̶i̶n̶u̶x̶ ̶i̶s̶s̶u̶e̶ ̶a̶f̶t̶e̶r̶ ̶a̶l̶l̶.̶
I've edited the title to make it more descriptive and easier for others to find it.
Thank you all for help.
T̶h̶r̶e̶a̶d̶ ̶o̶v̶e̶r̶.̶

[rene]

In that case apologies for leading you astray. 5.4 is Ubuntu's 20.04 LTS kernel series so it was to be expected especially on your not brand new hardware that it would in fact be the better choice: many more users with potentially the same issue (and fixes for serious kernel issues discovered later get backported).

[kubajz 22]

I'm afraid I spoke too soon. While it doesn't happen nearly as often, I still experienced another crash this evening while watching youtube.

[rene]

You probably know what my advise is

[kubajz 22]

It will be very difficult. I wasn't able to run windows to go even though it was specifically made to boot under UEFI. Actually it loaded once but after I selected windows 7 it just blackscreened and now it doesn't even show up.