PC crashes one to several times a day

Ploppz · 2017-08-18 23:08:57

It has only happened the past few days. (My build is a few weeks old) At least once a day, my PC running Arch Linux crashes.

What happens is that the screen just freezes and I have to do a hard reboot. There's nothing particular in journalctl (only some sparse dhcp stuff).

On boot, this message is always marked red in journalctl: `sp5100_tco: I/O address 0x0cd6 already in use`.

I unloaded the `sp5100` module during the previous session. This time (that is, the last time) when the crash happened, it wasn't a screen freeze, but rather the screen just went black and the music (from mpd) stopped playing.

I also tried to disable the module by putting `blacklist sp5100_tco` in `/etc/modprobe.d/blacklist-custom`, but upon reboot the module still gets loaded. Anyway, as the sp5100_tco module wasn't running when the crash happened this time, that's probably not the problem.

Any ideas?

CPU: AMD Ryzen 5 1600
GPU: Radeon HD 7850.
Mobo: MSI B350M Mortar

Last edited by Ploppz (2017-08-18 23:09:25)

circleface · 2017-08-18 23:17:18

What kind of cooling do you have? A while back there were lots of Ryzen owners whose systems kept freezing and crashing due to the processor overheating. Does it tend to crash under load, or just when idle, or both?

Ploppz · 2017-08-18 23:33:45

I'm using two 140mm intake fans in the front, and one 120mm exhaust fan in the back. I monitor GPU temp (lm-sensors can't find any others) and it's always around 38 Celsius. Sometimes the crash happens while I'm away from PC and at that time, there should be very low load on the CPU. I will try to make the fans go faster. But it's strange to me that it should overheat. What temperature is 'overheat'?

As a blind shot.. I recently changed which power outlet I use for the PC. Could a non working ground connection in the wall socket have anything to do with crashes like these?

Last edited by Ploppz (2017-08-18 23:34:24)

circleface · 2017-08-18 23:41:04

I don't actually use a Ryzen myself, but some people were reporting temps near boiling (90 celsius)... However, that was a kernel bug I believe and it has since been fixed. I doubt a non-working ground could cause a crash, but uneven power certainly could. You might want to check the whole outlet with a meter and make sure the power is steady. On that same note, a bad power supply could also cause crashes.

Ploppz · 2017-08-19 08:14:15

As for power supply unit, I have `be quiet! - Pure Power 10 500W`. I heard it should be good. I might check the wall socket, or see if crashes continue when I change socket.

In the meantime, any other ideas welcome of course.

alex.theoto · 2017-08-19 08:27:46

Try the lts kernel, although these symptoms are from faulty hardware.
Check if you have bad capacitors on your motherboard, graphics card and power supply.
Run memtest at least 2-3 times in the row.

Edit: Your motherboard doesn't suffer from old capacitors.

Last edited by alex.theoto (2017-08-19 08:37:57)

Kabbone · 2017-08-19 18:00:44

There have been (or perhaps still are) some problems with Ryzen. Are you already using the newest BIOS for your mainboard?
I also got random freezes with a Ryzen 7, always when in Idle or kind of. It's definitely not a temperature problem here. Right now I'm trying the newest BIOS to check if it's gone. I think this is the only way you can go right now if the power supply or any hardware has no damage.

Ploppz · 2017-08-19 19:39:32

@alex.theto:
How would lts kernel help? (I don't really know what it is)
How can I check for bad capacitors? Just a check on the surface? I shouldn't open the PSU, it will void the waranty.
I will try memtest.

@Kabbone
I haven't upated the BIOS. Seems like there is one update since I built the PC (the latest https://www.msi.com/Motherboard/support … RTAR.html) (Edit: actually it currently runs version 1.1 so there are quite a few revisions available). I might try to update the BIOS. If I understand correctly, all I have to do is copy the downloaded files to an MBR/FAT32 mem stick, boot to the BIOS and start the flashing.

Edit: When `sudo mkinitcpio -p linux`, I get these warnings. Could they be related?

==> WARNING: Possibly missing firmware for module: wd719x
==> WARNING: Possibly missing firmware for module: aic94xx

modinfo: https://bpaste.net/show/752b9d06f377

Last edited by Ploppz (2017-08-19 20:26:54)

ewaller · 2017-08-19 20:49:14

Ploppz wrote:

How would lts kernel help? (I don't really know what it is)

This is probably a dead end. Regardless, https://en.wikipedia.org/wiki/Capacitor_plague

WFV · 2017-08-19 21:59:15

How would lts kernel help? (I don't really know what it is)

the long-term-service kernel

How can I check for bad capacitors? Just a check on the surface?

yes, capacitors usually are very visible when they fail, like the picture in the link ewaller sent, tops open, or sides bulge, or corrosion, or burn visible on them..

I shouldn't open the PSU, it will void the waranty.

if you're familiar with troubleshooting at the hardware level and have another acceptable or known good PSU to try, then try it, the caps in the picture appear to be mobo caps but PSU's will have caps too.

When `sudo mkinitcpio -p linux`, I get these warnings.

those packages are available in the AUR, installing them should remove the warnings.

Could a non working ground connection in the wall socket have anything to do with crashes like these?

The pc should have a solid ground connection (its the most important wire in any circuit), basic desktops and home built pc's are not really designed to run ungrounded, they may run, but probably going to have some problems sooner or later.

Last edited by WFV (2017-08-19 22:00:22)

ewaller · 2017-08-19 22:24:04

WFV wrote:

The pc should have a solid ground connection (its the most important wire in any circuit), basic desktops and home built pc's are not really designed to run ungrounded, they may run, but probably going to have some problems sooner or later.

Can you explain that a bit more? My impression was that the Earth ground purposes were related to safety and to electromagnetic compatibility (electromagnetic susceptibility, ESD and electromagnetic emissions)

WFV · 2017-08-20 00:39:36

@ ewaller
The ground traces in the computer are mostly commoned to the PSU A/C ground, this includes the metal frame of chasis - this ground is intended to connect to the service ground (earth ground) house or building ground/grid etc. When the A/C is ungrounded the chassis metal frame, ground traces, and any heat sink connected to the earth ground traces, becomes a "floating ground". The neutral path of the electronics in the pc is segregated from the earth path - floating grounds unless specifically built to be "ungrounded" are insufficient to provide the necessary discharge for grounded circuits, they are sufficient to maintain a potential difference between positive and negative circuitry but inevitably will discharge back into components and cause problems - electricity follows path of least resistance. Usually intentional ungrounded systems will have very large heat-sinks, not for the purpose of dissipating heat generating components, but for the purpose of dissipating heat caused by converting the ungrounded components discharge to heat (e.g it has to go somewhere if not earth). Often times in older homes or buildings, the receptacles are ungrounded so people use 3x2 prong adapters to connect the ground to the neutral wire - and though this is acceptable electrical practice (in USA NEC) or install 3 prong receptacle with the ground post connected to the neutral at the receptacle (aka "California ground") it does not adequately prevent unintended consequences of discharge of electronics similar to ungrounded. In older homes with ungrounded receptacles, I've run a ground wire to them from the ground buss of the distribution panel and installed a bona fide grounded receptacle just for the computer. I don't recommend this as a do-it-yourself job if you are not an electrician (or EE etc), I am since 1987 although primarily heavy industry.

Edit: "A/C ground" = lay term for the ground (normally non-current carrying conductor) in any 3-wire A/C cable, e.g. "A/C hot", "A/C neutral", "A/C ground"

Last edited by WFV (2017-08-20 17:15:55)

ewaller · 2017-08-20 02:32:43

First, I must apologize. I may have been baiting you a little. Okay, a lot. I am electrical engineer who specializes in embedded systems and have been practicing since 1984. The systems I have designed have met the emissions requirements of FCC Title 47, Part 15; MIL-STD-461; and EU. As to safety, UL, CE and C-Tick

The ground traces in the computer are mostly commoned to the PSU A/C ground, this includes the metal frame of chasis - this ground is intended to connect to the service ground (earth ground) house or building ground/grid etc

Signal grounds in computers are most definitely *not* tied to "A/C Ground", whatever that is. At least not low impedance paths at DC. True, for safety purposes, metal chassis of computers are tied to the "Earth" ground that is provided to many systems. That Earth ground is carried back to the service panel and is not designed to carry current under normal, non-fault conditions. It does provide a nice benefit in that it prevents non-intentional radiators in the computer from acting as radio transmitters beyond the metal grounded box.

The "Neutral" signals at the power supply input are horrid and are never, ever tied to signal ground. Neutral is tied to Earth ground at the service panel, but may have 10s of volts on it depending on the current drawn by other devices on the same breaker, or by other devices on another breaker on the opposite phase that happen to share the same neutral.

Extreme care is taken in modern computers to isolate signal ground from either Earth ground or neutral. There are often "AC" paths between these paths, but they always include capacitive blocking, often include ferrite filters, and generally include pi-filters. Heatsinks do pose a problem. They often conduct to chassis both thermally and electrically. If signal ground is present on a part to be cooled by a heat sink, a thermal path must be provided, but the electrical path must be blocked using Mica or a silicon thermal pad.

Modern power supplies produce very low DC voltages at 1.25, 1.5, 3.3 5, and 12V referenced to a "signal ground" that is completely isolated from the common and line inputs. The computers work just fine when operating from this "floating" reference. At issue, as you have said, are ESD and safety. In addition, electromagnetic emissions. To minimize those hazards and effects are why we provide some bridging from signal ground to Earth. Systems will generally function without an Earth reference. Case in point, My 1950's home is missing grounds at most outputs. The this system operates just fine. I have measured the emissions from this system in an anechoic with and without grounds and see about a 10dB delta. Without ground, I might not meet the FCC Title 47 part 15 emissions requirements. Regardless, this laptop has two wires connected to the power supply, "Ground" and 20VDC. Neither of these signal have a DC path to Neutral, and ground has about a 20k path to Earth ground.

Most systems are protected by ESD by using a Faraday cage. The laptop certainly does not care about having a ground. Desktops do not extend Earthed metal enclosures around things users touch such as keyboards or mice.

As to cooling, I am not sure what you are saying. Most certainly, the Earth ground on computers are not used as a thermal conduction path. For cooling, you get your choice between radiation, convection, or conduction. Heat sinks use conduction at first to flow heat from the cases of parts, but then depend upon forced air convection (fans) or being tied to the chassis as a heat (cold) reservoir. As you point out, one must provide a thermal path, but not a DC electrical path.

tl;dr. The problem with the OP's system is probably not a missing Earth ground. Most systems don't care. An exception is ESD or susceptibility in the presence of local radiators like radios, or non-intentional radiators.

vuudochile · 2017-08-20 07:54:57

I am not an engineer by profession, not an electrical one, but as a mechanic I have learned some things. Reading the speculations before the elaborate explanation I was ready to jump in. All pc's I've ever seen in "decades" have colorful dc outputs coming out of the supply. This is where AC makes its nearest voyage on the pc. Replace this supply with above mentioned voltages with a battery and you have a dc only system (like a laptop on your lap without a ground).
But there may be static loads building up in large (by comparison) metal surfaces and they may become significant enough that if they escape and go through the circuit, as they are not regulated, they may burn something up. This is why you would want to keep electronics isolated from the cage and power supply. The power supply itself if it has any "leaks" may give you a bit of a shock if not grounded but your dry fingers are not usually its preferred route. Big electrical motors with brushes are notorious from scaring you if they are not grounded.

Back to the problem of the OP

Do you have a screensaver or does your system go to sleep get suspended? I was expecting an answer from someone asking you if it happens under load or when it is left idle. What are the power settings you have?
What kernel are you running? Sometimes what we call a freeze is lack of the system to react to our inputs, usually keyboard and mouse. I have seen it happen (recently) where the system is just fine it has just stopped recognizing input devices. I pulled the two usb plugs off and plugged them back and it was fine the rest of the day/session.
There is an xorg bug about this and it is still open and clueless. But it seems as it appears in very limited number of people and with great variation in hw. My fix of wearing out the plugs and ports is not universal, the rest use the power button.

Kabbone · 2017-08-20 12:08:10

Ploppz wrote:

@alex.theto:
How would lts kernel help? (I don't really know what it is)
How can I check for bad capacitors? Just a check on the surface? I shouldn't open the PSU, it will void the waranty.
I will try memtest.
@Kabbone
I haven't upated the BIOS. Seems like there is one update since I built the PC (the latest https://www.msi.com/Motherboard/support … RTAR.html) (Edit: actually it currently runs version 1.1 so there are quite a few revisions available). I might try to update the BIOS. If I understand correctly, all I have to do is copy the downloaded files to an MBR/FAT32 mem stick, boot to the BIOS and start the flashing.

Edit: When `sudo mkinitcpio -p linux`, I get these warnings. Could they be related?
==> WARNING: Possibly missing firmware for module: wd719x
==> WARNING: Possibly missing firmware for module: aic94xx
modinfo: https://bpaste.net/show/752b9d06f377

Yes, that's a way to flash the bios, but if the pc doesn't run stable there is a kind of risk the flash can fail. In my case, it only happens at idle with really none input, so I considered it kind of safe.
I also have an MSI Board (Gaming Plus) and even with the newest BIOS I got a freeze today, but I would still recommend to update the BIOS, especially with Ryzen, there were a lot of improvements.
I also didn't try the LTS kernel. It's worth a try, but I don't think it will fix the problem.

Perhaps I will try a constant CPU core voltage, so I can eliminate the possibility the stock undervolting in complete Idle causes this.

The warnings are normal, nothing to fear there.

Ploppz · 2017-08-20 12:58:34

Regarding gorund, the wall socket should have ground, I just haven't physically tested it.

I tried to update the BIOS, but the result is that it can't boot UEFI anymore... I will try to flash the old version again.

vuudochile wrote:

Back to the problem of the OP
Do you have a screensaver or does your system go to sleep get suspended? I was expecting an answer from someone asking you if it happens under load or when it is left idle. What are the power settings you have?
What kernel are you running? Sometimes what we call a freeze is lack of the system to react to our inputs, usually keyboard and mouse. I have seen it happen (recently) where the system is just fine it has just stopped recognizing input devices. I pulled the two usb plugs off and plugged them back and it was fine the rest of the day/session.
There is an xorg bug about this and it is still open and clueless. But it seems as it appears in very limited number of people and with great variation in hw. My fix of wearing out the plugs and ports is not universal, the rest use the power button.

No screensaver, no sleep (I have not set up any sleep so I assume there is none). Not sure about power options. I just recently installed Arch Linux and haven't customized any such options. I don't think it's only keyboard and mouse, since e.g. the clock on screen stops, and once the screen went black.

Last edited by Ploppz (2017-08-20 12:59:23)

dockland · 2017-08-20 14:02:16

Are your memsticks overklocked? If so, try lower the MHz to default, and disable XMP if you are using/enabled that.

Ploppz · 2017-08-20 16:49:28

Mem sticks are not overclocked. They are 2400 MHz by default. I heard that this should work fine with Ryzen.

Edit: Regarding the problem that it wouldn't boot UEFI after updating and downgrading the BIOS: I found what was wrong, and my system now boots again! And unfortunately, the problem is too obscure to really learn anything from it. For some reason, the ESP of the SSD was formatted with filesystem fat16. I formatted it with fat32 instead and now it works fine.

Now, I will wait until the next crash, and then I will update the BIOS to the newest version.

Last edited by Ploppz (2017-08-21 19:17:19)

Kabbone · 2017-08-22 18:46:29

I can tell you in my case newest BIOS and a constant core voltage didn't solve the problem. I'm testing the LTS kernel right now, but it wasn't much in idle yet.

drcouzelis · 2017-08-22 20:10:01

This is a known issue for Ryzen users (including me).

You can read about the issue here: (and also the unrelated "seg fault" issue) https://community.amd.com/thread/215773

The only known workaround is to go into BIOS, find the settings for "C6 states" and set it to "Disable". This will prevent any of the cores from going into super-sleep mode and locking up.

My computer used to reboot every 2 to 6 days. I'm currently on day 2 after disable C6 states. We'll see how it goes.

Kabbone · 2017-08-22 20:31:54

I know about the GCC seg faults, but didn't ran into it yet.
Thanks for the information about C6! Do you know how much C6 influences the power consumption? If I remember right, I saw something about 20W.

drcouzelis · 2017-10-13 18:05:00

FYI! https://bbs.archlinux.org/viewtopic.php … 0#p1742570

Arch Linux

#1 2017-08-18 23:08:57

PC crashes one to several times a day

#2 2017-08-18 23:17:18

Re: PC crashes one to several times a day

#3 2017-08-18 23:33:45

Re: PC crashes one to several times a day

#4 2017-08-18 23:41:04

Re: PC crashes one to several times a day

#5 2017-08-19 08:14:15

Re: PC crashes one to several times a day

#6 2017-08-19 08:27:46

Re: PC crashes one to several times a day

#7 2017-08-19 18:00:44

Re: PC crashes one to several times a day

#8 2017-08-19 19:39:32

Re: PC crashes one to several times a day

#9 2017-08-19 20:49:14

Re: PC crashes one to several times a day

#10 2017-08-19 21:59:15

Re: PC crashes one to several times a day

#11 2017-08-19 22:24:04

Re: PC crashes one to several times a day

#12 2017-08-20 00:39:36

Re: PC crashes one to several times a day

#13 2017-08-20 02:32:43

Re: PC crashes one to several times a day

#14 2017-08-20 07:54:57

Re: PC crashes one to several times a day

#15 2017-08-20 12:08:10

Re: PC crashes one to several times a day

#16 2017-08-20 12:58:34

Re: PC crashes one to several times a day

#17 2017-08-20 14:02:16

Re: PC crashes one to several times a day

#18 2017-08-20 16:49:28

Re: PC crashes one to several times a day

#19 2017-08-22 18:46:29

Re: PC crashes one to several times a day

#20 2017-08-22 20:10:01

Re: PC crashes one to several times a day

#21 2017-08-22 20:31:54

Re: PC crashes one to several times a day

#22 2017-10-13 18:05:00

Re: PC crashes one to several times a day

Board footer