Crazy temperature with nstress-ng?

[yohplala]

Hi there,
My setup is as follow:

• a MSI Vector 16 HX AI laptop,
• Ubuntu 25.10.
• I tweaked BIOS to undervolt and reduce strongly TDP (see this tuto), with these values:
  • TDP (Advanced > Power & Performance > CPU PMC > Config TDP Configurations):
    • Power Limit 1 to 45,000 and Power Limit 2 to 80,000 and Time Window to 40 seconds,
  • Turbo Clock (Advanced > Power & Performance > CPU PMC > View/Configure Turbo Options):
    • Energy Efficient Turbo to Enabled
    • P-core Ratio0 and Ratio1 set to 50,
    • the remaining P-cores to 46,
    • and all E-cores to 34.
  • Undervolting (Advanced > OverClocking Performance > Processor), voltage setting for individual cores (VF Configuration Scope > Per Core):
    • for all P-cores, set the Voltage Offset to -60
    • for all E-Cores to -50.

I installed msi-ec drivers & MControlCenter from Github repo (so last version in main).
I run a stress test with nstress-ng.

pierre@pierre-A2XWJG:~$ stress-ng --cpu $(nproc) --timeout 30m --metrics
stress-ng: info:  [8446] setting to a 30 mins run per stressor
stress-ng: info:  [8446] dispatching hogs: 24 cpu
stress-ng: info:  [8446] note: 24 cpus have scaling governors set to powersave and this may impact performance; setting /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor to 'performance' may improve performance
stress-ng: info:  [8446] stopping 24 stressors
stress-ng: metrc: [8446] stressor       bogo ops real time  usr time  sys time   bogo ops/s     bogo ops/s CPU used per       RSS Max
stress-ng: metrc: [8446]                           (secs)    (secs)    (secs)   (real time) (usr+sys time) instance (%)          (KB)
stress-ng: metrc: [8446] cpu            48228828   1606.11  38056.67     33.68     30028.30        1266.17        98.82          8688
stress-ng: info:  [8446] skipped: 0
stress-ng: info:  [8446] passed: 24: cpu (24)
stress-ng: info:  [8446] failed: 0
stress-ng: info:  [8446] metrics untrustworthy: 0
stress-ng: info:  [8446] successful run completed in 26 mins, 46.13 secs

I stopped manually this test because at approximately 24mn, fans started to run at ?max? speed, and temperature displayed in MControlCenter for CPU has been 255°C

I did not panic, because at same time, I had sudo turbostat --interval 1 running, telling me a different picture, with a temperature at max 50°C.

After stopping nstress-ng, situation returned to normal in MControlCenter without me doing anything.

Similar in turbostat

I am then thinking there might be a bug lying in MControlCenter, when tested for a while with nstress-ng, but I don't have much more clue?

Another thing, I don't think it is related. I think it has been reported already.
There can be conflict between Ubuntu energy management system and MControlCenter. I don't know if this can be improved, but both can compete together regarding how managing loading/disloading battery. Maybe a message should indicate in MControlCenter not to use Ubuntu's (Gnome?) 'preserve battery life' button?

Thanks again for this great tool you are making available to common mortals. It is definitely a great utility!

[mutchiko]

@yohplala you talk about multiple subjects in one issue, lets start with the wrong temp value:

reproduce the issue again (doesn't have to be stress-ng), and when MCC reports 255, do an EC dump, you can use the ec_sys method if you dont want to unload msi-ec, then share the dump here.

fans started to run at ?max? speed

it seems like you triggered emergency mode, but you mention that the other CPU sensor reported 50C, so there's something going wrong

[yohplala]

@mutchiko , well I tried to reproduce, running stress-ng (I don't know other method to create this), but this time it ran for 30mn without trigerring this, without a crazy 255°C temperature.

you mention that the other CPU sensor reported 50C

I am no hardware expert, but I would think these are the same sensors MControlCenter / msi-ec is using?
I had turbostat running along MControlCenter.

turbostat (the screenshot I shared) has a CoreTmp column, showing temperature for each core, you can see temperature is between 48 and 50, while the column Busy% show you the load per core: all above 99%.

At the same time, MControlCenter was showing a CPU temperature of 255°C, and had the fan running at ?max? speed. (>4500rpm)

I made a EC dump, but as said, I have not been able to recreate the 255°C temperature in MControlCenter.
So if it does not help, we can close, and if one day I meet this trouble again, I will know that I should do an EC dump.

sudo hexdump -C /sys/kernel/debug/ec/ec0/io

00000000  00 80 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
00000010  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
00000020  00 00 00 00 00 00 00 00  0a 05 00 00 00 24 39 5b  |.............$9[|
00000030  07 01 00 0d 00 00 50 81  2b 16 28 3c 37 02 c0 00  |......P.+.(<7...|
00000040  2b 16 3d 00 f6 15 00 00  30 0d bb 3d b4 0b c0 44  |+.=.....0..=...D|
00000050  00 00 00 00 00 02 00 00  00 00 00 00 14 00 14 00  |................|
00000060  80 00 00 00 1f 00 2b 43  32 00 37 3e 45 4b 51 57  |......+C2.7>EKQW|
00000070  69 19 0a 19 28 37 41 55  69 00 28 31 38 3e 44 4a  |i...(7AUi.(18>DJ|
00000080  2c 00 32 37 3c 41 46 4b  61 00 00 14 28 37 46 5a  |,.27<AFKa...(7FZ|
00000090  73 00 2b 30 35 3a 3f 44  02 00 00 02 00 00 41 00  |s.+05:?D......A.|
000000a0  31 35 4d 33 45 4d 53 31  2e 31 31 33 31 31 32 36  |15M3EMS1.1131126|
000000b0  32 30 32 35 31 34 3a 35  39 3a 32 38 00 00 00 08  |202514:59:28....|
000000c0  00 00 01 25 00 00 11 00  01 00 00 00 00 00 00 00  |...%............|
000000d0  00 00 c1 80 8d 00 05 bc  00 09 00 00 00 08 00 00  |................|
000000e0  e2 00 00 f6 15 00 00 c1  00 00 00 00 00 83 00 00  |................|
000000f0  00 00 70 00 28 64 00 2a  64 00 00 00 00 01 00 00  |..p.(d.*d.......|
00000100

[mutchiko]

@yohplala must've been the wind. if it ever happens again, don't hesitate to re-open the issue.

And if you have unanswered questions, you can ask in discussions.

[yohplala]

Hello, it did it again!

dump:

00000000  00 80 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
00000010  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
00000020  00 00 00 00 00 00 00 00  0a 05 00 00 00 24 3b 5b  |.............$;[|
00000030  03 01 00 0d 01 00 50 81  2b 16 28 3c 37 02 c0 00  |......P.+.(<7...|
00000040  2b 16 33 00 73 15 00 00  c2 0a 5d 3c be 0b c0 44  |+.3.s.....]<...D|
00000050  00 00 00 00 00 02 00 00  00 00 00 00 58 5a 61 60  |............XZa`|
00000060  80 00 00 00 a3 00 e9 13  ff 00 37 3e 45 4b 51 57  |..........7>EKQW|
00000070  69 28 0a 19 28 37 41 55  69 00 28 31 38 3e 44 4a  |i(..(7AUi.(18>DJ|
00000080  34 00 32 37 3c 41 46 4b  61 28 00 14 28 37 46 5a  |4.27<AFKa(..(7FZ|
00000090  73 00 2b 30 35 3a 3f 44  82 00 00 02 00 00 49 00  |s.+05:?D......I.|
000000a0  31 35 4d 33 45 4d 53 31  2e 31 31 33 31 31 32 36  |15M3EMS1.1131126|
000000b0  32 30 32 35 31 34 3a 35  39 3a 32 38 00 00 00 08  |202514:59:28....|
000000c0  00 00 01 25 00 00 15 00  00 46 00 40 00 00 00 00  |...%.....F.@....|
000000d0  00 00 c4 80 0d 00 03 bc  00 09 00 00 00 1f 00 00  |................|
000000e0  e2 00 00 73 15 00 00 c1  80 00 00 00 00 da 00 00  |...s............|
000000f0  00 00 70 00 2c 64 00 35  64 00 00 00 00 05 02 00  |..p.,d.5d.......|
00000100

I activated cooler boost after noticing this.
Context is similar as previously: important workload, I was hearing the fan, doing something else, and getting back to laptop when noticing no more fan noise.
I checked and saw CPU temp at 255°C, but fan CPU running slow.
So as to make sure it does not damage CPU, I activated Cooler Boost.

[yohplala]

Can you please re-open the issue maybe? (I don't have the required rights)

[yohplala]

Other feedback, I switched off MControlCenter and then restarted it (without restarting the laptop) and now temp is correctly displayed. (varying between 70°C and 82°C)

I would say it does seem to be a risk to me:

• appears when CPU is high / has workload
• the temperature detected at 255°C (visual artifact likely) but fan CPU going to sub 2k rpm
• all the while the CPU continues its work and continues to heat
  -> there is a risk of overheating the CPU because CPU fan is lost and not running fast enough

[Xabi08YT]

So I just read the issue. Since I will soon have the Vector 16HX AI on hand I will try to reproduce the issue. Have you tried resetting the EC ? MSI have a pdf on how to do that.

[yohplala]

Thanks for reopening this issue.

I am sorry, I don't know what implies "resetting the EC"?

I fear that level of occurence is very low, and I hope the dump helps.
It only occured to me twice, in both cases when stressing CPU significantly, but I am stressing it a lot, and over all this time since end of Feb, it only occured twice.

1st case reported end of Feb, with stress-ng to validate some undervolting CPU
2nd case now, running intensive calculations with CPU+GPU (cuda based), but I have started to run these intensive calculations over the past 3 weeks i would say.

[Xabi08YT]

There you go:
Here is the official docs.
https://www.msi.com/support/technical_details/NB_EC_RESET

It is an easy process and it may solve the bug. Also, maybe you should think about updating the BIOS if you did not do that already, as some BIOS updates are also updating the EC firmware.