Ethminer: CPU at 100%

[vampyren]

Hi folkz,

This my first time i use this miner, before i used claymore but i thought i could try this and see how it works.
Sadly when i run this my CPU goes to 100%. Another thread says to change farm-recheck to 5-10 but it did not help at all. Sadly i have to continue using claymore for this reason, thought i should mention it. I run this for a very short time but i did not see my total hashpower. Also the color on claymore is helpful.
Maybe use that for improvement work in the future ;)

[chfast]

What operating system do you use?

[chfast]

And what GPU?

[smurfy]

Also what version do you use? master or another one?

[vampyren]

Hi,
I'm on Windows 10 x64, CPU is a i5 4Ghz.
I just downloaded the latest version today.

[MariusVanDerWijden]

What Graphics Cards (GPU) do you use?

[mgedzius]

I'm getting a similar issue now. When I notice that I'm submitting an abnormal amount of stale shares I try killing ethminer. Instead of ethminer closing my cpu shoots too 100% and the process becomes defunct and not even a sigkill will stop it. I try shutdown -r and the system just hangs. I have to hard reset the system. I also ran nvidia-smi when this happened and it gave me an error something about not detecting the cards and a system reboot was needed. From what I could gather it seems it might be that my cards are overheating and shutting down. I'm trying to recreate the issue now. I will post back if I learn anything new.

Ubuntu 16.04.1
Nvidia GTX 1080 - 384.69

[mgedzius]

Just lost my first card...
It also looks like the CPU is spiking before I even try to kill the process.

[vampyren]

GPU: nVidia 1070
Driver: 388.31

[derubm]

set all of your cards to P0 mode before starting ethminer. this gives the cards a actual boost of 200 mhz,
crashes and os hanging are produced by oc´d gfx ram on p2 state. if your card switches to P0 state it adds extra 200 mhz to actual speed, which causes crashes , os hanging, bad things. Note: you need t set gfx card ram to 200 mhz less then now, as the p0 state gives your card a 200 mhz push from start on.

example:
card oc´d to +710 memory on p2 state ( standard settings ) == cardo´c to 910 + memory on p0 state.
mostly users are oc their gfx card ram to the maximum possible as it gets you more mh/s.

lets say your tested max memory speed is +910 with standard settings on P2 state.
you run your rig and all is running fine. out of nothing nvidia card switches to P0 state>>
you gfx card will go up to 1110 + memory- which causes the crash/ system hang.
don´t ask me why nvidia cards do this, but they DO.

solution: set card to P0 state, run with +710 ( == 910 on p0 state forced by card itself) in your case just subtract 200 mhz from your max memory speed you tested allready.

Now the card cannot set a additional 200 mhz and your rig will run stable.
Also make sure you set power limit @ 64-66 % ( you will not lose a lot of hashing speed, but cards will run very cool.

for windows you can use nvidia inspector, on linux you need to find the settings for nvidia-smi
if you expirience lag with the log running, reduce output with "-v 1"

[Jacxz]

@mgedzius the output in your screenshot looks a lot like what I see before ethminer crashes with a "broken pipe" which is reported in #171 during this before the crash I've not looked at CPU intensity, I'll do that next time if I catch it live...

[fekoh]

CPU also at 100% -- If more info is needed, please let me know:

./ethminer --version

ethminer version 0.12.0
Build: Linux/g++/Release

$ lsb_release -a

No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 17.10
Release: 17.10
Codename: artful

$ top

Tasks: 206 gesamt, 1 laufend, 205 schlafend, 0 gestoppt, 0 Zombie
%CPU(s): 22,1 be, 4,1 sy, 0,0 ni, 73,9 un, 0,0 wa, 0,0 hi, 0,0 si, 0,0 st
KiB Spch : 8125492 gesamt, 6564420 frei, 817556 belegt, 743516 Puff/Cache
KiB Swap: 8343548 gesamt, 8343548 frei, 0 belegt. 6996264 verfü Spch

PID USER PR NI VIRT RES SHR S %CPU %MEM ZEIT+ BEFEHL
3893 rig1 20 0 15,134g 149464 99064 S 100,0 1,8 374:54.92 ethminer :zap::zap:

$ nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.90 Driver Version: 384.90 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce GTX 1070 Off | 00000000:01:00.0 On | N/A |
| 44% 66C P2 136W / 230W | 2457MiB / 8110MiB | 100% Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 1624 G /usr/lib/xorg/Xorg 21MiB |
| 0 3893 C ./ethminer 2379MiB |
| 0 4025 G /usr/lib/xorg/Xorg 52MiB |
+-----------------------------------------------------------------------------+

Running stratum proxy (eth_proxy)
$ ./ethminer --farm-recheck 200 -G -F http://127.0.0.1:8080

[bmatthewshea]

@felixkohtz / others -
I had same problem - binary using 100% cpu - though, it was (GPU) mining fine.
Ubuntu 16.04.3 LTS

I happen to notice my start file had -G because I had been running amd card with:
ethminer -G -SC 1 -SP 1 -S 127.0.0.1:8008

But i had swapped cards out with nvidia and forgot to change my start file.
was running opencl with nvidia card..
I switched to -U/cuda and that resolved the cpu load issue.

Not sure this isn't a bug though, with opencl+nvidia / linux (?)

[ddobreff]

Its still unresolved regarding OpenCL.

[smurfy]

This happens on CUDA too sometimes. (Occurred on Win10) I noticed a problem of 30-100% ethminer cpu load.
IT happens after a while sometimes.
It may be related to other cpu load. i noticed that if i do something cpu intensive the likely hood of ethminer buggs and take higher cpu load is higher.

[bmatthewshea]

@smurfy Hmm never noticed it consuming at all under Windows/cuda, but as far as MS machines here, I'm still old school on Win7 lol. I will keep an eye out, though..

[fuelman31410]

I was able to resolve this by changing my page file size, I set min to 20480 max to 51200 . I am running 6 gtx 1070's and was getting terrible output before like 68mh now up to 158mh. This is my current batch file

setx GPU_FORCE_64BIT_PTR 0
setx GPU_MAX_HEAP_SIZE 100
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_SINGLE_ALLOC_PERCENT 100
ethminer.exe --farm-recheck 2000 --cuda -S us1.ethermine.org:4444 -FS us2.ethermine.org:4444 -O mywalletaddress..minername --dag-load-mode sequential

Hope it helps

[chfast]

@fuelman31410 can you confirm this is what happened:

1. You had ethminer performing bad with 100% CPU usage.
2. You increased page file size (in Windows?) and now it works better?

[Quozzo]

I was having the same problem with ethminer.exe using two cores and keeping them constantly at 100%. Having a 7900X overclocked to 4.8GHz I didn't care all too much but was curious if that was normal.

Using what fuelman posted and changing the -G to --cuda and removing the last variable means it now uses less than 1% of my total CPU power. On top of that. My system power draw was reduced from 500W to less than 450W which is more than a 10% reduction. Unfortunately the hash rate has also been reduced on my 1080Ti from 37.5MH/s to around 35MH/s.

Not sure what the difference is between -G and --cuda is but it is definitely doing something different. Just not sure if the ~7% drop in hash rate is worth it.

[chfast]

-G uses OpenCL, -U uses CUDA. @Quozzo do you use Windows?

[fuelman31410]

Yes, I am running my miners on Windows 10

[Quozzo]

@chfast Yes I use Windows 10

[chfast]

So this issue is only for Windows and Nvidia GPUs when using OpenCL? (@fuelman31410 did you mentioned having 100% usage also when running in CUDA mode?)

[chfast]

I've found this similar symptoms: https://github.com/pandegroup/openmm/issues/1541.

[Quozzo]

@chfast Yeah it looks similar. I should mention I have two 1080Ti's which is probably why I've got two cores pegged at 100% instead of a single core.

[uvoz]

Same here on Ubuntu 17.10. 100% CPU usage. Replaced the -G flag with - -cuda. Processor is near idle now.

[fossum]

I'm running Ubuntu 17 and was using the --opencl option. I had 300% CPU even after a process restart. I switched to --cuda and all is better now.

[Quozzo]

@fossum Was that using version 0.13.0 because that was fixed for me (on windows).

[DeadManWalkingTO]

Please check #695.

[fossum]

Yes, I think it was with 13. I use the latest stable release.