Ethminer unusable with Celeron CPU

[n4ru]

So, I have 7 GPUs running on a measly Celeron with 8GB of RAM. When it's just one, I can run it just fine. When I bump it up, the text output becomes so slow that it ends up freezing and won't mine properly. CPU usage spikes up to 100%.

This issue doesn't happen with any other miner. I'm assuming it is related to the way text is output to the console and I'm wondering if there's either a flag I'm missing or if there is another fix available.

My pagefile is larger than the sum of all GPU's RAM.

[ajthemacboy]

Same issue for me. Claymore's miner and other coin miners don't have high CPU usage like this for me. I have an Athlon II x2 CPU, definitely low end, but since other miners work I think it's related to ethminer.

[jimmykl]

Try the --farm-recheck flag to reduce the speed of the text output:
--farm-recheck <n> Leave n ms between checks for changed work (default: 500). When using stratum, use a high value (i.e. 2000) to get more stable hashrate output

[jimmykl]

Also if the miner is hanging when it's starting up loading the DAGs onto the GPUs you could try changing the --dag-load-mode flag to sequential or single:

--dag-load-mode <mode> DAG generation mode.
	parallel    - load DAG on all GPUs at the same time (default)
	sequential  - load DAG on GPUs one after another. Use this when the miner crashes during DAG generation
	single <n>  - generate DAG on device n, then copy to other devices

[n4ru]

No dice to any of those. As soon as "cudaminer0" text shows up, it shits the bed regardless of settings.

[jimmykl]

@n4ru Sorry to hear that. I don't have any other suggestions at the moment sorry.

@ajthemacboy Yeah even my pentium 4560 gets pegged for a while during the setup. Hopefully someone can work on an actual fix.

[ajthemacboy]

Ok, clearly I should have tested a bit more before responding the first two times.

With --dag-load-mode single 0, my startup time is a lot lower. However, once I get up and running, it randomly and frequently goes up to 100% CPU and the screen only prints one character at a time.

Additionally, my hashrate is only 92 MH/s max with 4 GTX 1060s, using --cuda-parallel-hash 2. My hashrate mining with Claymore's miner is 94 MH/s constantly. I'm not sure why these optimizations aren't working for my cards.

[jimmykl]

--cuda-parallel-hash 4 should be optimum for 1060s. I was about to reply to you in the other thread – davilizh said:

This code is optimized for GP106 with 1GPC GDDR5 case, and may be helpful for other chips. You can scale the “#define PARALLEL_HASH 4” in dagger_shuffled.cuh from 1 to 8 to check whether can improve performance for your GPU.

(from #18) I'm not sure what 1GPC means though, maybe 3GB card?

[omenos]

What operating systems are you guys running? I'm using CentOS without any of these issues. Is this potentially Windows specific?

[n4ru]

I suppose it could be. I am on Windows 10.

[ajthemacboy]

I am on Windows Server 2016.

[justinwyer]

This only happens on windows, ethminer works okay on Linux with a Celeron. I have rigs running both.

[gigi1231]

ethminer.exe -U -v 2 -L sequential --dag-load-mode single 0 -S eu2.ethermine.org:4444 -SP 1 -O

wallet.name --farm-recheck 2000 --cuda-parallel-hash 4

tried this, no luck. still takes forever to load ...

[ghost]

Same issue here, Windows 10 x64 with 7x GPU (1060s)

[ghost]

Running multiple instances with 2 GPU per instance seems to be a valid workaround for this issue.

ethminer.exe --farm-recheck 200 -U --cuda-devices 0 1
ethminer.exe --farm-recheck 200 -U --cuda-devices 2 3
ethminer.exe --farm-recheck 200 -U --cuda-devices 4 5

But all in all I would prefer if somebody could do a fix to allow more then 2 GPU in a single instance on a Celeron CPU on Windows 10.

[BentHaase]

100% CPU usage regardless of number of GPUs. Ubuntu Linux 16.04.2.

[omenos]

To be clear, are you all running the release candidate?

[BentHaase]

In my case running ethminer 0.9.41-genoil-1.1.7 built from source. @omento

[ajthemacboy]

@omento This occurs for me on the RC and 0.10.0, as well as on Genoil's miner.

[omenos]

Just trying to consolidate as much info for @chfast as possible, could you guys provide the following info so it's globbed together? Apologies if this feels redundant.

1. Operating System (Include Version)
2. Graphics Driver Version
3. Graphics Cards in Use
4. Version of Ethminer used (all versions)
5. Built from source or pre-built binaries?
6. If source were there any errors/warning in the build process? What are they? Any issues retrieving libraries? What GCC/CMAKE version was used?

[omenos]

@ET-Bent Could you try this binary? Just built it from the latest dev source (with CUDA 8 support).

ethminer.tar.gz

Edit: Forget that. I can't tell which branch is ahead of which, master or release/0.11. Just try with the Linux binary found on the Releases page for RC1. I've been running it for 14 hours solid.

[ghost]

@omento

1. Windows 10 x64 (newest release, including Creators Update)
2. latest nvidia stable driver
3. 7x GTX 1060 (6GB)
4. the latest RC release
5. pre-built binary (RC release)

The issue behind this is that the text output becomes really really slow because of a full load of the CPU (100% CPU usage). I think the text output and mining is not seperated from each other?

Shouldn't the CUDA-feature prevent from 100% CPU usage, even on a very slow G4400 CPU?

This issue happens on all "ethminer" I found (e.g. the genoil miner or the miner of this repo).
This issue does not happen on Claymore's Dual Miner.

As I wrote above a workaround is to run multiple instances of this miner with only 2 GPU per instance.
But I really don't like this solution.

[n4ru]

I have identical settings as JHGitty except I use 3GB cards.

[ajthemacboy]

Operating System: Windows Server 2016 Standard, build 1607
Graphics Driver Version: 382.33
Graphics Cards in Use: GTX 1060 6GB
Version of Ethminer used: 0.10.0 and the pre-release with optimizations
Built from source or pre-built binaries: Pre-built binaries tested only

[BentHaase]

@omento I appreciate all your efforts to help me.

I just rebuilt all my nodes today so they use the latest ethminer version built from source (0.11.0.dev0). Now everything is fine and we are soaring around < 1% CPU usage 👍

Cheers!

For future time travellers reading this: Building the latest version from source >= 0.11.0.dev0 will fix the abnormal CPU usage. Also make sure you are not using cpp anymore ;)

[ajthemacboy]

@ET-Bent I'm glad to know it's fixable. However, I'm having trouble compiling it on Windows.

• Do I need VS 14 2015 to compile?
• How do I install Hunter?!?!? I can't find any binaries anywhere.

I really have no clue what I'm doing, any help is appreciated.

[thghdbs]

Win 10 Pro
Celeron G3900
6x1070

Confirming that rc1 will not load. After the first few lines of loading the system crashes.

[thghdbs]

@ET-Bent could you link a zip with the exe and dlls (and everything else) compiled for Windows 10 that you mentioned does not have the Celeron 100% CPU bog down issue?

[guillermo82]

If you are using the latest binary, try using the --cuda-schedule option
sync should put cpu to about 3% but is the slowest
yield cpu is at 100% but responsive system (sort of)
spin screws up the system if you have more GPU cores than CPU cores (it might be an even/uneven issue, but no way I can test that here)

[cnilsecure]

Please please share, I been looking everywhere,All my rigs are Celeron based and I cant use this optimization.

[jimmykl]

@marcovillar If you've tested it sucessfully you should make a PR for review and wider testing

[ChronoXNL]

@chfast : Latest build seemed to have solved the issues for me on 2 machines. Tested with latest build #112.

Setup: 7 x 1070, celeron 3930, 8gb mem, z270a chipset.

[ajthemacboy]

The new release works great for me! Mining at about 98 MH/s, tyvm @chfast <3

[gigi1231]

works great, thank you! used https://github.com/ethereum-mining/ethminer/releases/tag/v0.11.0

[ghost]

I can confirm - latest v0.11.0 release is fixing the issue - somebody can close this ticket.
Thank you very very much for fixing this @chfast !!!

It's working fine on my 7-GPU (7x GTX1060) machine. Tested under Windows 10 x64.

[ebbbang]

It still doesn't work for me ...

[derubm]

if it is still a text issue to you, try >> -v 1 <<
disables allmost everything excpet information and error output.
shows set work and found shares only.

[ebbbang]

Still no luck ...
CPU stayed at 100% and system restarted after a crash ...

[derubm]

ebbang could you provide a bit more information about your rig?
CPU Model, Ram, GPU model , OS, Miner Version, driver version etc ..

[ortakpayda]

I also have this problem although I have Pentium.
My rig:
Intel Pentium G4400
5x 8GB Asus GeForce GTX 1070 Strix OC
1x 8GB Asus GeForce GTX 1070 Strix
geforce 382.53
Windows 10 with the necessary updates done
miner version 0.11 and 0.12dev

I really want to stop using claymore but with ethminer I can't mine properly 😞

[derubm]

@ortakpayda
check if you can generate the DAG sequential or single, that might take load from the system.

-L, --dag-load-mode <mode> DAG generation mode. parallel - load DAG on all GPUs at the same time (default) sequential - load DAG on GPUs one after another. Use this when the miner crashes during DAG generation single <n> - generate DAG on device n, then copy to other devices
also try "-v 1" to reduce spam in the miners window

[ebbbang]

@derubm My Rig:
Intel Pentium G4400
6x 8GB Gigabyte GTX 1070 G1 Gaming
GeForce 384.94
Windows 10
miner version 0.11 and 0.12dev

I tried with -v 1 & also with single/sequential DAG generation mode. Still it does not work ...

[jimmykl]

@ebbbang + @ortakpayda Have you updated to the latest mobo BIOS and upped Windows' virtual memory? (See Increase virtual memory on [url]http://1stminingrig.com/best-windows-setup-configuration-tweaks-for-mining/[/url])

[ebbbang]

@jimmykl I'll check for the mobo BIOS today ...
And I have done all windows tweaks ...

[jimmykl]

@ebbbang It's a pain to do but you could also unplug cards individually to see if one is acting up. I've had rigs which got unstable and all it took was unplugging everything and hooking it all back up which somehow resets things and they work afterwards.

[xixou2]

Remove the -G option and only keep -U

-G,--opencl  When mining use the GPU via OpenCL.
-U,--cuda  When mining use the GPU via CUDA.

When both -G and -U are specified, the miner works but uses one CPU core.

[ajthemacboy]

@xixou2 No one here was using both.

[xixou2]

Maybe they were by mistake.

[dest4]

Using a 1070 on linux / Cuda 8, ethminer 0.13.0.dev0, i5 Quad core, I had 100% CPU usage on a single core whatever the logging options, using the OpenCL kernel (-G).
I recompiled ethminer with the following flags
cmake .. -DETHASHCUDA=ON -DETHASHCL=OFF
Now I'm using the Cuda kernel (-U) and I get minimal CPU usage.

[dougbacelar]

Tried to mine eth for the first time using this miner with the following code in a .bat script:

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 200 -G -S eu1.ethermine.org:4444 -FS us1.ethermine.org:4444 -v 1 -O <Ethereum_Address>.<RigName>

CPU overloaded and restarted my Rig twice, reached temperatures of over 100C.

I am back to zCash mining for the time being. My CPU stays at a stable 40C while mining zCash.

ethminer 0.12.0
4x GTX 1070
i7 4790k
win10

[derubm]

@dougbacelar replace -G with -U

[dougbacelar]

@derubm thanks! I dont understand why, but it worked with -U 👍 💯

1. I replaced -G with -U
2. increased the value of farm-recheck to 30000,
3. increased -v from 1 to 2 to get hash rate info

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 30000 -U -S eu1.ethermine.org:4444 -FS us1.ethermine.org:4444 -v 2 -O <Ethereum_Address>.<RigName>

[Logic-Net]

I had the same problem, mining with claymore and CPU stay in 50 percent..
the problem resolved by disabling windows search and superfetch:

1. Press the Windows logo key + R at the same time, and type in services.msc
2. Locate Windows Search option, double click it to enter its Properties window.
3. Change the Startup type to Disabled. Then click Apply and OK to save your change.

and the same for superfetch (Properties - Stop - and Startup type Disable)

I hope these settings help other people
I struggled for ten days until I solved the CPU issue 50%

If you think I was useful to you, you can donate me a beer by paypal :))

[dhanaraj1]

@derubm Thanks man !!

[Logic-Net]

and if the problem persists, you have too much overclocking on one GPU
usually the GPU which has fan on zero.
reduce overclocking to that board and everything becomes stable.

[AndreaLanfranchi]

If I may add just one more hint.
Please note that when power-limiting the GPU below their "natural" consumption, the CPU load increases a lot. (direct experience on Linux !)
This is however related to the NVIDIA driver, not ethminer fault.