Need pointers for mining ethereum on xubuntu 16.04 with

[Rickirs]

Hi,

I have couple of issues related to ethereum mining with Nvidia 1060 6B and I need proper pointers here to solve them.. (already have been through lots of reading online from last couple of days).
Currently, I can get 23.2 Mh with OC Memory clock = +1250 and power 80W , increasing GPU clock has no effect.
Using Nvidia 387.x latest drivers. (had same result with nvidia 384.x drivers)

1. First and foremost, I can't make headless setup, I am not new to linux but never worked on GUI remotely. If I start machine without monitor, I can't overclock using nvidia-settings. I can do from CLI but it needs to have monitor connected at first.
Hence I need to know which is best possible solution, I read there is this thing like fake display which can be added to /etc/X11/xorg.conf , then some suggest adding dummy drivers. (ofcourse I don't want dummy plugs).
So what is practical and easy way to achieve headless setup where X (gui ) is started so that I can do overlocking settings after it. Need guide and advice here.

2. I cant find way to save overclock settings, currently I am using things like and running through rc.local:

Code:

nvidia-settings -a [gpu:1]/GpuPowerMizerMode=1
nvidia-settings -a [gpu:1]/GPUMemoryTransferRateOffset[2]=1250

Now I can't find a proper link or read where I can understand exactly what GpuPowerMizerMode=1 do, I understand that it set performance level but I need little more detailed info to understand. But can't find one. I can set all those settings from Nvidia Gui (nvidia-settings) but it get lost on restart, hence using cmd.

3. This is extension of first issue, if I get headless setup working with GUI , how can I access it remotely as if doing locally without restriction because if I use xrdp etc then it creates completely new session where overlocking is not available (as that remote gui is not through GPU) in Nvidia X server (nvidia-settings). So any input / advice is much appreciated on it.  What would be most practical approach  to work around this issue ?

4. Its about overclocking, while reading around, I see people are saying they were able to achieve ~24 mh on Nvidia 1060 6GB with 65W power limit, but for me it resulted in 17 MH only .. what I am missing here  as in linux I can only touch GPU clock, Memory transfer and Powerlimit.

Btw, I am using Claymore 10 for mining.

[MATHReX]

If you don't want any headaches setting up Linux for mining environment, I will suggest you give nvOC a try. It's a prebuilt Linux distro with everything ready for mining.
Link: https://bitcointalk.org/index.php?topic=1854250.0

1. I'll suggest getting a dummy plug because it's good for the long run and it'll cost 8 USD and it'll save you a ton of time configuring for a headless mode.
2. GpuPowerMizerMode=2 for better hashrate.
3. Use OpenSSH.
4. Core Overclock: -300 and Mem Overclock: +1600 for 24MH/s. (GTX 1060 Samsung Mem)
    Core Overclock: -300 and Mem Overclock: +1000 for 22.5MH/s. (GTX 1060 Micron Mem)
    Core Overclock: -300 and Mem Overclock: +200 for 17MH/s. (GTX 1060 Hynix Mem)

[Rickirs]

Thanks mate for for suggesting nvOC, had given it good consideration but then dropped the idea preferably because I had already done things I need. Only stuck in few tit-bits and hence asked help here.

I know that Xorg is quite versatile and you can do headless setup, just wanted good pointer. Have been already managing it through SSH and am more comfortable with it.

Alright already using GpuPowerMizerMode=2 as per nvidia-smi , saying P2 under GTX

For OC, I am not sure who to know make of VRAM in Linux , as I can get around ~23.6 , I suppose its Samsung. What about PowerLimit , currently I am on 80W ?


 Btw, little Off-Topic, any advice which dual mining I should consider with ETH at the moment. Sia looks useless, however, in trial was able to get ~22.1 Mh with ETH along with SIA ~300 Mh/s with dcri of 40

[MATHReX]

Thanks mate for for suggesting nvOC, had given it good consideration but then dropped the idea preferably because I had already done things I need. Only stuck in few tit-bits and hence asked help here.

I know that Xorg is quite versatile and you can do headless setup, just wanted good pointer. Have been already managing it through SSH and am more comfortable with it.

Alright already using GpuPowerMizerMode=2 as per nvidia-smi , saying P2 under GTX

For OC, I am not sure who to know make of VRAM in Linux , as I can get around ~23.6 , I suppose its Samsung. What about PowerLimit , currently I am on 80W ?


 Btw, little Off-Topic, any advice which dual mining I should consider with ETH at the moment. Sia looks useless, however, in trial was able to get ~22.1 Mh with ETH along with SIA ~300 Mh/s with dcri of 40

For VRAM, just set the above clocks, if they are stable at the given clock speeds, then it has the corresponding VRAM vendor. If you have Samsung, just push it more to get 24.5 MH/s which is possible.
Set power limit to 65W. GPUs should work perfectly fine at 65W for mining ETH.

I wouldn't suggest dual mining because it will draw much more power than single mining, you'll likely face stability issues. Moreover, Single mining at 24.5 MH/s with 65W will be a bit more profitable than mining ETH at 22 MH/s and SIA at 300 MH/s at 80W. Just place these values in whattomine and you'll understand.

[Rickirs]

For me, it is kind of confusing, I have set offset 1250 , I can get around ~23.6 Mh/s however, have to set PL to 80W, anything less than that reduces mining rate. Further, if I try to push memoryclock further, lets say, 1400, then miner starts showing error and soon system freezes.

On 65w, I get 17 Mh/s only.

Seems I can't push more than 23.6 Mh/s , no matter what settings.

[MATHReX]

For me, it is kind of confusing, I have set offset 1250 , I can get around ~23.6 Mh/s however, have to set PL to 80W, anything less than that reduces mining rate. Further, if I try to push memoryclock further, lets say, 1400, then miner starts showing error and soon system freezes.

On 65w, I get 17 Mh/s only.

Seems I can't push more than 23.6 Mh/s , no matter what settings.

Oh! It's okay, every card is different and the highest hashrate you can get depends solely on ASIC quality of the GPU.
The difference between the numbers can be because I got my numbers using nvOC distro.
I still recommend you to give nvOC a try and see if you can get more from your system.

[Rickirs]

Well, I guess I have to live with ~23Mh/s for now. However, I will definitely give try to nvOC, will set it up on a new disc to try out on same rig. Just wondering why he is using 15gb image while same can be done on less than 1 GB image (all the customization he has done in image)

However, I have seen trend that sometimes in night I see system freezing, can't connect remotely however, If found that Xorg was using 100% cpu while  GPUs were  all gone from nvidia-smi  . Only hard restart fixed the issue , any idea what could be the issue ?

[MATHReX]

Well, I guess I have to live with ~23Mh/s for now. However, I will definitely give try to nvOC, will set it up on a new disc to try out on same rig. Just wondering why he is using 15gb image while same can be done on less than 1 GB image (all the customization he has done in image)

However, I have seen trend that sometimes in night I see system freezing, can't connect remotely however, If found that Xorg was using 100% cpu while  GPUs were  all gone from nvidia-smi  . Only hard restart fixed the issue , any idea what could be the issue ?

Yeah, I know 15GB image file for Linux Distro with every software related to mining. It is kind of a large file but it works pretty well. I'll recommend to flash it in 32 GB USB drive for better stability.

The problem with Xorg might be related with ./Xmodmap
Here are a few links for you to check them out and see if this is the culprit behind the freeze:
https://askubuntu.com/questions/914148/xorg-high-100-cpu-on-xubuntu-17-04
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/998310