Topic: [ mining os ] nvoc - page 250. (Read 418546 times)

***Fixed. Found another USB, loaded nvOC on it, got it running and cleaned up the other USB with Gparted***

I was messing with my 1bash file using nano via SSH from my mac and managed to screw it up. Next time I will make a backup version before I do.

My thoughts were the easiest fix was to just start from scratch using HDD Raw and redo the USB.

My problem is I can't seem to do anything to the USB. HDD Raw doesn't recognize it. Disk utility on my Mac sees the linux partition but won't let me delete it.

I am sure I could wipe it out in Linux but I can't do that since the Linux is running on the USB.

How does one go about rewriting a new copy of nvOC onto a USB that already has it on there?

In my experience with 1060's and single mining ETH, when I push memory overclock to the max (+1840 Samsung memory), and power limit to very low limit of 75 Watts, negative Core clock causes instability. Further testing is required but for now +100 Core with +1840 mem and 75 PL works stable for couple of days.

My Celeron's are on 8x rigs and they cause too much instability. I would not recommend them at all. Invest little bit more in better CPU like i3 (I haven't used Pentium CPU's on mining rigs but I guess those are not much better than Celerons), it will save lots of frustration and downtime and it will pay itself off. A lot of times my Celeron rigs were frozen for 6-8 hours while I was at work -> lost revenue. I don't feel like babysitting them 24/7 and they will be replaced soon

Hi tomlev5. Just to clarify, we are not talking about one computer power supply that can handle 26 GPU's, you need multiple power supplies for that. We were talking about the electrical panel where the main power line enters the house, the box where all the fuses (circuit breakers) are located. You want pic's of the rigs or the electrical panel?

To get the memory type you have to install the card in windows machine and run GPU-Z tool. To my knowledge there's no such program that can identify memory brand under Linux.

As I mentioned I had few cards with Micron memory and I was able to OC memory to +500 under windows (which equals to +1000 under Linux). I never had cards with Hynix memory but I read that's the worst brand to overclock. I hope your's are not with Hynix which could explain the low OC and the problem lies somewhere else. Have you tried setting PL to 76, core to +100 and memory to +800? See if that's stable, then increase the memory by another 50 and so on until the rig becomes unstable or crashes.

Yes, I have all 13 slots filled on the asrock boards. 8GB of DDR4 memory. 2 x Corsair HX1000i power supplies per rig (it is overkill but you need lots of PCIe and Sata connectors for 13 GPU rigs). The CPU's are i7-7700 (also overkill, but I got them on sale and I plan to utilize them for CPU mining when I find some spare time to experiment). On that note, two of my Asus Z270-A are running with Celerons G3900 and those are not very stable, one is running on Win 10 and the other on nvOC-0018 with same settings and same MH/s, no difference between windows and Linux at all, although the windows machine is running little bit cooler with lower fun speeds (I suppose that's driver issue, or maybe Maxximus' temp control is not on par with Afterburner, hehe). Although both win and Linux machines with Celeron CPU's are hashing same per card as the one's with I7's, they are occasionally being restarted by watchdogs, or they completely freeze and have to restart them every few days. I tried lowering the OC on them but it didn't help. The first chance I get I will replace the Celerons with at least i3's. Although the mining community recommends Celeron CPU's at large to cut costs I would not recommend that garbage of CPU to anyone, not even for web browsing...

First of all mucho thanks to Fullzero and everyone else who go this to where it is today. Amazing software.

I built my first mining rig today (Biostar 250 BTC & 6 GTX 1070 cards)

I downloaded V19, HDD Raw, made the two recommended changes in the BIOS, Changed the NiceHash address to mine and selected NiceHash in the 1bash and loaded the image on a 32 GB USB and booted up and voila it worked. The first try it only saw 5 cards and one of my risers was loose. Fixed that and all 6 are running. It is as simple as falling off a log.

I am not even sure this belongs in this thread since it is such a general question but here goes. I chose NiceHash simply because that is what my son has been using on his windows gaming PC and it seems to work well and pays directly to a BTC wallet (I am using block chain).

I have the parts here for a second identical rig I am going to put together tomorrow and may order a couple more. My question is should I be using the nice hash or are there much better options built into the nvOC? Salfter profit switching? I am sure many people prefer different coins for various reasons but I am really too busy too keep up with the daily news on each one and the idea of something automated seems great to me. I am sure on  any given day the argument could be made to choose one over another but is there a best choice for one who just wants to set it and forget it?

Right now it shows around 163 MH/sec. Is that what should be expected or do I need to tweak things?

Thanks Again.

wow, this power supply sounds really great. Can you share some photos of your setup? Where can you buy these panels and breakers?

Thanks for the feedback - I would love to discuss more as, while I cannot seem to get more than 21 or so per 1060 (while I am not sure what the memory type is yet, I feel like I should be able to get more than +600-700 in Linux for any brand).

The reason for the panel upgrade is that I am out of breakers entirely, and my panel doesn't support half-breakers. So really, I will be having a 200 amp panel with additional space installed, but keeping my service at 100 amp for now, Then I'll have new outlets added on separate breakers, as the goal is to increase beyond the 26 cards I am running now.

If I may ask, since your pushing 1060's on the same boards:
- Do you have them full (using all 13 slots)?
- What processor are you using? I did not think it mattered, but am trying to pinpoint any differences).

If you don't mind, I may have more questions about your rig, either in this thread or outside of it. Either way, I appreciate the information.

Will a software update through the ubuntu updater affect the performance of nvOC? The updates i thought could be disruptive are the nvidia drivers and xorg.

When Genoil incorporates Claymore: D
http://imgur.com/xx6UnGD

Greetings fellow nvOC miners. I've been using nvOC and following this thread for about a month now, I am pretty impressed with it. There are few glitches here and there but most of them can easily be solved. Many thanks to fullzero and the rest of the community for creating this great mining OS.

I'd like to share my experience with 1060's and possibly help others.

I am using Zotac 1060 AMP!  6GB cards on several rigs with Asus Prime Z270-A and Asrock H110 Pro BTC+ boards. All my cards are with Samsung memory which is great for overclocking the memory compared to cards with Micron or Hynix memory. My cards with Samsung memory can OC to over +1800 on linux (+900 on windows) while the few other brand cards I had (and returned) with Micron memory could only OC to +1000 (+500 on windows) and weren't much stable.

Few people asked for settings for 1060's, here are mine:

Mining ETH only with Claymore 9.7 and 9.8 (same MH/s)
Power limit: 76 W
Core OC: +100
Memory OC: +1820 (Samsung only, don't try with Micron)
Temperature setting: 54 C (fans run at 50-70%)
Getting 25 MH/s per card (3 watts per MH/s is pretty amazing)

By no means these are final settings, there is still room for improvement... I am testing +1850 memory at 75 W power limit. Do not go below 75 Watts on power limit setting, hashrate starts dropping a lot. For ETH (and similar) there's no need to go above 77 watts for single currency mining, there's no gain whatsoever.

fullzero, I noticed you have same Zotac 1060 AMP's, give them a try with the above settings. These cards rock!

JudoFlash, there's no need to install new panel for 26 1060's if you mine ETH and not GPU core hungry algos, I run 42 1060's at and 8 1070's on a 100 amp panel. I just added 3 more 20 amp breakers. Please note that what is classified as 100 amp /240 volt panel is actualy 200 amp / 120 volt capable panel, there are two 120V main cables entering the panel (to combine for 240V) and each is capable of delivering 100 amps. Hovever, don't do any electrical work if you are not qualified and consult/hire professional electrician to do the work.

Thanks

nvOC supports CPU mining, but only for Monero (XMR), as far as I know. And it makes sense, as most (all?) of the other coins are not worth to mine with a CPU. Even Monero may not be worth it, unless you have free or very cheap energy...

Is it possible to mine with CPU with nvOC?

I only use teamviewer for the first time, for long-term use the best way is to use ssh.

The stak client ran a little bit slower IIRC.  That it requires a config file to run (only command-line option is to specify the name of the config file) is also annoying.

In other news, I replaced the Asus Prime Z270-AR I had been using with an Asrock H110 Pro BTC+.  I was hoping I'd finally get all three GPUs (two overclocked MSIs and a PNY) running on passive risers, but ended up putting the two MSI boards on the 4-port PCIe port multiplier I had been using with the other board:

http://www.ebay.com/itm/PCI-E-1-to-4-Ports-PCI-Express-16X-Slot-Power-Expansion-Riser-Card-Adapter-Board/132267216692

I'm still running v17, but have downloaded v19.  I think I'll throw that on a spare flashstick next and see if the passive risers work any better with it.

Anyone else having problems with 1080 TI FE cards on 0019?

I came to notice that in xorg.conf the Devices were set to 1050 TI.  Correcting those values and rebooting did not fix the fan speed issue.

I then realized that "lspci | grep VGA" shows NVIDIA cards are installed not the model.

Just  "01:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)"

EDIT: Doing first boot with a monitor plugged in seemed to sort whatever issue I was having with v19 and fan control.  Headless worked fine after that.  Thanks to fullzero and others contributing to this awesome project

I did, doesn't work.

I'm also not sure, when I used teamviewer at that time I connected to the monitor.

I did experience the same thing when I raised the power limit, though I will need to wait about a week to test more - all of this is on a 20 amp circuit - 26 GPU x 75 w = 1950 w, plus the rest of what is plugged in, I am already close to the 2400w theoretical max and certainly above the 80% that would be more ideal.

Next week I am having a new panel installed with more breakers and additional outlets.