I tried 2 scenarios to install that and both had failed.

#1: ubuntu-16.04.2 server with already installed amdgpu-17.10 (running system).
Attempt to install rocm failed at amdgpu module rebuild stage (version 17.10).
Attempt to boot new kernel resulted in black screen during kernel start at the moment when a video mode should be changed.

#2: ubuntu-16.04.3 desktop (new installation, no yet amdgpu driver - used built-in one).
rocm installation succeeded, after reboot the new kernel is running well.
Attempt to install amdgpu-17.30 failed at module rebuild stage: "headers 4.11 not supported" message, build log contains errors with incompatible argument types of atomic vs struct.
Original amdgpu driver (built-in) mining with the same hash rate as before (2 MH/s lower that it does on Windows).

Any hints from those who has it working: your ubuntu version, amdgpu version (if not built-in), order of installation?

Linux!!! Linux!!!

Sorry if this was already fixed or is a duplicate report. Didn't read the thread for a long time.

I found that my nvOC system boots too long (around 2 minutes). The cause of that is a swap partition in /etc/fstab that does not exist on USB drive (was /dev/sda5 at install time). After commenting it out a 120 seconds swap disk mount wait time has gone.

Have anybody tried to deposit more than 10 ETC with no result?

I have an idea that they may not deposit small amounts. It is like 1 ETH minimum for Ethereum, it could be, say, 10 ETC minimum for Ethereum Classic, and sending more than 10 will fix the issue? I am tempting to send some ETC just to check it, but afraid it will hang as well.

Can anybody confirm that even 10+ ETC deposits are stuck?

I don't believe they're able to cope with that amount of tickets. They receive around 3000 tickets/day (according to ticket numbers I placed twice with 15 days interval for different issues). If they felt that 5 chat moderators can help, they do not have enough staff to handle it.

So I've only been waiting to see my ETC occasionally deposited.

Meanwhile I withdraw my funds from Poloniex within daily limits every day to sleep well.

9 hours after withdrawal the ETH was transferred to the wallet with different TxID. I read it is a "normal" practice for Polo to instantly show fake TxID and then transfer with different one. I don't know why. But today's withdrawal of 9.5 ETH was really instant, few seconds. Still no ETC deposits...

It seems that something is seriously wrong with Poloniex.

I started to withdraw funds from it today. Funny, ETC withdrawals are almost instant (but still no deposits). But ETH withdrawal generated Tx in seconds, but after 8 hours there is no such transaction on the network and of course there is a zero balance. WTF...

fullzero wrote that there are some issues with X11 configuration mixing the drivers. I guess it is ok in manual mode for particular installation/rig, but a bit harder to make it for any configuration of rig.

Same issue here, 15 days left, no ETC deposits on the balance (still on the deposit address), 12 days of no response to support ticket. Started to mine to a wallet.

Anybody receives ETC to Poloniex?

Claymore's logs are huge.

AFAIR, they are disabled with '-dbg -1' option. Change this if you need logs.

Rig or miner? Have no experience with genoil, but 'uptime' shell command can give you the rig uptime.

Git suggestion was only to simplify the implementation of all changes and make easier to track what's changed.
Using the script as a learning example is the point, but I also think that using bash loops is a better example than copy/paste style.
Finally, oneBash is good, but I would split it into separate parts like overclocking, miner selection, watchdog, monero mining, etc to make easier to tune/restart some parts.

In any case, I really appreciate your efforts. I did not use nvidia GPU till now, and your distro was the best way to start using nvidia with Linux. Thank you for working on it and sharing.

I was going to rewrite exactly these parts of script since I also don't like such code. But before doing so I asked (in another rxOC thread) to put oneBash files to GitHub repository, so such changes will not be like a monkey work to copy/paste a lot. Until that I feel no motivation since I have only couple of rigs with 1050 running nvOC. It might change when I receive 12 PCIe MOBO (waiting for Biostar board from Aliexpress). But I am pretty sure that git is the best way to handle a lot of change requests and integrate them into this great distro or run just own branch.

PS. I was talking about putting oneBash only to a repository, not about full distribution.

It seems like no one has succeeded with that because no one posted a success story
I am ok to use Polaris BIOS Editor, or do manual hex patching. No matter what I was not able to fix the issue with high power consumption under linux with rx470 and asked for anyone who WAS able to do that to share how.

Actually, I tried to remove all nvidia-smi commands at all and reboot. Still higher power consumption with modded BIOS comparing to Windows.
That is why I asked anyone who was able to run rx470 undervolted on Linux: please post your PBE screenshots and linux settings.

Didn't you think about putting oneBash to GitHub? Makes it much easier for custom modifications.

For instance, I would group similar options for cards (mostly for nvOC version) into loops using bash (you can pull changes if like it). Another option for me may be to split oneBash into few option files like 10-power.sh, 20-miners.sh, etc... Next step would be to add a kind of rig autodetect (say, using NIC MAC) and load rig-dependant options automatically, still having single settings repository.

Will do it for myself, but integration of user changes will be much easier using git.

You've missed the point. I have modified BIOS using PBE, but it seems that drivers ignore flashed settings and use some driver defaults for voltages.

I guess I do something wrong. Anybody who is running rx470 under linux with low voltages, can you show your PBE values and corresponding rxOC settings (powerplay, etc)?

After trying nvOC I decided to try rxOC. It works, but I have few questions.

My test rx470 rig has modified BIOS with low voltage levels (around 950mV for GPU, for instance). Under Windows I used WattTool to alter clocks and voltages as well. But if I didn't run WattTool, I had BIOS-defined low voltages and clocks.

Now under rxOC I set GPU clock to 6 (of 7 max). It runs but I have 200 more watts for 6 GPUs comparing to Windows system. Using rocm-smi, I realised that GPU still runs at level 7. After explicit command to set it to lower levels (say, 5 or 6), I can see it set for few seconds, and then it immediately reset to level 7. I read that change is active until another OpenCL program is started. So I tried to change it before Claymore's miner, and after it started - nothing helped.

In addition, since power usage is much more than under Windows, I guess that drivers ignore BIOS voltage levels and set them explicitly to some defaults (at least, I read about that on Windows some day).

So in short, all rx470 run and maximum clocks and use some high voltages and not those written in the BIOS.

Any suggestions? I wish I can have the same power and performance as on Windows system. Hash rate is a bit lower comparing to Windows one (around 0.5-0.7 MH/s per card), and it might be the difference between Linux and Windows miner timers (it always was for fglrx as well). But what about voltages? Please advise!

I totally agreed with you. Also the semicolon-separated format is not quite good as a JSON even if easy parsed.

But I try do not suggest a lot of changes. Once I suggested in PM to rework a bit config file format, so it could contain both common and instance (rig) specific options. Claymore rejected that first, then said he will consider, and then I have not heard from him about that. And this was not implemented, so you can't have a single config file for all rigs that unlike epools.txt needs more than just an individual rig name.

As a result, I understand that he is not willing to make massive refactoring of his miners. The change I suggested is easy to implement and it will look better I think. I really don't understand why he did that this way instead of another extra field.

In the end I don't think that it ever will be changed at all

Excellent.
But I don't really like version strings with spaces and dashes. Maybe until you updated all miners you may consider something like "9.7/CN" ?

In addition, if the utility will provide key/certificate option and the miner will check it, it may replace a missing password protection or other authentication options for monitoring/management.
And THAT is where I highly appreciate SSL/TLS.