Topic: [OS] rxOC easy-to-use Linux AMD Mining v_stopgap - page 10. (Read 31393 times)

Yes it worked with one card. I am wondering if it doesn't like the mix of different cards rx470 / 570s

If you try booting with a single gpu direct in the primary pcie slot; does the OS boot properly?

I tried running rxoc but I get a scrolling error, its hard to read but looks like error PCIe

Running an ASUS Prime H270-Plus, 8Gb RAM 4x RX470/ RX570. Have tried flashing the BIOS, enabling Above 4Gb decoding and setting PCIe to Gen 2.
Anything else I can try?

This is the next thing to do on my list after building out some hardware.  I will start working on it in earnest this weekend; although I am not sure when it will be done.

It is technically in percent:

OVERDRIVE_PERCENT

Although it is further complicated by the clocks having distinct ( levels ) which are essentially static clock speeds which can be selected.  Why AMD didn't use a standard offset as Nvidia does and previous AMD drivers did; I have no idea.  AMD has announced they are making a new mining optimized driver and api; at this point I would only make an AMD rig if someone offered to sell me GPUs at significantly discounted rates.

Any estimates on when the next version is coming out?

Yes, I will use this driver for the next version.

Thanks for your reply.

Is the overclocking in the rxOC oneBash file a level i.e. 1-10 rather than MHz as in the nvOC?

Any plans to implement the blockchain driver from AMD?

Currently no; rxOC does not support individual OC: mostly because the AMD API is shite.  I am now working on a new rxOC version.

To have a high resolution team viewer either; leaver a monitor connected to the primary GPU or attach a dummy plug to the primary GPU.

Is it possible to set a resolution for Teamviewer, when there is no monitor connected to the rig? Because now I get a very low resolution.

I already looked here: https://wiki.ubuntu.com/X/Config/Resolution, but I am still a Linux beginner.

What do I have to do when I want to set the resolution to 1280 x 1024 for example?

Can you set individual OC for different cards?

Thank you for the quick response and sanity-check.

I've had to abandon my late night trouble-shooting due to other affairs, but I'll be back to continue in a few weeks. In case you're curious, I left off attempting a clean Ubuntu desktop install (kernel 4.10.5) with AMDGPU-Pro 16.60 drivers. Also found it surprising that wolfamdctrl binary was so difficult to obtain - however, managed to pluck v1.2 from a "simple" OS as was trying that next as well...

Anyway - appreciate all your efforts into this distro. I'll certainly let you know if I come across anything substantial before you figure it out yourself.

I still need to mess around with the direct amd api commands. In my experience so far; the current amd api is absolute rubbish for the exact reason you are describing above.  AMD has announced they are making a new mining driver and api, hopefully it is as good as nvidia's.  I will look into making the current amd api OC better (it has to be possible as it can be done by the windows OC applications) probably after I finish nvOC v0019 and translate as many of the v0019 features to rxOC as possible.

Hi fullzero and all.

Really...REALLY like the NvOC OS for my nVidia cards - using v17 without any hiccups for over a month now.

However, trying to spin up a 12-rig tb250-btc pro rig at the moment using rxOC v0012 and finding that I'm completely incapable of lowering the power consumption of the GPUs.

How can I verify that the MEMORY_OVERCLOCK_LEVEL and __CORE_OVERCLOCK_LEVEL parameters are actually doing what they're supposed to? The rocm-smi utility shows that the lower levels aren't being utilized properly...I'm either at the lowest level or highest, depending on the PERFORMANCE_LEVEL variable alone. So...sadly, it seems I can't get a middle ground.

No issues with these same cards (mix of RX580 Nitro+ and Pulse) within Windows.

Did some digging around with other mining OS and found that two of the big ones use wolfamdctrl to modify core-state and mem-state as well, but I haven't had any luck with that utility, either. The only values I seem to be able to change between rocm-smi and wolfamdctrl are the Fan speeds, memclock/core-clock, and overall Performance Level. Also, modifying the pp_tables directly by using 'echo manual > /sys/class/drm/card*/device/power_dpm_force_performance_level" and similar shows that nothing sticks.

All the work is being done under root - so I know it's not permission. It's almost as if something is overwriting every change I'm making, although nothing is jumping out at me...

Any ideas?

Here's an easy to see example of what I'm talking about:
^{root@bad:/home/m1# echo "manual" > /sys/class/drm/card0/device/power_dpm_force_performance_level
root@bad:/home/m1# more /sys/class/drm/card0/device/power_dpm_force_performance_level
auto}

10% overdrive is probably too much for one, some, or all of your GPUs.  I would try setting overdrive to 0 to see if it is the problem.

Thank you FullZero

Yes i use plug in my monitor on mother board ...im using Biostar TB250 BTC pro, pentium G4650,16gig memory sticks,usb 3.0 sandisk 32 gig 6 rx570 and 6 470 all 4 gig,,(for test i only plug 1 gpu) no nvidia card mix on

mobo bios is updated,bios setting is configure as on the instruction

Attach your monitor to the primary GPU; the one connected to the 16x slot closest to the CPU.  rxOC and nvOC don't support using integrated graphics.  This is most likely the problem.

Thank you fullzero for your awesome work.
Having configured the BIOS according to the guide and added "pci=noaer" as well as "pcie_aspm=off" to the grub boot parameter, I can successfully boot the machine. Currently I face a few problem with my 12 R9 390X rig. First, it boots into low graphics mode, which isn't too bad as there is still a terminal for me to run oneBash. However oneBash always hangs here.
http://imgur.com/thIws8b
I tried running Genoil and Claymore in ~/eth/Genoil and ~/eth/9_5 respectively and both ran into the "GPU fault detected:147" just after a few seconds of mining.
http://imgur.com/E36sdze
Dividing the GPUs into separated Claymore sessions helps but I have never been successful in running all the 12 GPUs. While ubuntu recognises all 12 GPU judging from lspci output, if more than 8 GPU is mining the error "GPU fault detected:147" will appear.
Any hint on how to solve these problems?

Hardware: Biostar TB-250 Pro Ver.6, Intel Celeron G3930, 4GB Ram, 3x 1200W Server PSU, 12x R9 390X.