Ian&Steve C.

Just a question, with your approach, I still get that cool nVIDIA app that lists all the GPU's right?

Ian&Steve C.,

As soon as the HD image is done and I can do this repeatedly much faster, I will try your approach.

Dave

you do not need to do that at all. read my previous post. installing the repository driver then purging it is not necessary. i would recommend you just install the drivers from the Ubuntu PPA. these will get you the latest drivers and cleanest install designed to work with your operating system.

after setting up the BIOS and fresh installing the OS, simply run the following:

1. add the Ubuntu PPA for nvidia proprietary drivers

sudo add-apt-repository ppa:graphics-drivers/ppa

2. update

sudo apt-get update

3. install nvidia drivers from the Ubuntu PPA (substitute 396 for 410 if you insist on that version, but it makes no difference)

sudo apt-get install nvidia-driver-410

4. install OpenCL components

sudo apt-get install ocl-icd-libopencl1

5. reboot
6. check system that it's all added
7. shutdown and start adding some cards and check behavior

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

BTW TBar, You asked a valid question:

"Why does eng4hire feel the need to re-install the OS over and over."

ANSWER: eng4hire is NOT a Linux guru and once the nvidia driver install fails into the logon loop I know of no other way to try again except a new OS install... I cant recover from the logon loop!

Wow.

Guys, Take it down a notch huh? Why this cant be made into a flawless script or included as a part of an install CD is beyond me so you two and the steps are all I have.

Geez... Listen to this chain of steps... PLEASE CORRECT MISTAKES....

1.) Install Ubuntu 16.04 LTS stock (No updating)
2.) To get rid of / disable the Nouveau kernal/nvidia driver install the repository nvidia driver (384.130) using "Additional Drivers"
3.) Restart
4.) With NVIDIA-Linux-x86_64-396.54.run in the home directory and the execute bit set, enter the console with CTRL ALT F1
5.) As TBar set fourth which has worked before for me, Log In
6.) Type, sudo service lightdm stop press [ENTER]
7.) Enter Password, once it's stopped, press [ENTER]
8.) sudo apt-get remove --purge nvidia* [ENTER]
9.) sudo apt-get autoremove [ENTER] to remove the repository driver.
10.) sudo ./NVIDIA-Linux-x86_64-396.54.run [ENTER]
11.) Respond to the Screen prompts
12.) The install script will Always fail it means nothing, continue installing
13.) Don't have 32-bit comparability, keep going
14.) Allow the install to register the Module
15.) Don't have a custom xorg.conf, let it build one
16.) After it's done, sudo reboot

Did I miss something or did I do something out of sequence?

LOL gladly.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

If you want to take over here you are welcome. Have fun. I have other things to do than deal with your ignorance and arrogance.

Bye.

Well you’re 1 of 2 people that have a mining board currently. So it’s not exactly some huge feat. There just aren’t many people trying it. I could try to lament about how I’m the “only” one running 6x 1080tis, but who cares if I’m the only one that’s even tried? Wildcard looks to have moved on to different things. We can only speculate about the issues he had at the end since he never posted, but he was running well enough and long enough to have a 600-700k RAC at one point. You had some issues of your own with that system. I saw a few days ago that it dropped a card, only listed 10, but you had Stderr.txt logs spammed with retries from failing to initialize device 11. So don’t act like you’re impervious to issues as well.

A “mining” board isn’t special. It’s using normal consumer components. It doesn’t have a special chipset or anything for use with mining. They are just using the hardware to more of its capacity. Like hooking up all of the available PCIe lanes from the chipset to physical slots, where non-mining boards would either not use those lanes at all or use them for other devices like M.2 slots. Notice how your board doesn’t have things like that? Because the B250 chipset only supports 12 total lanes and they are using 11 of them for PCIe x1 slots. It’s not special. It’s just configured differently. Nothing about it is inherently harder than setting up for any other system. It does not require special treatment.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

Tell me ONE Person that has a Mining board Working correctly in Ubuntu by using a Repository Driver...just one.

You don’t HAVE to use the nvidia run file install method though. If you use the PPA drivers from the Ubuntu official repository, they install fine. And they are packaged to work properly and have the best compatibility with Ubuntu.

The drivers at their core are still nvidia closed source drivers. Just a different install method.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

Actually the Beta version Ubuntu 18.04.1 & Lubuntu 18.04.1 ISN'T working fine. They will Not allow you install the drivers from NVidia without first installing and then removing the Repository driver. The Catch is there is something Seriously wrong with using Autoremove on 18.04.1. Both Myself and Keith have both had to reinstall the 18.04.1 OS after running Autoremove . I've Never had that problem in 16.04, and I've never heard of anyone else having that problem with autoremove Except those running 18.04.1. Considering we are down to the driver being the problem with this particular User, it would be Intelligent to stay with an OS that has a History of Not having Driver Installation problems. I've installed dozens of 16.04 systems, Never had a problem installing the Driver. I have yet to install the Driver Once in 18.04.1 without having problems.

The question is, What is the problem eng4hire is having that makes him think he has to reinstall the OS? He shouldn't have to reinstall the OS as Both drivers have a history of working in 16.04 which means if he can't get one to work then just install the other or just remove the one you can't get to work from the Recovery Mode and boot without a driver, that Will work BTW.

You shouldn't have to reinstall, why do you think you have to reinstall?
To remove the driver from NVidia simply boot to Recovery Mode, enable Networking and Root, then run nvidia-uninstall
You already know how to remove the Repository driver from Recovery Mode.
Read this again, I would suggest you install the driver First, then install the other GPUs, then reinstall the driver again telling it to update the xorg.conf with all the cards installed.
It says to install the Driver Twice, NOT every time you add a card. Once before and then Once after ALL the cards are installed to update the xorg.conf with All the cards added.

The ONLY time I've EVER had to reinstall an OS after a Driver installation attempt was with 18.04.1, nothing from 10.04 to 17.10 has ever had the problem 18.04.1 has. I speak from experience.
BTW, I've just managed to get a FIFTH GPU working in Mac OSX, something else that is quite a feat to accomplish.

i'm going to list what i think you should do, maybe some of it is already done, but it's good to check everything.

-remove all but 1 card in the system. make sure this card is in the main x16 PCIe slot and connected to the monitor.
-disable the onboard video/GPU in the BIOS
-disable mining mode in the BIOS
-enable 4G decoding in the BIOS
-set all PCIe speeds to gen3 in the BIOS
-reinstall Ubuntu OS fresh (is there a reason you're sticking with 16.04? 18.04 is the latest LTS and it works fine)
-add the ubuntu official PPA for nvidia drivers "sudo add-apt-repository ppa:graphics-drivers/ppa"
-and update "sudo apt update"
-install nvidia drivers with "sudo apt install nvidia-driver-410"
-sometimes this doesnt install the OpenCL components, but that is easily remedied with one command "sudo apt-get install ocl-icd-libopencl1"
-reboot
-verify that the one card is working properly with BOINC/SETI
-shutdown and start adding cards one by one. let the system boot and verify that it's all working before adding more. DO NOT change which card is connected to the monitor, or move the first card from that main x16 slot.

this general process has worked several times for me and on several different systems.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

Yea, well, it didn't work. So time consuming to get this working.

Now I have to do a full re-install of the OS to get back to square one.

I'm going to make a copy of the hard drive and cut this down to size.

I have no idea why the installation of the 396.54 doesn't always work the first time.
I must be doing something out of sequence and not realizing how sensitive this is to that.

Maybe allowing Linux to update the all that crap in the OS is a mistake?

Is it okay just to reinstall the existing nVIDIA driver or do I need to go through...

sudo apt-get remove --purge nvidia*

sudo apt-get autoremove

before blindly hitting "y" to run autoremove, you should actually check what packages will be removed. i've seen it pop up a few times where autoremove ends up wiping out many important pieces of the OS.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

Ah! So reinstall the driver for each new GPU added!

Key concept!

you really shouldn't need to do that no matter what hardware you're using. the system should auto-detect a new GPU and set it up for you. this is how it has worked in every ubuntu system i have. adding, removing GPUs, doesn't matter. 7- GPUs on my main host and drivers were installed once via the official Ubuntu PPA. i simply added the additional cards afterwards and they "just worked".

most people prefer the Ubuntu PPA drivers because they provide an easier install process and the best compatibility for Ubuntu and wont break anything during install. i've only ever needed to use the nvidia provided runfile on my Ubuntu 17.10 hosts where the PPA wont push the latest drivers and i needed the 396+ drivers to try out the CUDA9.2 app.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

Well, Murphy must have moved in to stay at my house.
1st attempt to install 396.54 the old way failed with the login screen loop...

Going to try and oust that driver through console to avoid a re-install...

UGH...

Is it okay just to reinstall the existing nVIDIA driver or do I need to go through...

sudo apt-get remove --purge nvidia*

sudo apt-get autoremove

Ah! So reinstall the driver for each new GPU added!

Key concept!

Using your 396.54 as prescribed...

On second thought, I wont use the latest if there is no improvement just to be consistent with your efforts.

IS THERE ANY WAY TO MAKE A DISTRIBUTION CD WITH ALL OF THIS TINKERING IN TACT ??