Blackmagic Forum

Sun Mar 11, 2018 3:31 pm

This took me a three tries where I removed everything and started the whole process over again, but I finally got a painless install. Here it is:

Assemble the PC as specified. My particular build used:

ASUS micro-ATX motherboard (why is a mini-ATX smaller than a micro-ATX?) with an LGA1151 socket,
3.6GHz , 7th gen Intel Core i7 processor,
64Gbytes of DD4 ram,
Asus GeForce GTX 1070 Ti Graphics card,
DeckLink Mini Monitor (Mandatory under Linux)
1 512 MByte SSD hard drive for the OS
1 4 TByte WD Red hard drive for the /home directory.

This particular motherboard only had 1 16x PCI slot, 1 16X PCI slot running at 4x and two 1X slots. The original intent had been to use an older Nvidia dual DVI card to drive the monitors from the 4x slot and dedicate the 1070 ti to Resolve, but the system would default to the 16x slot since it was first. I eventually gave up and just used the 1070 to drive both HD monitors as well as Resolve. I don't do any intense 3D stuff anyways. That did leave a 4x slot open that I could have used a better DeckLink card in, but I had alread purchased the 4K mini monitor. Oh well.

Once the hardware build is finished:

Download Centos7 from https://www.centos.org/download/ I used the "DVD ISO" since this would be an Internet connected PC and I can add software later as needed. Burn the iso to a DVD. Note the "DVD ISO" will fit on a standard DVD, the "Everything ISO" needs a dual-layer disk.
Install Centos7. I have an internal BlueRay DVD for this, but you can use a USB connected DVD player if you don't have an internal. Been there, done that. I installed the GNOME desktop package. I did not install legacy stuff. I don't have any old programs that require it. By default, I install the compiler packages - you'll need them.
If you have Internet access, update the install at this point. You will need Internet to finish the install, but you can disconnect after completing the build. You will probably have to wait for awhile after you first boot it up as the automatic updater is checking what packages need to be upgraded. I prefer to open a terminal and simply use the command (as root or with sudo) "yum update".
Install the "extra" repositories. Please make sure you do this after you have upgraded the base packages. Otherwise you may run into compatibility problems. I'll explain in a moment. "yum install epel-release"
Install RPM Fusion. Bring up your browser and navigate to "https://rpmfusion.org/Configuration" and pick both the "free" and "non-free" packages. I prefer the command line setup as the GUI setup almost always fails with the GUI software installer. They list the exact commands you will need. Do not install any packages yet except the Nvidia drivers.
Install the Nvidia drivers. You can find what drivers are available by issuing the command "sudo yum list available | grep nvidia". I prefer the "akmod" package as it will compile the nvidia drivers whenever it needs too. I've had issues in the past with the "kmod" package because it doesn't always get updated correctly. If you upgrade your kernel and the new nvidia kmod package isn't updated on the repository, you have issues. Lately, they've been pretty good about it, but I've been burned so I just use the akmod. I had to install the cuda package as well "sudo yum install xorg-x11-drv-nvidia-cuda.x86_64", but that may have been because I failed to install opencl-filesystem. I didn't veryify because I didn't want to go through the whole process again. You may be able to just install opencl "sudo yum opencl-filesystem" and just the akmod package. If it doesn't work, you can always come back and install the cuda package. Resolve will let you know if it can't find what it needs. Note that installing Nvidia from these packages automatically takes care of other things like blacklisting the nouveau drivers. Reboot your system when you have completed this step and it should come up under the nvidia drivers rather than the nouveau drivers.
You probably need other packages like "vlc". These will be under the nux desktop repository. There are conflicts between nux and rpm-fusion though. At this time, edit the rpm-fusion respository files in "/etc/yum.repo.d/" There are four files, but usually only two of them have to be altered. Use your favorite text editor and change "enabled=1" to "enabled=0" in all four files. Make sure you do this step before installing any other packages. rpm-forge usually has more updated software than nux, but things like vlc will fail to install due to dependancies. Once you have those disabled, you can enable the nux repository. At a command prompt, issue the command "yum -y install http://li.nux.ro/download/nux/dextop/el ... noarch.rpm" Once that is done, install any of the desktop multimedia packages that you need. (like vlc, etc)
Download the Davinci resolve package from "https://www.blackmagicdesign.com/support/family/davinci-resolve-and-fusion" and get the declink drivers (Desktop video) from "https://www.blackmagicdesign.com/support/family/capture-and-playback" and install per their respective instructions. I used Davinci Resolve Studio, but I assume Davinci Resolve will install the same way. System sound with Davinci Resolve 14 does not work at the time I wrote this so you have to have a DeckLink monitor card to have system sound. My preference would be to have a capture and playback card, but as described above, it didn't work out that way for me.

My system was intended to be a dual monitor system, but because Linux can't really use the DeckLink without some editing or playback software, it's now a three monitor system. It worked out for me though because I picked up a 4k monitor and my descktop monitors are just HD. I have the 4k monitor mounted on the wall above the workstation and tilted down to reduce glare and optimise the viewing angle.

Note: if you need additional software packages off of rpm-forge (or if you need to update the nvidia drivers), you will have to re-enable the repositories, but just know that it may cause some issues. Thankfully, Nvidia doesn't update all that often.

Sun Mar 11, 2018 7:17 pm

Thanks for this!

I’m getting ready to switch from macOS to CentOS.

Just curious—any reason you didn’t use BMD’s custom ISO of CentOS 7.3 that has PostgreSQL preinstalled?

Sent from my iPhone using Tapatalk

Tue Mar 13, 2018 10:12 am

In one of the iterations to find an easy install path I used BM's image. As far as I could tell, it only offered two things:

A postgresSQL server
Nouveau driver already blacklisted

I wasn't interested in the pg server because I already have a 100TB raid array that also hosts both a mySQL (mariadb) and a pg server. A built-in pg server made no difference to me and it's not all that hard to install and set one up using rpm package managers. It was also very annoying to have the nouveau driver blacklisted before a replacement was in place.

If you install from Nvidia's site, you have to manually blacklist stuff and that is a pain, but if you install from rpm-forge, things are much easier. the package blacklists it for you.

All-in-all, there is nothing wrong with BM's install image and if you are not already familiar with Centos, it's just fine to use. The lack of a frame buffer on these giant monitors until you install nvidia was annoying so I opted not to use it.

Tue Mar 13, 2018 10:55 pm

The NVIDIA driver IS installed as part of the Resolve CentOS 6.8 and 7.3 build image installation. That's why it's easier for people with little Linux familiarity to use the Resolve build image.

Wed Mar 14, 2018 11:39 am

I installed from the iso that was included on the SD card that came with the USB key and there was no nvidia driver installed. nouveau was definately disabled and instead of a frame buffer for the install, I had a 640x480 display that was incredibly difficult to read. It couldn't fit the necessary info on it and it was very fuzzy. switching to text console mode revealed no nvidia driver modules installed in the kernel and a search of the disk using "find" revealed no nvidia files anywhere on the iso.

nope - no nvidia on the iso

It's all good though - it was easy enough to install the akmod packages from rpmfusion.

Wed Mar 14, 2018 6:47 pm

Believe it or don't, I don't have the ability to access the current SD cards, and I'm surprised there is an ISO on the SD card. Could you tell me the name and size of the ISO that's on the SD card?

Wed Mar 14, 2018 7:48 pm

well -- installing nvida drivers and cuda libraries is nowadays a simple one-liner on nearly any popular actual linux distribution (o.k., on "alpine linux, which in the meanwhile has become the most common choice for container based systems, it's a little bit more difficault, because they use musl instead of glibc, and on intels "clear linux", which should be otherwise respected as the fastest resp. most perfomance optimized linux, it's also tricky, because of its large amount of kernel modifications).

i really don't see, why resolve makes so much buzz about this particular distribution and it's mandatory use? sure, e.g. foundries nuke or sgos mistika products are also officially only supported on distributions of the redhat heritage, but they do not handle it like a religious dogma. in fact, both of them use more modern linux installers, which are well maintained and will work on most other distros just as well in a really user friendly manner. something, which doubtless can not be said about resolves actual installer.

but BMDs modified CentOS installation media also used to show a few other old-fashioned peculiarities, which differ even from real CentOS! they are not coming as an hybrid disk image and only support old "legacy" BIOS boot, not the more recent EFI system initialization. little nasty deficiencies, which can become really challenging on some hardware or in virtualization contexts.

but -- o.k- -- if someone actually had luck and want's to describe/characterize his experiences as "painless", why not?

Wed Mar 14, 2018 9:29 pm

Ain't tech fun :-)

results of ls -l:
4287393792 Sep 7 2017 DaVinci-Resolve-Linux-14.0-CentOS_7.3.iso

I purchased it from B&H Photo late last month.

Dwaine Maggart wrote:Believe it or don't, I don't have the ability to access the current SD cards, and I'm surprised there is an ISO on the SD card. Could you tell me the name and size of the ISO that's on the SD card?

Wed Mar 14, 2018 9:57 pm

Ephraim, BMD's latest build of CentOS 7.3 with 14.3 pre-installed is available to download here: http://downloads.blackmagicdesign.com/D ... OS_7.3.iso

With that, you can create your own installer disk.

Wed Mar 14, 2018 10:50 pm

@Ephraim: That is a valid ISO with a valid size. Not the current ISO, which Seth has kindly provided a link to, but an ISO that does include the NVIDIA driver. So not sure why it didn't get installed in your case.

Good to know we are putting the ISO on the SD cards.

Thu Mar 15, 2018 5:44 am

Seth Goldin wrote:With that, you can create your own installer disk.

most dvd burning applications are able to transfer this kind of disk images to a physical media, but it should also work from usb sticks etc., because not all machines have optical media devices anymore, and some hardware (e.g. alienated apple machines) aren't able to use traditional pc BIOS but EFI for the boot process -- and that's the point, where the mentioned deficiencies of the BMD installation media come into play...

Mon Apr 09, 2018 11:30 pm

Dwaine Maggart wrote:@Ephraim: That is a valid ISO with a valid size. Not the current ISO, which Seth has kindly provided a link to, but an ISO that does include the NVIDIA driver. So not sure why it didn't get installed in your case.

Good to know we are putting the ISO on the SD cards.

Dwaine, I'm a bit confused. I tried installing with this CentOS ISO file, and was surprised to discover that only PostgreSQL 9.2 is included on the ISO. I though 9.5.4 was required?

Mon Apr 09, 2018 11:37 pm

It's not a huge deal, though--here are instructions from PostgreSQL to install via yum. I just thought that the whole purpose of including an ISO was that the right version of PostgreSQL was supposed to be included.

https://www.postgresql.org/download/linux/redhat/

Thu Apr 12, 2018 7:56 pm

Is there a config file to setup the advanced panels? The left and right panels do not function. These panels have been tested and are functional on a mac. I have tried all the tricks on the forum.

Mon Apr 16, 2018 10:53 pm

one the most frustrating issues regarding CentOS, should be seen in the fact, that it represents one of the few remaining linux distributions, which don't support libglvnd (=OpenGL[X] vender neutral dispatching library). that's a very significant shortcoming! it doesn't allow the proper installation and simultaneous operation of different graphic cards on the system. this becomes very important, if you prefer to use the intel iGPU for video output, and the nvidia device just as an auxiliary compute device. this kind of setup is quite useful in practice, because the nvidia VRAM doesn't get polluted by unnecessary tasks and you can always reset this GPU in case of errors, without loosing control over your whole machine. but again, this kind of parallel utilization of different GPUs resp. GLX providers unfortunately isn't possible in CentOS until now.

This are the real issues, why experienced users try to steer clear of CentOS whenever possible.

Thu Apr 19, 2018 10:54 pm

As an aside, I was just testing with a generic CentOS 7.4 install, and used yum install kmod-nvidia to get the NVIDIA driver installed. And that worked as expected.

However, I wanted to manually install a different NVIDIA driver, and the NVIDIA installer complained of not having libglvnd devel libraries available. After adding the EPEL repos to yum, I was able to yum install libglvnd and yum install libglvnd-devel and that made the NVIDIA installer happy.

I'll also note that Resolve would not launch, complaining about not being able to load libGLU.so.1. I don't know if that happened because I didn't select the proper packages during the 7.4 install, but a yum install mesa-libGLU solved that issue and made Resolve happy.

Thu Apr 19, 2018 11:55 pm

Dwaine Maggart wrote:However, I wanted to manually install a different NVIDIA driver, and the NVIDIA installer complained of not having libglvnd devel libraries available. After adding the EPEL repos to yum, I was able to yum install libglvnd and yum install libglvnd-devel and that made the NVIDIA installer happy.

on installing the nividia driver and libraries you can usually choose both variantes -- an lilglvnd setup and a legacy installation without glvnd redirection. and as long as your aren't using any other vendors drivers in parallel, it doesn't make much difference. but if you actually install also the CentOS 'mesa-dri-drivers' package and use your intel iGPU for video output beside the computing capbilities of your nvidia card -- i.e. the case, where libglvnd isn't anymore only a useless dependency, but should play an important role in routing GL[X] commands to the right library and device -- it doesn't work anymore on CentOS. it's very easy to reproduce by a simple 'glxinfo' command, which will throw an error in this case, if the intel iGPU is in use but the libglvnd still points to the nvidia libraries...

but o.k., it's perhaps just an issue, which will only affect a very small minority of resolve users.

Dwaine Maggart wrote:I'll also note that Resolve would not launch, complaining about not being able to load libGLU.so.1. I don't know if that happened because I didn't select the proper packages during the 7.4 install, but a yum install mesa-libGLU solved that issue and made Resolve happy.

yes -- i observed the same behavior utilizing ready made CentOS images prepared by nvidia. but 'mesa-libGLU' seems to play nice together with the nvidia OpenGL libraries and doesn't force the installation of any more incompatible dependencies.

Sun Apr 29, 2018 2:27 am

Im trying to install from generic CentOS, but I’m finding the CUDA documentation a bit wacky. Can anyone confirm if you actually need the persistence daemon?

Going through the CUDA documentation step-by-step, the documentation says that you must add a systemd unit, but I’m finding that persistenced wasn’t actually installed and the unit fails and causes gdm to hang on boot.

Elsewhere, NVIDIA says that the persistenced is only necessary for headless systems, and that X will keep track of keeping the GPU active. So if I’m running GNOME, it’s not necessary, right?

Sent from my iPhone using Tapatalk

Sun Apr 29, 2018 3:02 pm

Seth Goldin wrote:Im trying to install from generic CentOS, but I’m finding the CUDA documentation a bit wacky. Can anyone confirm if you actually need the persistence daemon?

Going through the CUDA documentation step-by-step, the documentation says that you must add a systemd unit, but I’m finding that persistenced wasn’t actually installed and the unit fails and causes gdm to hang on boot.

Elsewhere, NVIDIA says that the persistenced is only necessary for headless systems, and that X will keep track of keeping the GPU active. So if I’m running GNOME, it’s not necessary, right?

Sent from my iPhone using Tapatalk

You don’t need to install CUDA separately, that is purely for development purposes. On Windows and Linux the ability to utilize CUDA is included in the driver.

Cheers,
Mike

Sun Apr 29, 2018 3:16 pm

MikeRochefort wrote:You don’t need to install CUDA separately, that is purely for development purposes. On Windows and Linux the ability to utilize CUDA is included in the driver.

Thanks! I guess I’ve picked up weird habits from macOS.

Sent from my iPhone using Tapatalk

Sun Apr 29, 2018 6:41 pm

MikeRochefort wrote:You don’t need to install CUDA separately, that is purely for development purposes. On Windows and Linux the ability to utilize CUDA is included in the driver.

it a really butzzeling question, because there are indeed lot's of redundancy in both variants (cudart vs. driver included libcuda), but the differ in memory handling and granularity and may sometimes trigger or prevent different flaws because of this varieties. but in general i have to agree with mike, the cuda support coming with the driver package is usually sufficient to run resolve -- i had to proof this recently, because cudart was removed from debian testing a few days ago, because it was depending on a outdated version of the c compiler.

Mon Apr 30, 2018 12:07 am

Dwaine Maggart wrote:I'll also note that Resolve would not launch, complaining about not being able to load libGLU.so.1. I don't know if that happened because I didn't select the proper packages during the 7.4 install, but a yum install mesa-libGLU solved that issue and made Resolve happy.

Curious. From my generic install, I got an error about `libpng12.so.0`, but `sudo yum install libpng12` allowed Resolve to run.

Mon Apr 30, 2018 12:28 am

Seth Goldin wrote:Curious. From my generic install, I got an error about `libpng12.so.0`, but `sudo yum install libpng12` allowed Resolve to run.

it just depends on the other software, which was already installed on the machine...

but on a really "generic" CentOS installation, you will not even find all of the tools, which the installer needs to work correctly.

Mon Apr 30, 2018 9:07 am

Hey guys!

Is the Linux version of Resolve more or less stable than the MacOS one?

I'm thinking to build a standalone 8700K WS for Resolve on Linux (with CUDA) to increase stability (hopefully!?)

Starting from scratch would you suggest me Ubuntu/Mint or CentOS 7.4+elrepo?

How is dual monitor support/behavior/stability in Gnome/Resolve under Linux?
I'd like to have 2 main screens connected to the VGA on top of 1 monitor connected to my Mini Monitor 4K.

As far as a simple 2 drive raid0 is concerned would you suggest mdadm or lvm?

Thanks a lot!!!!!!!

Mon Apr 30, 2018 11:53 am

Have you looked into fedora?

Sent from my iPhone using Tapatalk

Mon Apr 30, 2018 11:36 pm

Tommaso Alvisi wrote:Is the Linux version of Resolve more or less stable than the MacOS one?

Difficult to say. Both seem quite stable. Much of this depends on your hardware.

Starting from scratch would you suggest me Ubuntu/Mint or CentOS 7.4+elrepo?

Blackmagic Design only supports CentOS. If you're extremely familiar with Linux and you're comfortable customizing different directory paths, then you could try one of the other distros, but if you run into trouble, you're on your own.

How is dual monitor support/behavior/stability in Gnome/Resolve under Linux?
I'd like to have 2 main screens connected to the VGA on top of 1 monitor connected to my Mini Monitor 4K.

I can confirm that the NVIDIA 390.48 works great with GNOME and two monitors. The hardware guide recommends 2560x1440 or 1920x1200 for the best GUI layout.

Tue May 01, 2018 4:01 pm

Seth Goldin wrote:
Tommaso Alvisi wrote:Is the Linux version of Resolve more or less stable than the MacOS one?

Difficult to say. Both seem quite stable. Much of this depends on your hardware.

15b2 is indeed already quite stable and really acceptable for practical work on linx, but this should equally true on both other systems too.

you should also consider, that resolve on mac supports a few more available codes -- especially those, which many of us need in practice quite often --, and the system color management integration resp. color rendition on the GUI screens is much better realized on mac OS than on windows and linux. these are really important advantages, which you shouldn't underestimate in their consequences.

Starting from scratch would you suggest me Ubuntu/Mint or CentOS 7.4+elrepo?

Blackmagic Design only supports CentOS. If you're extremely familiar with Linux and you're comfortable customizing different directory paths, then you could try one of the other distros, but if you run into trouble, you're on your own.

frankly, the actual support for CentOS by BMD isn't significan better, than for all this other distributions!

the only actual benefit, which you may get, could be causes by the fact, that BMD utilizes CentOS in their internal tests and perhaps also the majority of resolve customers. but as much as you will not find any useful numbers about the real popularity and share of CentOS resp. redhat derivatives in comparison to other alternatives, which are not utterly distorted by obvious business interests and snooty association to this 'enterprise grade' labeling, it's really hard to objectively decide, whether or not already more users make use of this application on other distributions? the much wider share of other distributions in real world daily linux use makes it at least very likely.

it's also undeniable, that resolve doesn't come as a nice and proper CentOS .rpm-package neither!
sure, it's really horrible inferior installer will stumble over less obstacles, if you execute it within this particular environment, but experienced linux users should be able to work around all this annoying flaws anyway.

but i really do not want to claim, that somebody new to linux shouldn't use CentOS at all!
it's just another choice beyond others, and everybody has to decide for him/herself, what he/she prefers!

if you only consider the availability and support of closed source software products, CentOS may indeed look very attractive. but this doesn't come without obvious drawbacks! in regard to available and up to date open source software packages and voluntary support by a wider user community, it looks much less attractive than many other alternatives. in fact, i would value the significance of this community driven help and gathered documentation as much more important and useful in practice, than all the promised support and backing by software manufactures and enterprise oriented marketing...

How is dual monitor support/behavior/stability in Gnome/Resolve under Linux?
I'd like to have 2 main screens connected to the VGA on top of 1 monitor connected to my Mini Monitor 4K.

I can confirm that the NVIDIA 390.48 works great with GNOME and two monitors. The hardware guide recommends 2560x1440 or 1920x1200 for the best GUI layout.

that's a quite common setup. i don't see, why it shouldn't work?
it's just important, that you do not mix different kinds of actually used families of GPUs (e.g. Intel iGPU utilization + nvidia graphics output). this would would indeed make everything much more complicated and error prone...

Wed May 02, 2018 7:11 am

Thanks a lot guys!!!!

Your help is very appreciated!!!

I know about ProRes but I need a SOLID with no bells and whistles workstation more he he!

I agree about choosing UBUNTU for the much larger user base but in the end I'd use this WS just for Resolve so maybe the official CentOS is the way to go even if it's older? Maybe I was just burned on Mac with nvidia and Cuda drivers where these could make a mess and create a TON of incompatibilities with apps and even the OS itself...hopefully BM updates the image to be at least based on 7.4 for the v15 final release! ;-)

I planned to use the Intel GPU for the Dual Monitor GUI setup (2x 2560x1440 25 or 27 IPS)
and 1x 1080 Ti for CUDA-only processing, to maximize RAW power and memory usage on the 1080 Ti.

Do you think it could work?

Does any of you use a similar setup under linux?

Or would there be a conflict on Gnome/Resolve with both GPU drivers in use?

Would be the 1080 Ti active even if with no monitor attached?

Wed May 02, 2018 3:25 pm

Tommaso Alvisi wrote:I know about ProRes but I need a SOLID with no bells and whistles workstation more he he!

it's not only prores, but the insuficient MPEG support on linux!

it's not very complicated to workaround this deficiencies by use ffmpeg, but it makes simple unnecessary complicated and should be really seen as a very significant handicap.

Tommaso Alvisi wrote:Maybe I was just burned on Mac with nvidia and Cuda drivers where these could make a mess and create a TON of incompatibilities with apps and even the OS itself...

the nvidia divers on linux are also a very problematic source of troubles, it's in the meanwhile a lot better then a few years ago, but i still prefere work without this kind of hardware whenever possible. but there are any task, which simple can not be realized without GPGPU acceleration in a satisfying manner. if need advanced video processing or machine learning tools, you simple have to live with this kind of troubles.

Tommaso Alvisi wrote:I planned to use the Intel GPU for the Dual Monitor GUI setup (2x 2560x1440 25 or 27 IPS)
and 1x 1080 Ti for CUDA-only processing, to maximize RAW power and memory usage on the 1080 Ti.

that's more or less the way, how i usually utilize nvidia cards myself. it has some very important benefits over the more common setups, where the nivida cards are used for CUDA GPGPU and as main video output. if you do like me, you can always reset the GPU, do not wast VRAM by the display tasks and the whole sytem will be much more stable (this usually concerns features like, wake up from sleep mode etc.)

on centos it's a little bit more complicated, than on other systems, where both configurations are usually available all without any additional system reconfiguration. nvidias GLVND mechanism is used by them to route GLX output to any installed card. but this mechanism isn't available or system wide supported an CentOS. if you install the necessary mesa-driver for the intel iGPU output, it will break the nvidia display output capabilities and and vice versa. this are exactly the kind of issues, which are very characteristic for this kind of very conservative and old fashioned distributions. it's just one particular example, but you will face similar troubles quite often, if you want to use inux for all your daily work.

Would be the 1080 Ti active even if with no monitor attached?

yes -- the cards CUDA computing capabilities can be used completely independent from any utilization as video output device.

one of the very important benefits of such an setup could also be seen in the fact, that you can passthrough the whole grafic card to a virtual machine anytime, because it's not blocked by any other continuous running processes on the host system. that's a very useful feature, if you want to run win10 or mac os virtual machines resp. software on the guest instances, which requires GPU hardware too.

Wed May 02, 2018 3:46 pm

Martin Schitter wrote:that's more or less the way, how i usually utilize nvidia cards myself. it has some very important benefits over the more common setups, where the nivida cards are used for CUDA GPGPU and as main video output. if you do like me, you can always reset the GPU, do not wast VRAM by the display tasks and the whole sytem will be much more stable (this usually concerns features like, wake up from sleep mode etc.)

So Intel mesa driver is generally more stable for GUI display purposes if I'm reading this correctly? ;-)

Martin Schitter wrote:on centos it's a little bit more complicated, than on other systems, where both configurations are usually available all without any additional system reconfiguration. nvidias GLVND mechanism is used by them to route GLX output to any installed card. but this mechanism isn't available or system wide supported an CentOS. if you install the necessary mesa-driver for the intel iGPU output, it will break the nvidia display output capabilities and and vice versa.

I don't care about nvidia display capability as long as CUDA processing is still working 100%, would CUDA still work after installing intel mesa driver or there would be a conflict because mesa then expect the nouveau driver for the nvidia gpu?

Thanks again Martin!

Fri May 04, 2018 3:38 am

For future reference for others, here are my notes for getting Resolve up and running on CentOS. I didn't use ELRepo like Ephraim--I disabled the Nouveau drivers the "hard" way.

https://sethgoldin.github.io/install-da ... ve-centos/

Sat May 05, 2018 11:26 am

Seth Goldin wrote:For future reference for others, here are my notes for getting Resolve up and running on CentOS. I didn't use ELRepo like Ephraim--I disabled the Nouveau drivers the "hard" way.
https://sethgoldin.github.io/install-da ... ve-centos/

that's a nice description, how to work around most of this minor driver related issues and requirements on CentOS.
i would just two additional points:

on installing the nvidia driver my the manufacturies self-installer, you can specify a lot of options, to change it's default behaviour. sure, the default settings should work in most cases, but it's very hard to guess, if it will install the GPU libraries in a GLVND enabled manner or not on this particular distribution? this doesn't make a significant difference, if you do not want to utilize the intel iGPU or mesa tools and libraries coming with CentOS at all, but otherwise it's a very important decision, which will need some additional fine tuning.
before you start the resolve installer, you should definitely check, if all tools are already available, which the installer expects and requires for proper operation. the chances are high, that most of them are already available after installing a full gome desktop environment, because they got automatically pulled by .rpm dependencies of other applications. but on a more minimalist CentOS installation, they may not be present, and the installation process will therefor fail!
for a working list of this kind of requirements, you can look at my CI based processing of a resolve installation on CentOS: https://gitlab.com/mash-graz/resolve/bl ... Dockerfile resp. https://gitlab.com/mash-graz/resolve

Sat Jun 09, 2018 8:38 am

Dwaine Maggart wrote:As an aside, I was just testing with a generic CentOS 7.4 install, and used yum install kmod-nvidia to get the NVIDIA driver installed. And that worked as expected.

However, I wanted to manually install a different NVIDIA driver, and the NVIDIA installer complained of not having libglvnd devel libraries available. After adding the EPEL repos to yum, I was able to yum install libglvnd and yum install libglvnd-devel and that made the NVIDIA installer happy.

I'll also note that Resolve would not launch, complaining about not being able to load libGLU.so.1. I don't know if that happened because I didn't select the proper packages during the 7.4 install, but a yum install mesa-libGLU solved that issue and made Resolve happy.

#WowJustWOW. As an outsider u familiar with the command-line shennanigans and multitude of other machinations Linux hosted DaVinci Resolve users apparently must suffer -- especially with the aforementioned non-support for a most critical library entity that is libglvnd, one would have to imagine the Linux community of Resolve users ought to be clamouring for -- if not outright screaming for Blackmagic Design to summarily scrap the Linux b*stars child that is CentOS, with its inordinate amount of patching requirements prerequisite to simply play nice with various on-board graphics card options on Resolve hosting workstations. #ThisIsNUTS!

OK, then...
1) WHICH LINUX DISTRO (offers the best DaVinci Resolve Studio user-experience, hands down, even if Blackmagic Design for now at least, utterly refuses to certify it as a highly recommended Linux OS host for DaVinci Resolve Studio)?

2) Do tell us Linux Wannabe's, what specific message thread link(s) in BMD's Linux Forum discussions and also elsewhere -- would you recommend to begin the self-taught journey into the world of Resolve Studio for Linux?

3) And what warnings have you, in the way of audio pipeline limitations on the leading suggested ad-hoc alternative(s) to CentOS?

O how I wish to abandon the use of DaVinci Resolve on Micro$haft Windows, in deference to a far-more user-friendly Linux distro than CentOS (which seems egregiously incomplete IMHO, with its hand-wringing, totally jack-knifed, woefully unacceptable, oblivious-to-the-graphics-cards OpenGL library scheme!

Sent from my P00I using Tapatalk

Sat Jun 09, 2018 2:43 pm

Well, as a CentOS user myself, I’ve had zero issues with this system. Audio works great, DE works great (using GNOME 3), video codecs work great once you install ffmpeg from a repo or build it yourself, and it’s dead simple to manage. Personally I don’t see what the fuss is with CentOS beyond the core libraries being older, it’s no more difficult to use than any other distro in my opinion. But that’s just my experience, YMMV.

Maybe for audio DAW users there might be issues, but that’s just speculation since I’m not one of them.

And a side note on GLVND, it’s available in the EPEL repository, so that’s a moot point.

Cheers,
Mike

Sun Jun 10, 2018 1:47 am

MikeRochefort wrote:Well, as a CentOS user myself...Audio works great, DE works great (using GNOME 3), video codecs work great once you install ffmpeg from a repo or build it yourself, and it’s dead simple to manage.

Maybe for audio DAW users there might be issues, but that’s just speculation since I’m not one of them.

And a side note on GLVND, it’s available in the EPEL repository, so that’s a moot point.

Cheers,
Mike

GLVND seems like an essential for use with various 21st century graphics cards -- so
why on earth won't Red Hat or whomever the dinosaurs are, that won't include it in CentOS for best user experience?

Are not the community of Linux users publicly decrying its absence in CentOS AS they ought?

And Mike kindly tell us what that "DE" moniker stants for as I'm going to ve new to Linux (plys, I'm currently in Wilmington, DE (Delaware) as I write this, by the way. Haaaah!

Thanks for interacting on this!

Resolve Studio v14.3....014 | MiniMonitor | DTV 10.9.7 | Win64 8.1 | ASUS ROG G751JL, Intel i7HQ, 24GB, 1TB HDD, 500GB EVO 850 SSD, 2GB GTX965M | Mackie MCU Pro | Softube Console 1 Mkii | Shuttle Pro 2

Sun Jun 10, 2018 2:59 am

Red Hat’s focus is on enterprise situations where stability and reliability are key. So any package that is in the system is thoroughly tested and needs to be mature. The other thing is that when a RHEL major version is released the distribution does a package freeze within the year to months leading up to it. That essentially means that no new packages beyond what the current packages add for dependency updates can be added into the core system/repositories. That is carried through for the entire 10 year lifespan of support. GLVND is still relatively new so it makes sense that it wouldn’t be included when Red Hat did their package free for RHEL 7 back in 2013-2014.

Anytime we want something new, we get it through external repositories, just like how Ubuntu uses PPA’s. The small difference is that some of these repositories follow the principle of don’t alter/overwrite base packages. So certain projects won’t end up in these repositories if there’s that kind of conflict. Otherwise, repositories like EPEL (Extra Packages for Enterprise Linux) which is derived from the Fedora build system and ELrepo (Enterprise Linux repository) offer us mostly what we need for those extra goodies. There’s also the IUS repository (there are some conflicts here with certain web frameworks), Nux Dextop, and Software Collections (scl, synonymous with Red Hats developer toolkit’s for getting newer packages to use in an isolated environment). Other repos like RPMFusion are considered dangerous or unsafe as they are more likely to overwrite base packages.

In my eyes, GLVND isn’t that big of a thing as I don’t know many people who want or try to use multiple mixed vendor cards from AMD/NVIDIA in their systems. Particularly for OpenGL unless you're driving monitors with multiple cards. For me something like that for OpenCL would be preferable as I recently bricked/de-bricked my system dealing with a mix of Intel/NVIDIA OpenCL drivers trying to get Fusion/Resolve playing nice. That has to do with NVIDIA using OpenCL 1.2 and INTEL/AMD using 2.x.

To your other question, DE stands for Desktop Environment. It’s the collection of applications/UI that you use. Some examples are GNOME, KDE, MATE, Cinnamon, LXDE, XFCE, etc. RHEL/CentOS/Fedora/Ubuntu use GNOME as their default DE, while other flavors or spins might choose another. Over in Ubuntu land, there’s Kubuntu, Lubuntu, Xubuntu which are normal Ubuntu with different DE’s by default.

Cheers,
Mike

Thu Jun 14, 2018 4:50 am

So forgive me if this has been asked. I am building a new computer and thinking of putting Centos 7 on it as I really don't want to use Windows. I am a newby who has never used Linux before. Is there an automated installer like on Windows and Mac OS or do I have to do somethings manually with code?

Thanks

Thu Jun 14, 2018 5:01 am

Download an ISO from one of the mirrors on the following page:

http://isoredirect.centos.org/centos/7. ... os/x86_64/

If you are new, I would recommend the DVD iso so you can pick and choose packages, or the Live GNOME option that pretty much has the OS set up in a pre-built fashion. The Live ISO’s aren’t meant to be used for installation, but can be.

You won’t have to do any coding for installation, but the major points to keep an eye on are the disk space (install location) and networking. Make sure you go in and enable your Wifi/Ethernet poets otherwise you’ll have to activate them manually post installation.

Cheers,
Mike

Thu Jun 14, 2018 5:29 am

MikeRochefort wrote:Download an ISO from one of the mirrors on the following page:

http://isoredirect.centos.org/centos/7. ... os/x86_64/

If you are new, I would recommend the DVD iso so you can pick and choose packages, or the Live GNOME option that pretty much has the OS set up in a pre-built fashion. The Live ISO’s aren’t meant to be used for installation, but can be.

You won’t have to do any coding for installation, but the major points to keep an eye on are the disk space (install location) and networking. Make sure you go in and enable your Wifi/Ethernet poets otherwise you’ll have to activate them manually post installation.

Cheers,
Mike

Thanks man. So then also too every Resolve update should be easy to do as well after that?

Sat Jun 23, 2018 1:07 pm

MikeRochefort wrote:Red Hat’s focus is on enterprise situations where stability and reliability are key.

firefox/mozilla recently switched it's build system producing those binaries, which should run on any linux distribution, from CentOS to debian for good reasons:

https://gregoryszorc.com/blog/2018/06/2 ... ox-builds/

although "reproducible builds" are perhaps a goal, which is much more important in case of open source projects to archive an efficient differentiation between bugs, which are more related to the peculiarities of a specific build environment resp. varying library versions, and others, which indeed concern the actual applications code, it's still a very impressive documentation about such a drastic change and all the requirements to make such a huge linux application as compatible as possible with different runtime envrionments resp. actual utilization on different linux distributions.

Sat Jun 23, 2018 4:49 pm

MikeRochefort wrote:Red Hat’s focus is on enterprise situations where stability and reliability are key. So any package that is in the system is thoroughly tested and needs to be mature. The other thing is that when a RHEL major version is released the distribution does a package freeze within the year to months leading up to it. That essentially means that no new packages beyond what the current packages add for dependency updates can be added into the core system/repositories. That is carried through for the entire 10 year lifespan of support. GLVND is still relatively new so it makes sense that it wouldn’t be included when Red Hat did their package free for RHEL 7 back in 2013-2014.

Anytime we want something new, we get it through external repositories, just like how Ubuntu uses PPA’s. The small difference is that some of these repositories follow the principle of don’t alter/overwrite base packages. So certain projects won’t end up in these repositories if there’s that kind of conflict. Otherwise, repositories like EPEL (Extra Packages for Enterprise Linux) which is derived from the Fedora build system and ELrepo (Enterprise Linux repository) offer us mostly what we need for those extra goodies. There’s also the IUS repository (there are some conflicts here with certain web frameworks), Nux Dextop, and Software Collections (scl, synonymous with Red Hats developer toolkit’s for getting newer packages...

In my eyes, GLVND isn’t that big of a thing as I don’t know many people who want or try to use multiple mixed vendor cards from AMD/NVIDIA in their systems. Particularly for OpenGL unless you're driving monitors with multiple cards. For me something like that for OpenCL would be preferred...

To your other question, DE stands for Desktop Environment. It’s the collection of applications/UI that you use. Some examples are GNOME, KDE, MATE, Cinnamon, LXDE, XFCE, etc. RHEL/CentOS/Fedora/Ubuntu use GNOME as their default DE, while other flavors or spins might choose another. Over in Ubuntu land, there’s Kubuntu, Lubuntu, Xubuntu which are normal Ubuntu with different DE’s by default.

Cheers,
Mike

AWESOME insights, Mike! Thank you and thanks to every other of the helpful contributors here. This forum rocks.
-Pete

Sent from my P00I using Tapatalk

Sat Jun 23, 2018 11:23 pm

Peter Benson wrote:AWESOME insights, Mike! Thank you and thanks to every other of the helpful contributors here.

yes, it's indeed very useful to read divergent opinions about this topic!

nevertheless i wouldn't agree, if someone claims, that other more popular distributions aren't just as stable and reliable as CentOS...

Sun Jun 24, 2018 12:50 am

Martin Schitter wrote:nevertheless i wouldn't agree, if someone claims, that other more popular distributions aren't just as stable and reliable as CentOS...

I'll respond here, but I think this discussion belongs in the general Linux thread, not this one.

I don't think anyone here is calling any other distribution not as stable and reliable. If you're interpreting my words as such, you're looking much to deeply and between lines that don't exist. There's just a distinct difference in the philosophies and target audiences of the distros. Ubuntu (which is the most commonly stated distribution here) is a great distribution and holds its own. It's very stable and is a crown of the web and smaller team science/research fields. Do we see it very often in this industry? Not really. That's mostly to do with history and choices made many years ago. It's difficult to change an entire infrastructure and toolchain that's preexisting.

The point about the reproducible builds is cool, but it's very specific to the toolchain in use. Unless BMD is using Docker or VM's and self builds every tool (GCC, X11, Qt, etc) on their own, it's pretty much moot.

I said it at the top of this post, but if you want to discuss this, let's move it to the other thread.

Cheers,
Mike

Sun Jun 24, 2018 1:27 am

MikeRochefort wrote:Ubuntu (which is the most commonly stated distribution here) is a great distribution and holds its own. It's very stable and is a crown of the web and science/research fields. Do we see it very often in this industry? Not really. That's mostly to do with history and choices made many years ago. It's difficult to change an entire infrastructure and toolchain that's preexisting.

yes -- i also think, it's more a remnant from the past, but honestly i do not see any serious reasons, why clients in the industries shouldn't be attracted by just the same advantages, which made some alternatives much more popular/important in the meanwhile. nevertheless i also don't want to convince any actual CentOS user to change his habits. we just shouldn't irritate less experienced linux users by biased [mis-]information.

MikeRochefort wrote:I'll respond here, but I think this discussion belongs in the general Linux thread, not this one.
...
if you want to discuss this, let's move it to the other thread.

why should we change the thread? -- i also just wrote about the specific disadvantages of CentOS (e.g. the GLVND troubles and similar consequences of outdated software), which make it a very problematic choice for most linux users in practice.

Sun Jun 24, 2018 2:27 am

Sorry for highjacking the thread. Could you please point me to the reason why I cannot create a dual-boot installation?

I install CentOS 7 using this thread as a guide. It is running fine. I then repeat the same steps on the second SATA SSD and it is booting and running fine too, however I cannot chose between the disks to boot from neither in bios nor the bootloader. I can mount the partitions of the older installations and they look fine. I can even boot from the older installation if I physically remove the second SSD. I tried both MBR and GPT - no difference.

Sun Jun 24, 2018 4:07 am

Martin Schitter wrote:why should we change the thread?

This thread is for help regarding installation on CentOS. The comment I quoted was your statement about stability vs other distributions. They are not the same discussion. This is not a thread for general Linux discussion. It's easy to go off topic, but let's restrain ourselves from that. I have other comments I can make but I will not write them here.

As to your comments about GLVND (and determining state), I suggest you read NVIDIA's official statement: https://devtalk.nvidia.com/default/topi ... nt=5145218 as well as my statement above where I explain why it isn't in RHEL/CentOS by default as part of the core distribution. This could quite possibly change with the release of RHEL 8. Also look into the recent release of XOrg 1.20, which contained some great improvements to GLVND support for server side usage which is very much in the wheelhouse of Red Hat.

Misha Aranyshev wrote:Sorry for highjacking the thread. Could you please point me to the reason why I cannot create a dual-boot installation?

Are you trying to run two CentOS disks? If so, they share a common bootloader, so they won't show up separately. If you wanted them to show up differently in GRUB you would need to boot into one drive and make a custom edit to the GRUB menu pointing to the other drive.

Cheers,
Mike

Sun Jun 24, 2018 5:32 am

Yes, I'm trying to run two CentOS disks. One for Resolve 14, another for 15.

Sun Jun 24, 2018 11:00 am

Mon Jun 25, 2018 6:14 am

Karel Voners wrote:However when going to the media browser to import media it crashes when hitting R3D media folders. It also does not give a thumbnail preview. Any ideas what could be wrong?

If you're on the last couple of betas, you will get a crash reporter dialog.

If not, you can use these steps to create a crash archive zip file on your desktop - viewtopic.php?f=21&t=41211 - and create a new post here with the URL. It might be worth moving it out of the installation thread for visibility.

In either case, please add some context (steps to repo, which file types work, which don't, which crash, what action does it crash on etc). An uploaded sample file would be useful too.

For files and URLs, viewtopic.php?f=21&t=69837 has this:

When reporting your issue, be as detailed as you can with steps to reproduce the problem. Post online a screen shot or phone video of the issue if that helps explain.

If you don’t have permission to post images yet, you can also place these in shared download storage locations, Dropbox, OneDrive, Google Drive etc. and provide a link so that other users can see your description and the video.

As the forum does not permit paths to external sites, place a space in the path name so it doesn't look like a path to the forum but its human readable.

Tue Jul 03, 2018 10:35 am

MikeRochefort wrote:I said it at the top of this post, but if you want to discuss this, let's move it to the other thread.

Cheers,
Mike

You are doing a great job herein, Mike, but...
Which "other" thread should such pertinent Linux distro-variant questions be moved to though -- would you or someone in-the-know kindly link us to it here, please?

[size=80]ResolveStudio 14.3...014 | MiniMonitor | DTV 10.9.7 | "Micro$haft WindBlows" 8.1 x64 | ROG G751JL, 2.8GHz Intel i7HQ, 24GB DDR4, 1TB HD, 500GB EVO 850, 2GB GTX965M

Who is online