Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2026 Daz Productions Inc. All Rights Reserved.
Comments
Hi all, I was wanting to get back into installing DAZ on a Linux system (cachy would be nice), but I haven't been in the trenches for some and have lost track of what works why and how. I tried to find my way around this thread but keep sliding down too specific rabbit holes. Is there a place someone can point me to that has the basic Dos and Don'ts and maybe some guiding for installation? That would be so cool.
With 2K comments this dicussion could better be located into "3rd party software" together with Max Maya etc. as it is evident there's a strong community interest.
@tsroemi
The most important thing is to pick an OS that you are comfortable using. It is very easy to try out any number of Linux distros before you settle on one but actually giving advice on how to install the OS is a little trickier due to the variable nature of peoples' hardware and setup. Most modern mainstream distros have a very easy to follow step-by-step process. Many will offer Nvidia drivers as part of the setup procedure which is one less thing to worry about.
If you are new to Linux I would suggest you use an OS that is widely adopted and has an active community rather than something niche and odd. Linux Mint, Ubuntu, Fedora to name but a few. CachyOS is the easiest of the Arch-based distros to work with, but it's still Arch. It really doesn't matter which one you pick as both Daz Studio 4.x and 6.x can be installed on basically all of them. The only distros I would consider avoiding would be the immutable Cosmic series of operating systems, such as Bazzite, which can prove problematic. They are great for people who just want Linux to work and basically be indestructable. However, they are not so good if you need to start tampering with some of the core system files which you will have to do, unless you want to go down a dark, dark, very dark rabbit hole of technical hooliganism.
I highly recommend you use a separate drive if you plan to keep your Windows OS but at the very least you should install Linux on a separate partition. Don't try to install it over the top of Windows!
Have a think about what you plan to do with your Daz content. Linux file systems are case sensitive which will cause problems with some of your items: 'DAZ 3D' and 'Daz 3D' are two different directories as far as Linux is concerned, the same for 'textures' and 'Textures'. Linux OS support for NTFS drives is somewhat limited but it does render the above issue moot. However, in my experience, using Linux to write to NTFS drives for any period of time ends up with them needing to be fixed using CHKDSK or similar from within Windows.
So... if you plan to keep Windows, consider reformatting your content drives/partitions to something like exFAT which is usable by both Windows and Linux without issues. This could be quite inconvenient but it's less inconvenient than constantly repairing your drives. Or, keep your drives as NTFS but only write to them using Windows.
If you plan to ditch Windows completely, consider using ext4 for your data drives/partitions. An ext4 partition can be set to case insensitive so it behaves like a Windows drive, but this can only be done during the formatting process so you can't add this option retrospectively to drives with existing content.
Installing Daz Studio is possibly the easiest part. There are many ways to do this: One person here has installed it through Steam with a fairly clever hack to enable dForce, some people have used system WINE and some have used Bottles and Lutris (both WINE managers). Personally, I recommend Lutris due to the ease of setting things up and then making changes later if you need to. It's by far the simplest option, but the choice is obviously yours. Have a look back through this thread to see a how-to for each of them.
Finally, for reasons no-one yet really understands, Debian-based distros lose the ability to run the built-in Nvidia denoiser. If you absolutely must have this available, simply avoid Debian distros. One member here using a niche Debian OS, Tuxedo, brute forced it with symbolic links which briefly enabled dForce and then it promptly destroyed his entire Daz WINE setup. Hard to recommend this route :)
Have a think about what you want to do and if you have any problems or questions, let us know.
@Padone
Completely agree. I have no idea who to ask though.
@TimberWolf I suppose any moderator here can consider the request, or at least let us know if it is possible. To move this into "3rd Party Software".
@TimberWolf, thank you so much for your post. I'm not completely new to Linux fortunately, I've had Mint on my writing laptop for quite some time and am happy with it, if a little bored. I got Cachy on a magazine DVD coincidently and tried it out on a 2017 iMac that I keep wanting to use for CG work since it has this beautiful display, but keep running into problems everywhere - and the system was SO fast and pretty, it just amazed me, even starting from DVD. Then I read that many arch distros play nice with nVidia, and this got me started to look further into things.
I've decided on Manjaro instead of Cachy now because I don't think I can be dealing with such a rapid update system, especially with a fickle DAZ environment installed. Manjaro installed quite nicely except for the wifi which I saw on the Cachy site is a known problem, and I managed to fix it. Sound is still missing, but low on my priority list for now. Everything else just works. Well almost everything ...
The one thing that proves really hard is, of course, the egpu with the nVidia inside. I've gotten the system so far as to recognize and authorize the enclosure, it also seems to see the card as well as the driver - but it somehow can't connect the two yet. Looking further into things, I can also see that it will give me that annoying 'this device has not enough resources' message, because for some reason, it's multiplied the one bridge of the enclosure into many, and the one where the card sits has 0 resources allocated. I've had that error with the egpu on the original Mac system as well (installed bootcamp then, but I'm really trying to move away from Windows more).
Looking at online resources, I'm really really confused by now. Some say I should use the open source drivers, some say I absolutely shouldn't. I think I will leave the egpu alone for a moment and try and install DS first, does that make sense? Then I could at least use it in non-GPU mode already. I was wanting to use this guide:https://gist.github.com/llaurama/76c930a3f74741f83dd6453a22f411e1 . But I've mostly been a graphical UI Linux user so far, and just sloooowly starting to work my way more into the depths of things (because the darn egpu forces me to). It does seem to make sense to me though. Or am I overlooking something?
Would much appreciate any comments and ideas!
Edit: Yes, dual boot - I've kept macOS on the iMac and installed Manjaro into the Windows bootcamp partition. The iMac still thinks it's windows because the boot manager is still hanging around, which is kind of funny, but the systems seem to work really well side by side. I think on a Mac, I wouldn't ever ditch the macOS completely.
@tsroemi
I've had a look through that installation guide and I can't see any problems with it. Omit the Valentina CMS step though - that is genuinely unnecessary. I think that process could be cut down significantly but as I have zero experience of trying to do any of this with the combination of Mac+Manjaro I will defer to the author of the guide. There are a lot of belt-and-braces steps in there (the CMS installation looks particularly over-cooked to me) that complicate the procedure but there's nothing wrong with it per se - it will work. However, what can be skipped on a dual-boot Windows PC might not be possible on a dual-boot Mac. I simply don't know. My experience of Macs can be boiled down to compiling Unity C# code on it for IOS and then running away from it as fast as possible!
I also have zero experience of eGPUs with Daz software, let alone trying to get them to run in Linux. What I can say is that with internal GPUs you should be using the closed-source proprietary drivers provided by the distro where available, not open-source or those available on the Nvidia website, but other than that I can't really offer any help.
A quick look on the Arch wiki suggests enabling your eGPU could be a fairly complex operation - https://wiki.archlinux.org/title/External_GPU
Which desktop did you choose? Manjaro offers desktops that use the older X11 (Xorg) protocol and the newer Wayland protocol; which you have will determine the approach you take to solving this.
On a more upbeat note I am sure we can fix your sound issues. One thing with Macs that makes troubleshooting them a little easier is a fixed set of hardware so it shouldn't be too tricky to get a Pipewire installation sorted out but this is definitely at the bottom of the to-do list.
@TimberWolf, again, many thanks for your take! I'm using the standard KDE Plasma that came with Manjaro. I thought that Xorg was the safe option since the iMac is so old, and was gonna go with that; I wasn't even aware that KDE has been switched to Wayland. I've been so out of the loop with everything, and I never had that much linux know-how to begin with ... This might not be a bad thing though, since the actual card I plan on using is an RTX 3060, so not decades old. I should be good with Wayland then, generally speaking, no?
The hilarious thing is that this is probably (at least part of) the reason why the card that I'm using for tinkering atm (the 3060 hasn't arrived yet) is not working. Because it's a GTX 1060, so what was that, Pascal? Way old in any way. No wonder I had no luck ...
Once I can put the 3060 in its Sonnet box, I'll probably be shot straight to the 'no resources' error which is what I got with a 5070 on the macOS itself. At least I'll be spared the driver problem hunting then I guess. In any case, you already did help me out majorly here, thanks!
The egpu thing, yeah I know, it's always a horror and will always be, no matter what the distros promise. But some egpu sites now seem pretty confident that Linux is not such a dead-in-the-water bet for egpus as it used to be, so I'm willing to give it my all this time. I'm so sick of being made to buy new electronics all the time. I have this lovely piece of hardware and I mean to use it as long as it will let me. Well, if it will let me, that is ...
Concerning sound, it would be nice to have it back at some point. Pipewire is set up, volume controls on the desktop are there and linked to the Audio Pro just fine, and they are even reacting. But still no sound. I'll absolutely get back to this once I figured the graphics stuff out and ask you to help me out with this as well if I may.
As for now, I think I'll try and set up DS tonight with the Wayland routine from the tut and see how that goes. I'll just not connect any card yet and leave out the Nvidia-related parts of the routine.
Purely anecdotally, that was my experience when I was running Ubuntu, but went away when I switched to Mint. After mature reflection, this may have been because Ubuntu would sometimes freeze and need to be power-cycled; and it was my impression that power-cycling (or resetting) Linux while it has an NTFS volume mounted is a lot more likely to cause problems than doing the same thing under Windows. In fact on more than one occasion the disk was left in such a state that CHKDSK froze while trying to fix it.
To be fair, that Ubuntu set-up was my first Linux daily driver, and I very likely broke something while learning the ropes.
@tsroemi
I am sure you have already found this, but on the off-chance you haven't, there is a script to enable Thunderbolt eGPUs specifically for the KDE Plasma desktop available on the egpu.io forums. Although the initial thread is a few years old, the code appears current:
https://egpu.io/forums/thunderbolt-linux-setup/all-ways-egpu-script-for-wayland-linux-desktops/
Repo here:
https://github.com/ewagner12/all-ways-egpu?tab=readme-ov-file#installation
I wish you the best of luck with this but I suspect you've got a fight on your hands...
@English Bob
My issues stemmed from using both Windows and Linux to write to the drive. Linux (both Fedora and CachyOS) would eventually render the file index unreadable by Windows which then needed a rebuild. Linux NTFS support is about as barebones as it gets and I personally wouldn't recommend it for drives that are used to store data. If you have executables that need to run natively in both Windows and WINE under Linux then the file system will have to be NTFS but exFAT is a safer choice for a drive or partition used for Daz content imho. I have played around with Windows ext4 and btrfs drivers but none that I tried could cope with anything other than a basic installation of the file system. I have never managed to get a casefold-enabled ext4 drive recognised by Windows.
@TimberWolf, the enclosure itself is already authorized, that seems to have worked!
Edit: Ah sorry, read too quickly, the link as well. I was thinking that I would only use the egpu for specific programs though, for DS and Blender and Davinci mainly. So I'm hoping to set it up specifically like this, not as the main primary video. Especially since the mac refuses to boot with it plugged in. Anyways, thanks again for the link and for the wishes, am going to need them sorely probably
@tsroemi
So essentially you want to use the iMac's Radeon card as the display GPU and the Nvidia eGPU as a compute device. Hmm. You should be able to get your Manjaro OS to boot with the eGPU attached but I am not convinced you will be able to get the exact setup you want. I would suggest just trying to get it to work in whatever configuration the script will allow, certainly initially, before attempting anything more ambitious.
Your sound problems stem from the fact that there is no driver built into the Linux kernel for the Cirrus Logic CS8409 processor used by your Mac - you'll have to add the driver manually but it's fairly straightforward as the drivers are stable and have been in the Arch AUR repositories for a few years now. I was a bit surprised by this as most common sound and networking hardware is covered by the kernel these days but it's a straightforward trip into the console to get it working.
@TimberWolf, my little old iMac is even more basic than that, it doesn't have a dGPU at all, just the Iris chip ...
Yes of course, I'll take basically any working configuration the system will give me, as long as DS is somehow part of the mix. But the problem with the Mac not starting properly at all with the egpu attached runs deeper than Manjaro, because the macOS won't start with it either. I seem to recall that last time I tried, the screen went black right after the Mac icon, so before I even got to the boot loader for the OS's. I'm not keen on repeating that ... I couldn't try anything with the Mac at all today for lack of time unfortunately, but I'm hoping for tomorrow.
Cool about the sound, that really should be an easy fix. I didn't imagine it would be something so basic!
@TimberWolf, can report first success, DAZ Studio (and DIM) installed without much hassle and now up and running. YES!
What's been really annoying in the process is that Manjaro can't read my hitting the ALT key, and so I have to copy @ and such into Konsole and everywhere else, sigh. I'm reading online that it's something that needs to be configured in Plasma which I think is just plain weird, but oh well. At first I thought it was the Mac keyboard's fault, but it's the same on a generic Dell one. I tried to configure a new shortcut in SystemSettings but have a hard time atm finding my way around in there, so I'll just leave it as is for now. Unless you'd happen to know a really easy fix, of course
.
Next up, I'll just play around a bit and wait for the 3060 to arrive an Friday. That's when the real fun will begin I reckon ...
@tsroemi
You haven't specified what layout yout keyboard is but I'm slightly confused as to why you would need the ALT key to generate an @ symbol on a generic Dell model (SHIFT 2 for a US layout ??) but my first port of call would be to check that the OS is using the keyboard layout you expect it to, i.e. check the internal OS mapping matches the physical layout of your actual keyboard. It's never the easy answer (!) so I'm going to guess that it is set up correctly but it's worth a look. In the console:
localectl status
I do suggest bookmarking the Arch wiki - have a read through this as it may offer a solution:
https://wiki.archlinux.org/title/Linux_console/Keyboard_configuration
In the Manjaro UI, under keyboard settings, also check the keyboard model matches - the Dell should be set to one of the generic keyboards - 105 key, for example, if you have a full number pad.
---
I highly recommend you find and join a Manjaro forum you like the look of and appears active. Because we're both running Arch-based distros I can offer some help but it's inevitably going to be limited if you have to dive into the weeds to solve an issue. OS-specific Linux forums have a bit of an undeserved reputation for being brutal to people who ask beginners' questions but I think this is a stereotype. You will generally find most people are more than willing to help a new convert to their religion and most folks using them will know far more than I do :)
Great that you got Studio et.al. installed but, as always with unconventional hardware setups, this is going to get very technical, very quickly.
@TimberWolf, I'm so sorry if it seems that I monopolized you, I really didn't mean to! Of course I'm sifting through the forums as well, and will certainly join one once I got a feel for them. I guess it's just kind of natural for me to turn to the DAZ forums first and foremost (which is really very strange in matters of Linux and such, come to think of it ...). But I've even got myself a heftfy Linux compendium yesterday, so I should be starting to find my own way around more soon in any case. Well, hopefully, that is ...
Concerning the keyboard, this is probably a specialty of German keyboards - the @ is an alternative to Q there (and actually achieved via Alt Gr - the right side Alt I mean -, not Alt - was too hasty there). I did tell the system that I'm using a German keyboard, but it doesn't seem to have stuck. I''ll just put it on my Check and Fix Later list for now, as the trick with the text file works pretty well. I do need the @ so I can log into DIM to download, and what's more, I need the tilde for the terminal of course. This, on the German layout, is an alternative to * and +, also reached via Alt Gr, so.
DS actually works really well so far, I have installed a couple of things by now and done some renders and draws. No problems. CMS / smart content also seem fine. So as far as basic installation goes, Manjaro seems like a decent enough home for DS. Now for the NVidia endgame tomorrow ...
@tsroemi
It's not about my time - it's about my lack of specific knowledge of your distro! The keyboard settings should be common across all Arch distros, though, so I added a German keyboard to my settings, switched to it and ALT GR + Q did indeed produce an '@'. Had to pull up a German keyboard image to change the settings back as Linux does not know what a 'kezboard' is :) All WINE applications functioned as though my keyboard had a German layout.
Does the keyboard work correctly outside Daz Studio?
Edit: There is one simple change you can try. Start Lutris and from the hamburger menu at the top right, select Preferences > Global options and scroll down until you find the drop-down box for Locale. Set this to German rather than System (and remember to hit the save button) as you might have a mis-matched locale/language/keyboard in the OS itself. With this change, even with my locale and keyboard set to the UK, my keyboard behaved as if it had a German layout.
@TimberWolf, just a quick update, I was down with the stomach flu over the weekend and so couldn't work at all on the Mac, but I've got some spare time today and already managed to get the og apple keyboard working properly. The whole keyboard situation was a total mess, different values in different places. But when I switched back from the Dell to the Magic Keyboard, Manjaro detected it properly for the first time, and then with selecting 'apple aluminium iso' as the model and 'german macintosh' as the layout, it now works perfectly. So I'll now brace myself and get into the nitty gritty with the egpu.
Should I maybe just post my next updates here without @-ing you? I don't want to make this seem as if you're my personal arch assistant, you know ...
Okay, so here's the first day of trying to get the egpu to run for DS. Maybe it'll be something of interest for others playing with DS and Linux as well.
In the beginning, the egpu box (sonnet breakaway) was recognized and authorized right away, but that was about the only thing that worked out of the box. It took me most of the day and much help from the internet plus an LLM for most of the terminal commands (because I really am such a Linux noob) to work through various layers of problems.
The first big cluster was that the box itself was there (its thunderbolt bridge), but the nvidia card wasn't detected at any point in the system. (I should add that I cannot boot with the egpu because the imac runs into graphic glitches when I try, so I can only ever hotplug.) From some earlier adventures with another egpu, I had already added pci=realloc=on, pci=assign-busses and pci=hpiosize=268435456 to grub, but that wasn't enough. Things were complicated by the egpu somehow messing up the system as a whole and kicking out my wifi connection so that I had to google and chat with the LLM on an iPad and had to type out the myriad of things it wanted to know by hand, which took ages. Eventually, sudo dmesg reported that there was no bus number available for hot-added bridge. What didn't work to fix this: pci=busnr=128 or pci=nocrs added to grub. What did: pci=hpbussize=64 (also added to grub). Suddenly, in lspci -nn the gpu appeared in all its glory.
But of course, that was only the beginning. I installed propietary Nvidia drivers via mhwd (video-nvidia), but nvidia-smi kept telling me it couldn't communicate with the driver. Modinfo nvidia saw the nvidia module, but lsmod saw nothing at all. Dmesg then said nvrm: no nvidia gpu found, at which point I felt like dumping the whole kaboodle out of the window and be done with it. But I had already wasted most of my day, so ... Checking the actual address of the gpu, there was only one region listed (region 0, the small one with 16 MB), but no other one, especially not the large address one the nvidia was probably desperately searching for. Dmesg for nvrm said the assigned region was invalid (bar1 is 0M @ 0x0).
At this point, I was completely lost at sea and reliant on Chat to tell me where to go from there, and he wanted me to try adding pci=big_root_window to grub. But since he sounded like he was rather out of ideas as well, I rooted around in my saved info on the matter and rememered this article:https://cervesato.it/posts/egpu-thunderbolt-nvidia-bar-fix/ which seemed to me to deal with the exact problem: box present and authorized, gpu present and working, driver installed, yet no resources given to to poor thingie by the system. I had also noted earlier that there wasn't just one thunderbolt bridge enumerates but 8 or 10, which seemed quite a lot for 1 egpu with 1 card. This article said that for some reason, the system created lots of bridges for the egpu and allocated memory to them, and when poor nvidia came around asking for some of its own, it was turned away and never dared to ask again. This really seems to be the main problem.
There was a script included with the article, and I was so tired I basically just copied and ran it to see what it would do. It's meant to break down the whole bridge architecture, delete all the superfluous bridges, and then recreate it properly and allocate memory to the nvidia gpu as far as I understand it. It did run but quickly stopped dead because it could destroy but not rebuild the structure because it could not find the root port to rebuild from.
And this is where I stand now because I was just too tired to go any further from there. I think I'll try and take the matter to egpu.io or reach the author of the script somehow, because i'm really so far out of my depth now. Will post further developments. So that's it for today i guess, folks.
Evening All,
Just wondering if I can call upon the collective experiences here and ask whether it's actually worth moving from Win11 to Ubuntu or similar for running my main PC as a Daz box almost exclusively (maybe hosting a code server or a few other small services)? Oh and use it as a local LLM box with the two GPUs pooled when not rendering
The reason I ask is that am not completely new to Linux and eGPUs but have gone for a bigger case and motherboard to allow me to run multiple GPUs (5080 and 4070 SUPER) with a Ryzen 5950X CPU so generally performance is ok but am still hitting VRAM limits and the sheer number of pointless background "apps" and "services" I never asked for growing out of all reasonable need. Now I know I can't do much about VRAM other than maybe employ optimisation techniques and tools within Daz but is there any benefit (from a Daz perspective) to migrating fully over to Linux and freeing up 100GB+ of SSD space from all the junk Windows seems incapable of doing much about. I'm fed up of manually hunting around for caches and temp files that Windows apps don't clean up and would actively consider going full Linux. Have built Ubuntu, Fedora and PopOS boxes in the past so handy enough with the terminal but would rather not have to spend hours rewriting scripts to make it see two GPUs
I have built my own iRay server boxes in the past but get only caching errors these days and Render queue seems really temperamental. In an ideal world I would have a big box that hosts either a queue or a local iRay server that I can activate when I want (overnight?) and not faff with caches on other PCs and RAM/VRAM limits.
I have a separate NUC that has Windows on it that only runs Jellyfin and nothing else at present. Could be tempted to migrate to Linux and PiHole etc as have done that before but got fed up monitoring and updating everything every time I rebuilt networks etc... I have a massive Steam library but get no time to game, and have GIMP for image editing as required, plus a macbook pro for later image editing etc. I tried the NUC with an eGPU on Linux previously running iRay server but even with 32GB RAM and a 4070 SUPER in a Razer Core X enclosure (and working!!!!) it still crashed after anything other than a simple scene due to cache overloading... never managed to work out why but I was rendering scenes with 2-5 G9 characters, emissives etc (yes I know I should have bought a 5090 instead of the 5080 when I had the chance)
So short version - if I already have a 1TB+ Daz Content Library and am rapidly running out of space thanks to Winjunk, should I bite the bullet and reformat or get a new SSD and move direct to Linux? If so, which distro works best with Daz please?
@jonathan.torbitt
I can't really say whether or not you, personally, should switch to Linux but if I address some of your points hopefully that will make your decision a little simpler:
1 - Studio is obviously a native Windows executable however I can't tell the difference on CachyOS (Arch) between Windows 10/11 and WINE/Linux from a performance standpoint. It was the same with Nobara (Fedora). You won't gain anything but, equally, you won't lose anything. There really isn't one standout distro and/or desktop to choose. The only distros I would avoid are any that are immutable as if you need to make any changes at a system level (which you well might) then it becomes quite technical.
2 - Any reasonably modern distro will run Daz Studio and after installing one small package from Github iRay will work out of the box, as will dForce, with your GPUs. No terminal-based heroics required.
3 - All Debian-based distros we've tried collectively will not run Studio's built-in denoiser. It's an Optix issue that nobody has yet figured out. This would include Ubuntu. All Fedora and Arch distros will allow it to be used.
4 - LLMs and most python-based AI run significantly quicker in Linux distros compared to Windows.
5 - Consider what file format you intend to use for your Daz content. Linux support for NTFS is weak and if you dump Windows completely you will lose access to any disk repair tools that work with NTFS. Linux has a habit of making a bit of a mess of NTFS file indexes in my experience. This could be a bit of a pain depending on how you've got things set up at the moment. Also remember that Linux file formats are case sensitive which will need to be addressed if you decide to use them. I use ext4 with the casefold option enabled to make it case-insensitive but this needs to be switched on during formatting and can't be applied as an afterthought.
6 - Linux is considerably lighter on resources generally. I've just pulled up the system monitor - the background processes are using 567Mb of RAM, 0.3% of the CPU, and I don't have a particularly slimline installation. Disk installation size will obviously vary from distro to distro but most offer a fairly barebones setup to minimise the amount of disk space they initially take up if that's what you want; a fraction of the space of Windows, Win11 in particular.
Personally I switched to Linux because I absolutely cannot stand the current Windows ethos. I was fed up with the decade-old bugs, the foisting of unremovable software I neither wanted nor needed, the intrusive and constant data collection, updates I had no control over, etc., etc. I do, however, keep a small Windows 10 installation on its own SSD with no internet access as a backup OS in case something goes horribly wrong with Arch, and for the odd bit of software that just will not run under Linux.
If your main reason for switching is due to disk space issues then, yes, any Linux distro will be lighter but it might be easier just to get a bigger disk! If there are other reasons for switching as well you shouldn't run into any issues.
@tsroemi
Makes The Odyssey look like a stroll down to a local shop! Unfortunately you're dealing with hardware that I'm pretty sure no one else here has ever used or has access to so I think you're absolutely right to take it to a forum where people really do know what they're talking about. Half the battle is knowing which questions to ask and you clearly do - anything I or anyone else could offer is something you are perfectly capable of discovering yourself so, yes, take this to the experts.
Issues with Daz under Linux (and there are potentially many!) we can help with. Thunderbolt-linked eGPUs on a gently ageing Mac with Manjaro... not so much!
Viel Erfolg!
@TimberWolf, dankeschön!
I'll probably be back at the latest once I try to get dForce to run on my set up ...
You guys won't believe this, I hardly believe it myself: the RTX is actually running on my weird iMac / Manjaro / eGPU set up! And DAZ Studio sees it. YES!!!
The actual rendering doesn't quite work yet, because when i tried with the DAZ cat, it crashed and then Lutris told me it had been shut down by the kernel because the system was running out of memory - wait, but that could have been because of the dForce cat hair, maybe? I think I read somewhere that dForce needs extra steps. Regardless, I'm in the game! So next I'll pester everyone on how to get the rendering and dForce actually working.
@tsroemi
That is quite the achievement. It's such an unusual combination of hardware and software that I honestly didn't rate your chances of success but hats off!
Onwards. Linux will shut down a process before it consumes the entire system memory and brings down the whole OS. I have lost count of the number of times Blender has disappeared with the very same message when I've got a bit too ambitious. With an RTX 3060 (I'm assuming it's the 12GB model) you'll really need at least 32GB of RAM to use it to its full potential. You might just get away with 16GB if you avoid using items such as curves/strand hair which use a very large amount of RAM during the process of sending the scene to VRAM. Google tells me your iMac only came with 8GB and 16GB options - is this correct?
First thing I would do is render a primitve - create a cube, give it a colour and some gloss - render it. Make sure it is actually working.
I'm assuming you have installed the modified CUDA libraries from here - https://github.com/SveSop/nvidia-libs
If so, it really should all just work, both iRay and dForce. Check the hardware configuration in both Render and Simulation Settings. You should see your device listed exactly as it's named - RTX 3060. If you have any Blender experience you can also check it with the native Linux version which doesn't need any further intervention.
@TimberWolf, thanks for the appreciation! It's mainly a combination of boot params that made the whole thing work, basicall forcing the allocation of enough resources to the card. Maybe the render crash is due to me going a bit overboard in this regard and not leaving enough resources for everybody else in the old iMac to do their job - I'll figure this out by and by. Since I haven't yet installed anything special in relation to DS because I never expected things to work all of a sudden
, I rather suspect this is the reason, though. Those Libs, can I just install them over the ones that are already installed? I also seem to recall something like having to run DS via prime-run - but that is basically taken care of by the Lutris settings now, isn't it? Sorry if sounding a bit daft, it was a really late night ...
The iMac originally only had 8 GB but I had it upgraded, sadly 16 GB was the max it would take. But I work with 16 GB on a laptop with a 2060 just fine in DS, and since the 3060 is in an external enclosure, I figure it will not be able to unleash all of its power anyway. And one single cat shouldn't crash any half way decent system, dforce hair or no.
Edit: DS doesn't see the card name yet, only a CUDA device with a long number. So I really suspect I must work some more on the set up for DS.
@tsroemi
You will need to install the modified CUDA libraries for DS to see your GPU otherwise it will appear as you have described and will be unusable. Just check your system WINE version is >=10 (I am sure it will be) and select System WINE as the runner in Lutris from the Runner Options tab.
Files here - https://github.com/SveSop/nvidia-libs/releases/download/v1.0.2/nvidia-libs-v1.0.2.tar.xz
Installation guide - https://github.com/SveSop/nvidia-libs#how-to-install
You'll find the full path for the WINE prefix Lutris is using by right clicking on the icon/grey rectangle for DS, selecting configure, and then looking inside the Game Options tab.
Once this is complete you should see the correct identification of your GPU in both the Render Settings and Simulation Settings hardware config tab in DS. You will probably also see the original CUDA device which will need to be unchecked. See attached. You could install the NVML libraries which will remove the redundant entry in DS as well but they are not necessary and just add a further layer of complication in my opinion.
When I say 'install', what's actually happening is that symbolic links are being created from within the WINE prefix that point to wherever you've extracted and saved the modified .dll files. I can guarantee that this will work with any conventional internal GPU configuration. What I can't predict are any additional problems relating to your external device but I am mildly confident this will do the trick.
---
I loaded the DAZ Cat into DS and just hit render. RAM useage for DS alone peaked at just over 12GB and this was with no lighting or any other objects, until rendering started. I think we can be fairly certain why your Mac crashed! Bear in mind also that your render would have been run on the CPU only - switching off my GPUs showed a sustained RAM useage of 10GB just for the one model. Do try just the cube for a first test!
@TimberWolf,
ahhh, that makes so much sense now. I just thought DS was actually rendering with the GPU! What's a bit confusing is that nvidia-smi did show me DS running as a process on the card. But I'm guessing now that the card was installed just well enough, so to speak, to do some basic duty but not yet for dforce rendering. Will absolutely take your advice and try with a simple primitive first once I think I've got the set up complete.
Thanks very much for the thorough explanation on the extra libs as well, I'll do that next, and maybe some testing in Blender as well.
What a journey ...