New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Seeking Linux server (dedicated or not) w/ graphics
Hello,
I am seeking a 3d accelerated Linux vps with OpenGL support for a client. Not serious hardware, basic stuff. We'd like to run PyOpenGL and a web application API. More like the kind of hardware from the cheaper cloud Linux VMs like Linode but plus very modest 3d capabilities.
Maybe something like: dual core, good bandwidth, 4gb of ram or more plus the opengl support.
Comments
Updated post to confirm that it doesn't have to be a VM it can be a dedicated server
If you want a dirt cheap dedi with a GPU, try primcast.com, $7/m with $14 setup, or $80/y (iirc) with no setup. It's a low powered AMD processor with an on-chip older-generation Radeon GPU (2 CU's, or in marketing speak 128 "GPU cores"). To put that in perspective, the current high-end discrete Radeon cards have 56 or 64 CU's and I think the new Ryzen APU's have 10 or 12 on-chip. But the cheap thing is intriguing and I have thought of getting one to play with GPU code.
Would a Hetzner auction dedi or Kimsufi with integrated graphics suffice?
It doesn't say anything on their page about the $7/month server offering graphics support. Is there anywhere I can confirm this?
It's built into the CPU, AMD X2150. You can find the specs with a search engine.
I think it should work
https://www.ikoula.com/en/dedicated-server
starts at €14.99 with a GT 710 GFX card.
Important to note is that 15e is just a price for the first month. Then it's 30e.
Does it matter that I'm in the USA? These servers seem to be in Europe. Is it straightforward to make content delivery international?
I've gone with Primcast's Optetron 2150 server in NY for $22. Thanks guys
Hi,
it looks like the AMD Opteron (Kyoto) APU graphics are not compatible with Ubuntu. There are no drivers that exist for it.
Any other recommendations?
Got an update on this. Unfortunately it doesn't work on Ubuntu. AMD / Ubuntu don't have drivers for it.
If I'm in the USA and many users of my application are, how do I serve content closer to them if this is my dedicated server? Does it come pre-offered with CDN or do I need to serve all assets through something like Cloudfront or similar?
Nope, it's probably a bad option for you then if you and your users are in the USA
Dedicated Servers with on-board GPU or basic 3D graphics adapter 512MB-2GB for less than 30€ per month or lower annual. If you need specific model, please contact us directly.
Anyone have a Linux server with an Nvidia GPU (onboard or not) in the USA?
why nvidia GPU? You do realize modern AMD GPUs are actually better suited for shared environment use right? Or have you done no research and just latched onto the first brand you could remember? lol
What does "shared environment" mean? We're using a Linux dedicated server right now and it's taking hundreds of hours to make AMDs drivers work.
We need OpenGL support on Linux and many of AMD's drivers have been deprecated and don't work correctly with Mesa
We're seeking Intel or NVidia. AMD simply doesn't support Linux the way it should.
Hello,
On the off chance that this helps someone working on 3d rendering cloud software like we are, I will leave this information here. We've had no luck with a headless Linux server for 3D rendering. Today we found a host in Atlanta which offers a headless 3d accelerated i5 Mac Mini for $50/month. It is provided with a dongle which enables the GPU and OpenGL in headless mode. The GPU I believe is the Intel Iris HD5000. We're excited to put this server search behind us. I hope this information helps someone else in the community.
Do note that this is probably not acceptable use for any cheap nvidia GPUs like the GTX. You need to be on a Tesla or better to be allowed to use it for this case, or even download the drivers legally.
I expect it wont help at all unless you provide the name of the company you got it from at the very least?
Sorry, the host is here: https://www.macstadium.com/configure/?p=sm1b&c=VGS2.
The dongle they offer to attach to the mac mini is here:
https://www.macrumors.com/2015/11/10/newertech-hdmi-headless-adapter-for-mac-mini/.
I hope the market around Linux servers improves with regard to headless 3d rendering. We lost a lot of time on this one. Luckily this is a passion project and so we aren't too worried about taking a loss on the man hours.
I mean is exists already but more for high end and none budget, its not hard to find a server with a 1080 but its probably going to be $300 p/m
Yeah that's also part of our issue. A $20k application can't be justified with a $300/month server. There needs to be more low/mid end headless enabled hardware. $50/month is much more reasonable.
Also the key is also being UNIX based (mac or linux). Windows Server was not an option for us from the start so a few hosts were excluded.
I did not notice this thread until now, but I have a ton of GPU stuff for ML/AI. I am a bit confused as to what you are trying to do - I have always been able to directly access the GPU 100% from Ubuntu without issues. It definitely works.
Here the mac thing doing some roundabout scenic route stuff where you're doing the equivalent of plugging in a blank VGA cable not attached to anything. I have never needed to do this.
Are you trying to render in chrome or some other third party app you don't control ("web application api"), and then screenshot or save the result from it? What does your stack look like?
just a suggestion, but for your use it might be considering just getting a micro atx desktop adding your own GPU and finding a budget colo service that takes irregular stuff like Joe's DC.
We are simply trying to use direct rendering (hardware acceleration) however we've found that without a monitor plugged into Ubuntu (or faking a virtual frame buffer) Ubuntu refuses to allow us to use OpenGL *with 3d rendering. This may be unique to Ubuntu only, and perhaps we've picked the wrong Linux distro? However based on the fact that this dongle exists for Mac Mini it seems to me that this is a problem across ALL Unix distros and not just Ubuntu.
Thus I would question how you were able to get headless 3d acceleration? Maybe you have an Nvidia GPU? I've heard it's able to be achieved with Nvidia products. On the server we chose we were stuck with an AMD APU and we were not able to get headless 3d acceleration.
The integrated GPUs are generally ass support-wise (can't be passthroughed properly into a VM either). You want a discrete GPU if you want to be able to access it/pull it in as a hardware device.
I run on Ubuntu with Python (Tensorflow); discrete GPUs without anything plugged into them, just connected to PCIE port, show up as devices and are usable at full speed.
AMD RX5xx stuff (separate card, not "APU") has also worked fine for me - these are cheap as hell, and are very performant.
This seems more like a "trying to use integrated gpu" issue. They generally have problems being captured or used outside of "normal desktop use".
Unfortunately we couldn't find a host offering anything but AMD APUs. We found some hosts in Europe offering discrete Nvidia GPUs but we are in the USA.