Trick

Hi all, I was wondering if there are some best practices with respect to performance in the docker? I notice some people saying it performs better than windows VM's. For me though, using an RTX3080 and 12 cores pinned from my 7950x performance drops quite significantly. I currently run about 60fps in GTA V max settings, 90fps in World At War max settings and maybe 20fps in Crusader Kings 3 all at 1440p. I recall getting atleast an average of 100fps on GTA V on 1440p ultrawide and atleast 70fps average on CK3 maxed out ultrawide. I have the storage directly set without /user/ as described in earlier posts. I use 8.0-5 in steam to be able to play the games. I also see that usually my GPU is only at about 30-40% load, my CPU also looks like it barely gets used all cores anywhere between 0-20% while running the game. Really hoping there is some way to decrease the gap between my VM and the Docker. That said, do really enjoy this docker and the ability to use my hardware as a cloud gaming server, thanks for making this docker!

Thanks that was it indeed! Been a long time since I set that up, feels dumb that I forgot

So I have a 32 thread cpu in my system however as you can see below, only half of the threads seem to get used. The screenshot is when using jellyfin in docker. At this point I use a GPU for local AI however when I used local AI on my cpu all top half threads were at 100% and the bottom half were the same as below mostly 0%, very few at 1%. I do use the bottom half for my virtual machines, when I use the VM's those threads are used. However they are all turned off (for this screenshot I even turned VMs off as a whole in VM manager). I have checked all my dockers for pinned cpu's as well. I do not understand why this is the case, is there any way to check if these threads are reserved or could there be another reason why they are unused?

I was wondering if it is possible to use gpu's with this docker for example in order to test deep learning code?

Does anyone know how you can set this up with a selfhosted LiteLLM endpoint? In the docs they tell to add it to the docker-compose.yml. I have tried adding a variable (OPENAI_REVERSE_PROXY) to the docker template as well as setting it on some of the other 'proxy' variables in the template but nothing seems to work. My models do not show up in the UI. https://docs.librechat.ai/install/configuration/litellm.html#create-a-configyaml-for-litellm-proxy

A really late responce, but it does really look like the post you've attached, I don't know why but as of late I haven't been seeing any of this pinning anymore. So I guess untill I see it again I won't fix what isn't broken. Thanks for the responces!

Hi Everyone, I am running into a problem with my unraid server. I have noticed off late that my server starts pinning down mostly one thread to 100 % load. Sometimes it switches to another thread but it always keeps one of the threads on 100% load. When I see this happening and I stop all dockers the problem still persists however when i stop my array and start it up again the problem is solved for a little while. I have no user scripts running and it is not doing a parity check. I am running unraid 6.9.2. Does anyone have any idea what could cause this kind of behavior?

Thanks a lot, got it working now. Now I won't have to waste my night not sleeping

Here in the same situation, did you manage to fix this?