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.
is there GPU/Cores virtualization hosting service?
Hello, I know how GPU's are divided on it tiny cores are way different from CPU's so making VM that's share part of it way harder and security hazard.
What I make me think about it, is wen Google created Stadia they show using many AMD Instinct servers GPU of that time, and they talk how they're multiple players are sharing same GPU, ( it like renting it ).
I know what's I'm asking about is totally different subject (and I have no idea if it's something already done or not ) ,
And what kind of virtualization e.g google cloud offer for you for GPU VM, do you can access ans use the whole GPU for your self, or they'll put limitations ( e.g RAM size GPU utilization ).
Comments
In my local environment, I use incus/lxd to share and manage GPU computing resources, but it does not clearly display the VRAM capacity or number of CUDA cores allocated to each user. For team collaboration scenarios, I believe this level of sharing (within containers) is sufficient. In practical commercial applications, while numerous heterogeneous computing platforms supporting whole-card partitioning exist, no truly dominant market leader has emerged—this field is shared among many vendors of varying scales. In the open-source realm, https://github.com/Project-HAMi/HAMi is a notable project.
To be honest, these platforms are quite heavyweight, mostly built on Kubernetes or at least k3s environments. I haven't really considered using them practically since I don't have such demanding needs, though I do come across them in work projects.