gpu-virtualization 开源项目推荐（22 个）

Project-HAMi /HAMi

Kubernetes 上的异构 GPU 共享

cyberus-technology /virtualbox-kvm

KVM Backend for VirtualBox

Project-HAMi /HAMi-core

A transparent, in-container GPU resource controller that enforces memory and compute limits by intercepting CUDA calls without application or driver changes.

mental2008 /awesome-papers

Here are my personal paper reading notes (including machine learning systems, AI infrastructure, and other interesting stuffs).

NexusGPU /tensor-fusion

Tensor Fusion 是一款最先进的 GPU 虚拟化与池化解决方案，旨在最大化 GPU 集群的利用潜力。

reindertpelsma /nvkvm-pv

Paravirtual NVIDIA GPU for KVM guests — run unmodified CUDA, PyTorch and Vulkan inside a VM at host parity, on a GPU the host keeps using. No passthrough, no vGPU licence. Experimental.

noon-at-cgn /geforce-sriov

Enable SR-IOV vGPU for consumer GeForce cards

bird /vgpu-unlock-blackwell

Research: vGPU unlock on consumer NVIDIA RTX 5090 (Blackwell/GB202). 19 binary patches, full CPU-side pipeline working, GSP firmware blocked by fused-off VF PRIV registers.

reindertpelsma /kayfabe

The GPU your VM believes in. A stock NVIDIA driver runs against fabricated hardware; real compute is forwarded to a host GPU. Rust.

BudEcosystem /GPU-Virt-Bench

A GPU Virtualisation benchmarking tool - For LLM like workloads, overhead evaluations and on resource isolation metrics

NexusGPU /tensor-fusion-site

TensorFusion landing page and product docs

Nium32 /VGPU-MDEV-RTX-3090

NVIDIA vGPU on a consumer GeForce RTX 3090 (GA102) via legacy mdev. Working config, the exact kernel requirement, ~80 measured dead ends, and a tool to port the fix to other driver builds. Fixes the runlist-disable leak behind 'Immediate pteblit timed out' and silent no-render guests.

reindertpelsma /nvkvm-old

Archived - WSL2-style NVIDIA GPU ioctl forwarding for KVM/QEMU guests on Linux hosts - superseded by nvkvm-pv

emirhancabukk /gpu-over-ip-research

Vulkan on a remote GPU over plain TCP — Mesa Venus over the internet, with a full research log

xizha127 /WinTide

Windows apps. Linux flow. Wayland RemoteApps with smooth resizing, shared GPU rendering, and a reproducible libvirt setup.

punith1006 /LaaS-Strata

Sovereign private GPU cloud platform. Slices physical GPUs into 4–8 isolated, 60 FPS in-browser AI workstations streaming full graphical Linux desktops directly to any web browser with persistent storage.

ScaletKlazz /vcuda-hook

Transparent library for CUDA/NVML that virtualizes GPU access, enforces memory limits, and enables container-friendly sharing with minimal performance overhead.

dillanserrano /AmdGpuPassthrough

🎮 Enable GPU passthrough for KVM on full AMD systems, simplifying setup for Arch Linux and Windows 11 virtual machines.

api-evangelist /bitfusion-io

Bitfusion Io — independent third-party profile of a public API surface, by API Evangelist. Bitfusion (bitfusion.io) was an Austin, Texas startup, backed by DCVC, that built "elastic AI" infrastructure software. Its FlexDirect platform was a transparent GPU/CPU virtualization layer that pooled network-attached accelerators (GPUs, FPGAs, ASICs) acros

christian-dussol-cloud-native /hami

Educational repository for HAMi, GPU virtualization and sharing on Kubernetes. Episode 7 of the CNCF Project Focus series.