Rent NVIDIA RTX 3090 GPU by the hour
The NVIDIA RTX 3090 is an Ampere-architecture consumer GPU equipped with 24 GB of VRAM and compute capability 8.6. It delivers about 35.4 TFLOPS of standard single-precision compute and about 805 GB/s of memory bandwidth at a 350 W power rating. You can rent this GPU on ONIE with per-hour billing, full SSH access, and ready templates for PyTorch/CUDA, vLLM, Ollama, and ComfyUI.
Specifications
VRAM per GPU
Architecture
Compute capability
FP32 compute
Memory bandwidth
Power limit
Values reported by the infrastructure provider for machines currently listed; real performance varies by machine.
Available configurations
Live stock and prices, updated continuously. The final price is confirmed before you order.
Loading live prices…
What the RTX 3090 is good for
Local-scale LLM inference
Run quantized or mid-sized language models in 24 GB of VRAM using pre-configured vLLM or Ollama templates.
Image generation workflows
Deploy ComfyUI to generate high-resolution images and run custom diffusion pipelines without hitting memory limits on standard consumer cards.
Model fine-tuning and training
Use the PyTorch/CUDA template to fine-tune neural networks that fit within a 24 GB memory envelope and leverage about 805 GB/s of memory bandwidth.
General CUDA compute
Execute custom workloads compiled for compute capability 8.6 with about 35.4 TFLOPS of single-precision performance.
Is it the right GPU for you?
The RTX 3090 suits developers and researchers who need 24 GB of VRAM for memory-intensive tasks on a consumer-segment Ampere card. Choose a smaller GPU for lightweight inference that requires less memory, or step up to a workstation or datacenter GPU if your models exceed 24 GB of VRAM or require newer architecture features.
Compare with similar GPUs
Frequently asked questions
Is the RTX 3090 suitable for LLM inference and fine-tuning?
Yes, its 24 GB of VRAM and about 805 GB/s of memory bandwidth make it capable of running and fine-tuning many open-weight language models. You can launch vLLM, Ollama, or PyTorch/CUDA templates directly on ONIE to start working immediately.
How does the RTX 3090 compare to smaller or larger GPUs?
With 24 GB of VRAM, the RTX 3090 holds significantly more model weights and larger batch sizes than lower-tier consumer cards. However, if your workload requires more than 24 GB of memory or higher than about 35.4 TFLOPS of single-precision compute, you should consider a newer generation or a datacenter-segment GPU.
How do I connect to my RTX 3090 instance and what software is included?
You can connect directly to your instance via SSH access. ONIE provides ready-to-use templates for PyTorch/CUDA, vLLM, Ollama, and ComfyUI so you do not have to configure the software stack manually.
How does billing work for the RTX 3090 on ONIE?
Instances are billed on a per-hour basis. You only pay for the time your GPU instance is active.
Rent the RTX 3090 by the hour
Create an account, choose a configuration and connect over SSH. Billing stops when you delete the GPU.
Rent RTX 3090