Blog post

We are excited to announce that Ring Compute has brought two additional 8x RTX 5090 servers online. Both machines are racked, configured, and ready to handle demanding AI and high-performance computing workloads.

What Each Server Can Do

Each server is equipped with eight NVIDIA RTX 5090 GPUs, providing 256 GB of combined VRAM per node. This is enough to run the largest open-source AI models available today, including DeepSeek R1, Llama 3.1 405B, and Mixtral 8x22B at full precision. Whether you need to fine-tune large language models, run inference at scale, or train custom models, these servers deliver the performance required.

The RTX 5090 is built on NVIDIA's Blackwell architecture with significant improvements in AI throughput compared to the previous generation. Combined with NVLink interconnects between all eight GPUs, these servers are well suited for multi-GPU workloads that require fast inter-GPU communication.

Available for Rent on Vast.ai

Both servers are available for rent through Vast.ai, one of the leading GPU cloud marketplaces. You can spin up instances on demand with your preferred Docker image and start working within minutes. Whether you are a researcher, a startup, or an enterprise team, renting GPU time is a cost-effective alternative to purchasing and maintaining your own hardware.

With these two new servers, Ring Compute continues to expand its infrastructure to meet the growing demand for affordable, high-performance GPU computing powered by renewable energy.