English
Русский
Hosting provider FirstVDS has expanded its GPU Passthrough lineup with a new virtual server plan based on the flagship NVIDIA GeForce RTX 5090 graphics card. The new configuration is aimed at artificial intelligence, machine learning, and high-performance computing workloads, where the term “heavy load” has long ceased to be a figure of speech.
The servers are available with daily billing, allowing customers to deploy GPU resources for short-term projects without committing to long-term rental periods.
A Flagship GPU for AI Workloads
At the heart of the new offering is the NVIDIA RTX 5090, a flagship consumer GPU built on the Blackwell architecture. The card features 32 GB of GDDR7 memory, next-generation Tensor Cores, and support for FP4 computation, a format that is becoming increasingly important for running modern AI models efficiently.
FP4 support enables more effective utilization of computing resources during large language model inference. In practical terms, certain AI workloads can run faster while consuming less memory, and the electricity bill is less likely to develop ambitions of its own.
Image: nvidia.com
According to FirstVDS, the new plan is designed for:
- High-speed LLM inference;
- AI-powered video generation;
- Machine learning and data analysis workloads;
- Other compute-intensive applications.
Plan Specifications
The new offering is marketed under the name VDS-GPU-5090. Its specifications include:
| Parameter | Specification |
|---|---|
| GPU | NVIDIA RTX 5090 |
| CPU | AMD Threadripper PRO (up to 5.3 GHz) |
| vCPU | 4 |
| RAM | 16 GB |
| Storage | 250–500 GB NVMe |
| Additional IP Addresses | Up to 128 |
| Billing Model | Daily |
| Price, per day | From 1,750 RUB ($24) |
The option to assign up to 128 IP addresses is unlikely to be a priority for the average neural network. Infrastructure engineers, however, may view that figure with considerably more interest.
GPU Hosting Continues to Evolve
The introduction of RTX 5090-powered virtual servers reflects a broader trend across the hosting industry. Providers are increasingly adapting their infrastructure offerings to meet the demands of AI development and deployment.
Only a few years ago, rented GPU resources were primarily associated with rendering, scientific computing, and specialized engineering tasks. Today, large language models, image generators, and AI video platforms have become some of the most significant consumers of GPU capacity.