Russian IT vendor "Inferit" from the Softline group of companies has introduced a new Multi-GPU Inferit server designed for artificial intelligence training and inference tasks.
The server supports up to two Intel Xeon or AMD Epyc processors of the latest generation and up to eight Nvidia and AMD video cards, up to 2 TB of RAM and high-speed SSD storage:
- Maximum power — up to 8 kW.
- GPU — Nvidia: 5090, RTX 6000 ADA, L40, L40S, H100, H200, RTX PRO 6000; AMD: W7800, W7900.
- Single-processor configuration based on Intel: Xeon W-2400/2500 & 3400/3500, Xeon Scalable 4th and 5th generation, Xeon 6;
- Single-processor configuration based on AMD: Threadripper PRO 5000WX, 7000WX, Epyc 9004/9005.
- Dual-processor configuration: Intel Xeon Scalable Gen 4/5/6 and AMD Epyc 9004/9005.
The server's cooling system consists of two circuits: liquid cooling and air cooling based on three fans with a flow rate of up to 6200 rpm. The server's power subsystem provides the possibility of installing up to four 2000 W power supplies, which ensures fault tolerance in 4+0, 3+1, 2+2 modes.
The server is also equipped with a remote monitoring system that allows you to analyze the operation of cooling and other components in real time. This is especially important when operating in distributed infrastructures and data centers.
Inferit said:
The Multi-GPU Inferit server is designed specifically for high-performance AI tasks. Thanks to the flexibility of configurations and broad compatibility with modern OSs ("MCBCфера Server" 9, Ubuntu, Windows 10, Windows Server), the solution can be adapted to specific market demands. It can be used in healthcare, where AI is used for diagnostics and analysis of medical images; in finance — for fraud detection and transaction monitoring; in science and creative industries.
The server comes with a 2-year warranty, which can be extended from 3 to 5 years. Customer support is provided by "Inferit" throughout Russia.
