ITPOD AI and ML Computing Servers

The main articles are:

• Servers (Global Market)
• Servers (Russian market)

2025

Compatible with vStack HCP version 3.0

ITPOD and vStack have ensured a more complete joint integration of their flagship products. ITPOD server hardware, including AI and ML Computing, is now displayed in the graphical interface of the hyperconverged virtualization platform vStack HCP version 3.0, which facilitates component identification and maintenance. ITPod announced this on December 18, 2025. More here.

Adaptation to the objectives of pharmaceutical R&D centers

The company ITPOD has optimized the line servers with GPUs for pharmaceutical research and development tasks. Equipment configurations will help farm laboratories and R&D centers speed up calculations, analyze large amounts of data and reduce experiment time from weeks to hours with budget savings of up to 55%. The developer announced this on November 6, 2025.

The adapted models - ITPOD SY8108G-D12R-G5 (based on AMD EPYC 9005 "Turin") and ITPOD SY8108G-D12R-G4 (on Intel Xeon Scalable Gen 5) - use up to 8 graphic processors RTX 5090 with Blackwell architecture, 32 GB of video memory and 1.8 TB/s bandwidth. Abandoning NVLink in favor of data exchange over PCIe 5.0 x16 (up to 128 GB/s) made it possible to optimize the architecture for tasks with low inter-card traffic.

Servers allow for molecular modeling, protein structure prediction, faster search for active compounds, and analysis of large biomedical data. The platform supports simulations in GROMACS and OpenMM, working with AlphaFold and ESM models, processing genomic and omix data, and analyzing MRI and CT images. The large amount of video memory makes it possible to use large neural networks without swapping and accelerates model training to predict drug interactions and assess toxicity.

Optimized servers help pharmaceutical companies build digital infrastructure to accelerate drug discovery and develop artificial intelligence-based destinations. This solution combines high performance and energy efficiency with moderate capital expenditures and can become the standard for modern R&D divisions of the industry, - said Ilya Bornyakov, CEO of ITPOD, ITG Corporation.

The architecture of adapted ITPOD server models provides three times the return on investment in AI projects, up to 95% of GPU load efficiency and an eightfold increase in the number of requests processed per second compared to previous generation solutions.

Server view for AI and ML tasks

ITPOD has expanded the offer of servers for artificial intelligence and machine learning tasks to a full-fledged product line. The company offers two main areas of server equipment for various scenarios for the implementation of artificial intelligence technologies from pilot initiatives to industrial deployment of AI applications. The company announced this on July 24, 2025.

For large-scale projects with high performance requirements, ITPOD recommends ITPOD-SY4108G-D12R-G4 models based on processors Intel Xeon Scalable Gen5 and ITPOD-SYR4108G-D12R-G5 processorsAMD EPYC with 9005 "Turin." Both platforms are built according to Nvidia MGX principles and support the installation of up to eight graphics accelerators.

For companies that plan to launch pilot projects or create development and testing benches, a 2U ITPOD-ASR201-S08R server is intended. This model provides a cost-effective solution for experimenting with machine learning technologies without over-investing in infrastructure.

ITPOD servers support graphics accelerators from various manufacturers. The company integrates SOPHGO solutions that show high performance in video analytics tasks and working with large language models. The line also includes NVIDIA universal graphics cards from T4 to H200 NVL.

Particular attention is paid to supporting technologies for combining computing resources. The servers are compatible with NVIDIA H200 NVLink Bridge technology, which allows you to combine four accelerators and get the total amount of GPU memory in the 564GB. The platforms also support Cambricon accelerators with MLU-Link Multi-core Interconnect technology to create alternative high-performance machine learning clusters.

This line of servers is adapted to specific scenarios for using artificial intelligence. We see a growing demand for solutions that allow you to run your own language models in enterprise data centers. ITPOD servers provide the necessary processing power to train and infer neural networks, "said Ilya Bornyakov, General ITPOD, ITG Corporation.