3 September 2025
updated at: 18 September 2025
3 min
ITPOD-SY4108G-D12R-G4 is a powerful next-generation IT platform, targeting artificial intelligence, machine learning and large-scale parallel computing tasks in scientific and business environments.
4U form factor - optimized for dense installation and heat dissipation.
Flagship CPUs: 2 x Intel Xeon Scalable 5th generation, TDP up to 350W.
The main "trick" is support for up to 8 GPU gas pedals with direct GPU-CPU connection via PCIe 5.0 x16. This minimizes latency and guarantees maximum data throughput between the CPU and graphics cards.
32 slots of DDR5 RDIMM 5600MHz: fast data sharing, ability to handle huge models.
The memory subsystem is designed to minimize latency, which is critical for deep learning and distributed computing.
12 hot-swappable 2.5"/3.5" universal front slots, SATA, SAS, NVMe support.
2 M.2 NVMe (2280/22110) slots - fast system disk/cache.
Speedy handling of large data volumes plus optional integration with cloud services for scalable storage.
Four power supplies (2200/3200W) with 3+1 or 2+2 redundancy for reliable 24/7 operation even in the event of a unit failure.
Hot swap components without system shutdown.
Maximum 8 full-height double-width GPUs (tested: NVIDIA H100, L40S, A100, A800, Sophgo support).
Additional slots: 2 x PCIe 5.0 x16, 1 x PCIe 4.0 x8.
2 x 10GbE RJ45 network interfaces (Intel X710, optional) for rapid integration into high-speed clusters and data centers.
AI and ML: efficient training and inferencing of large language models, multimodal architectures, real-time analytics of video streams and big data.
HPC: molecular modeling, financial simulations, numerical problems where the power of massively parallel computing is needed.
Hybrid analytics and virtualization: enabled by fast component sharing and cloud integration.
| GPU | Description |
|---|---|
| NVIDIA H100 | For training advanced language models and neural networks |
| NVIDIA L40S | Versatile - suitable for AI and visualization |
| NVIDIA A800/A100 | Industry leader for complex AI/ML tasks |
| Sophgo AI gas pedals | Alternative chips for tasks where independence and power efficiency are important |
Maximum configuration of up to 8 cards of any type connected by high-speed NVLink interconnects.
3 year basic warranty (extendable to 5 years, 24x7).
Spare parts in warehouses in our worldwide locations.
Online support portal, personalized advice on configuration selection, architecture and sustainability/power efficiency requirements.
The platform lays the foundation for GPU Cloud: companies will be able to access the power of top-end gas pedals without large hardware investments.
Flexible scaling from one to eight GPUs for specific tasks with per-minute billing and technical support from our specialists.
ITPOD-SY4108G-D12R-G4 is not just a server, but a strategic tool for scaling AI and ML projects in a competitive global market. It combines modern technologies, flexibility, service and opportunities to realize the most ambitious tasks.
Servers
Servers
Servers
