Каталог оборудования
Каталог GPU-серверов и оборудования для ИИ
Корпоративные GPU-серверы и оборудование для ИИ: NVIDIA DGX, HGX, Blackwell, Hopper и системы OEM-партнёров — по запросу.
NVIDIA DGX
NVIDIA DGX B300
A high-performance enterprise AI supercomputer with 8x NVIDIA B300 SXM GPUs, 2.1 TB of HBM3e GPU memory, and 64 TB/s aggregate GPU memory…
NVIDIA DGX
NVIDIA DGX B200
Unified AI platform for develop-to-deploy pipelines with 8x Blackwell GPUs and 1.4 TB of HBM3e — up to 3x training and 15x inference performance…
NVIDIA DGX
NVIDIA DGX H200
Proven Hopper-generation AI supercomputer with 8x H200 GPUs and 1.1 TB of HBM3e — the workhorse for enterprise LLM training and inference with…
GPU Servers
NVIDIA HGX B300 Servers
OEM HGX B300 baseboard systems from Supermicro, Dell and partner vendors — Blackwell Ultra GPUs and NVLink fabric delivering record intra-node…
GPU Servers
NVIDIA HGX H200 Servers
141 GB HBM3e per GPU — nearly double the capacity of H100, enabling larger models in a single system with widely available lead times.
AI Servers
AI Server 1 GPU
Entry-level AI server with a single NVIDIA L4, L40S or H100 NVL GPU on AMD EPYC — ideal for development, inference pilots, and departmental…
AI Servers
AI Server 4 GPU
Up to 4 NVIDIA GPUs in a compact 2U chassis — maximising GPU density per rack unit for mid-scale AI training and concurrent inference.
AI Servers
AI Server 8 GPU
Up to 8 NVIDIA GPUs including H200 NVL and RTX PRO 6000 Blackwell in a single server, with dual AMD EPYC 9005 processors for balanced CPU + GPU…
NVIDIA MGX
NVIDIA MGX GB200 NVL72
Rack-scale AI supercomputer. All 72 GPUs are interconnected via fifth-gen NVLink, enabling seamless scaling of trillion-parameter model training.
GPU Servers
NVIDIA RTX PRO 6000 Blackwell Server Edition
96 GB of GDDR7 memory per GPU with up to 8 GPUs per server for unprecedented visualisation, rendering and inference power.
GPU Servers
NVIDIA RTX 6000 Ada Servers
48 GB of GDDR6 memory per GPU supports medium-to-large model training and inference workloads with comfortable headroom.
GPU Servers
AMD Instinct MI300X Servers
192 GB of HBM3 memory allows running 70B+ parameter models on a single GPU — reducing inference complexity and cost.
Готовы начать?
Постройте свою AI-инфраструктуру с уверенностью
Свяжитесь с нашей командой корпоративной инфраструктуры. Получите экспертную консультацию, цены на GPU и индивидуальный план внедрения — без каких-либо обязательств.