Moore Threads MCCX D800 384GB
All-in-one server engineered specifically for the high-performance training and inference of Large Language Models (LLMs). It serves as a comprehensive hardware solution for AI data centers to power demanding generative AI workloads
CPUs
2x Xeon Gold 6430 (64 ядра / 128 потоков)
RAM
1ТБ, (16x64 ГБ DDR5 4800MT/c RDIMM)
Accelerators
8x MTT S4000 48GB
Storage
4x3.84ТБ NVMe + 2x 480ГБ SATA
Available
Product status
CPU & Memory
CPU
2x Xeon Gold 6430 (64 ядра / 128 потоков)
RAM
1ТБ, (16x64 ГБ DDR5 4800MT/c RDIMM)
RAM Type
DDR5
Storage Configuration
SSD Model
-
Total Storage
4x3.84ТБ NVMe + 2x 480ГБ SATA
Disk Interface
-
GPU Configuration
GPU Vendor
Moore Threads
GPU Model
MTT S4000 48GB
GPU Count
8
Total VRAM
384ГБ GDDR6
Interconnect
MTLink 240ГБ/с Inter-card P2P connectivity
Network & Virtualization
Network Interfaces
2x 400Гбит/с IB + 4x 25Гбит/с Ethernet
Virtualization (MIG)
Поддерживается (vGPU / SR-IOV)
CUDA Compatibility
Musify (MUSA)
Power & Thermal
Power Supply
4x2400Вт Platinum (High Voltage DC)
TDP
5500
Cooling Type
Пассивное
Server Specifications
Technical ID
Moore Threads MCCX D800 S4000
Platform
Moore Threads MCCX D800
Form Factor
4U
Warranty
3
Primary Tasks
Обучение моделей, Инференс, Видеоаналитика
Price
On request