Chaitex
Chaitex
Back to Servers
YH001 PCIe Server 384GB  1

YH001 PCIe Server 384GB

The YH001 PCIe Server is a versatile AI acceleration platform designed for high-concurrency inference and model fine-tuning.

Utilizing the innovative RISC-V cloud AI architecture, it offers a flexible and supply-chain independent solution for modern data centers.

The server integrates eight YH001 PCIe accelerators, providing a total of 384GB of video memory (48GB per card). It is highly optimized for DeepSeek-series models, leveraging matrix processing engines to accelerate transformer-based workloads.

The PCIe form factor ensures easy integration into existing server infrastructures while maintaining high-speed data throughput.

By supporting advanced data formats like FP8 and INT4, the YH001 delivers a significant increase in computational density and power efficiency for enterprise-grade AI applications.

Application scenarios

HPC applications and LLM training with TP4/TP8 topologies.
Scalable AI clusters without CUDA via YHLink.
AI infrastructure in cloud data centers.

CPUs
2x Intel 8558(48 ядер,2.1ГГц)
RAM
1ТБ (16x 64ГБ)
Accelerators
8x YH001 PCIe 48GB
Storage
2x 480G SATA SSD + 2x 1.92T NVME SSD
Pre-order
Product status

CPU & Memory

CPU
2x Intel 8558(48 ядер,2.1ГГц)
RAM
1ТБ (16x 64ГБ)
RAM Type
-

Storage Configuration

SSD Model
-
Total Storage
2x 480G SATA SSD + 2x 1.92T NVME SSD
Disk Interface
-

GPU Configuration

GPU Vendor
YH
GPU Model
YH001 PCIe 48GB
GPU Count
8
Total VRAM
384ГБ HBM3e
Interconnect
YHLink 600ГБ/c

Network & Virtualization

Network Interfaces
DH E310 (2*25Гбит/с)
Virtualization (MIG)
-
CUDA Compatibility
Не зависит от совместимости с CUDA, нативно поддерживает основные фреймворки

Power & Thermal

Power Supply
-
TDP
4 600 - 6 100
Cooling Type
Пассивное

Server Specifications

Technical ID
YH001 PCIe Server
Platform
YH
Form Factor
4U
Warranty
3
Primary Tasks
Обучение LLM,распределённое обучение, высоконагруженный inference, построение масштабируемых AI-кластеров
Price
On request