
Graphics card *1
Dual GPU design with 48GB GDDR6 large video memory
PCIe 5.0 x8 + PCIe 5.0 x8 interface for high bandwidth and powerful computing performance
Dual-slot width design suitable for multi-GPU synchronous computing environments
Optimized software ecosystem with native support for PyTorch, ISV support, IPEX-LLM inference engine, and compatibility with vLLM
Supports Intel LLM-Scaler for large language model (LLM) acceleration
Compatible with various LLM models including FP8, INT4 precision and models such as DeepSeek-R1 series, Qwen3 series, LLMMOE, multi-modal and embedding models
Supports up to 16-card interconnection (Battlematrix) with total memory up to 384GB for smooth operation of 100B+ parameter models

R5 230-2G Heat Sink
CNÂ¥179.00

[R5 2202G] with fan model
CNÂ¥169.00
Suitable for AI content creation including 3D modeling, image/video generation, and speech generation
Intel Arc Pro drivers optimized for professional creative workflows such as video editing and 3D design rendering
Enhanced performance and stability for applications like Adobe Premiere Pro, DaVinci Resolve Studio, Blender, Autodesk 3ds Max, Cinema 4D
Consumer-grade platforms with efficient operation
Windows and Linux based LLM software stacks for local deployment of mainstream open-source models including DeepSeek, Qwen3-32B, Qwen3-VL, etc.
MS-iCraft Z890 Vertex
MS-iCraft Z890 Pacific
MS-iCraft Z890 Arctic
MS-Terminator Z890-A
MS-Terminator Z790-A
MS-iCraft Z790 WiFi V2
MS-WorkStation W790-112L
MS-PCFarm H770I D5
MS-Terminator H770 YTX D5 WIFI
End-to-End Request Latency (E2E Request Latency)
Time to First Token (TTFT)
Inter-Token Latency (Time Per Output Token Latency)
Request Prompt Length Heatmap
Finish Reason
Token Throughput
Scheduler State
Cache Utilization
Request Generation Length Heatmap
Queue Time
Prefill & Decode Time
Max Generation Tokens in Sequence Group

RTX5050
CNÂ¥2,899.00

[LP 5060 HalfHeight Blade Card]
CNÂ¥4,199.00