Moore Threads · Accelerator Card
MTT S3000
MTT S3000, built on the MUSA architecture, is the first server-class GPU based on the ChunXiao chip. With a complete MUSA software stack covering graphics rendering, video processing and deep learning, it provides general-purpose intelligent compute for cloud gaming, cloud rendering, video cloud, digital twins and digital content creation, targeting data centers, AI computing centers and metacomputing centers.
Key Highlights
- First-generation MUSA architecture
- graphics rendering, video processing and deep learning
- MT Mesh 2.0 for elastic containerized sGPU and vGPU partitioning
- MTVERSE metaverse full-stack platform
- passive cooling for data-center deployment
- MUSA Security Engine 1.0 with TEE.
Technical Specifications
| FP32 算力/FP32 Compute | 15.5 TFLOPS |
|---|---|
| 显存/Memory | 32GB GDDR6, 256-bit |
| 显存带宽/Memory BW | 448GB/s |
| 总线/Bus | PCIe Gen5 x16 |
| 显示/Display | 2× DP 1.4a |
| 散热/Thermal | passive cooling |
| 安全/Security | MUSA Engine 1.0, TEE |
| 尺寸/Dimensions | 268.6×111.0×38.7mm |
| 虚拟化/Virtualization | sGPU/vGPU elastic partitioning, SR-IOV |
Typical Use Cases
Cloud gaming, cloud phone, cloud rendering, video cloud, digital twins, digital content creation, Android-cloud acceleration (ACX), deep-learning inference and training.
概述/亮点英文取自官网英文页；量化规格表来自官网中文页（英文页未列数值表）；图片为官网产品封面图 cover.jpg。
Related Products
Moore Threads
MTT S5000
Universal GPU for AI Training and Inference View details →
Moore Threads
MTT S4000
Large-model AI inference and training GPU with Tensor Cores View details →
Moore Threads
MTT S2000
Low-power full-featured server GPU for cloud-native workloads (translation draft) View details →
Moore Threads