Moore Threads · Accelerator Card
MTT S2000
(Translation draft; official site has no standalone English S2000 page) MTT S2000 is a first-generation MUSA full-featured server GPU covering graphics rendering, AI, audio/video codec and scientific computing. It supports Vulkan, DirectX, OpenGL and OpenGL ES, mainstream AI networks and compute frameworks, and x86 / Arm / LoongArch CPU architectures, with GPU virtualization and Kubernetes container support.
Key Highlights
- (Translation draft) Full-featured GPU for AI, graphics, physics/scientific computing and UHD codec
- rich graphics APIs
- GPU virtualization (MT Mesh, passthrough and hardware vGPU)
- cloud-native container QoS and memory isolation
- first-gen Android Cloud Accelerator (ACX)
- low power and high efficiency
- multi-CPU-architecture support.
Technical Specifications
| 核心/Cores | 4096 |
|---|---|
| FP32 算力/FP32 Compute | 10.4 TFLOPS |
| 显存/Memory | 32GB, 256-bit |
| 最大功耗/Max TDP | 150W |
| 外形/Form Factor | single-slot, 279.5×126.4mm |
| 总线/Bus | PCIe Gen3 |
| 编解码/Codec | 2× first-gen MUSA encoder, 2× first-gen MUSA decoder |
Typical Use Cases
(Translation draft) Cloud gaming, cloud phone, cloud testing, cloud desktop, cloud computing and rendering, video cloudAndroid-cloud apps, and industry graphics/inference workloads such as BIM and GIS.
官网无独立英文产品页（en 路由回退为英文首页），英文列全部为翻译稿；量化规格来自官网中文规格表；图片由官网 hero banner 裁取卡体——官网未提供 S2000 独立封面产品图，本地 swiper 图为云游戏/云手机/云测试场景插图（非硬件渲染），未采用；doc 列为官网提供的 vGPU 驱动压缩包（非 PDF 规格书）。
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 S3000
Full-featured server-class GPU View details →
Moore Threads