MTT S3000, built on Moore Threads MUSA architecture, is the first server-class GPU product based on the ChunXiao chip. With a complete MUSA software stack that covers graphics rendering, video processing, and deep learning, MTT S3000 can provide diverse and intelligent computing power for cloud gaming, cloud rendering, cloud video, digital twins, and digital content creation. It's aimed at building a solid computational foundation for data centers, AI data centers, and metacomputing centers. The card also facilitates the innovation and implementation of applications within the metaverse.

MUSA Architecture

ChunXiao chip with a comprehensive performance uplift across four major engines

Intelligent Computing

Acceleration of deep learning computation through the MUSA architecture and software stack, balancing usability, scalability, and compatibility

Cloud-Native

Innovative cloud-native GPU virtualization and containerization technologies for significantly improved GPU utilization

Cloud Video

Second-generation smart multimedia engine for accelerated cloud-based 8K video applications

MTVERSE

An integrated hardware and software solution aiding in the construction and application of the metaverse

MetaCreation Platform

The starting point for developers using MUSA, igniting the "meta" dynamic for innovation

First-Generation MUSA Architecture

MUSA is a unified metacomputing system architecture adopted by Moore Threads' products. It includes a unified programming model, software runtime libraries, driver framework, instruction set architecture, and chip architecture.

Cloud-Native

Moore Threads offers a series of cloud-native GPU solutions based on the innovative MT Mesh 2.0.

MT Mesh 2.0 can automatically allocate GPU compute and memory resources based on application loads in the cloud, achieving elastic scalability of GPU power. It allows a single GPU to be arbitrarily divided among multiple containers or virtual machines, and also supports scheduling multiple GPUs for a single container or virtual machine.

Container Virtualization (sGPU)

Leverage the Kubernetes ecosystem and MT Mesh 2.0 for flexible and efficient deployment of containerized GPU resources.

Virtualization (vGPU)

Pioneering the industry with MT Mesh 2.0, Moore Threads has introduced resource elastic splitting technology that dynamically allocates and modifies virtualized GPU resources without the need for rebooting. This enables on-demand compute power and dynamic scaling, maximizing computational capacity utilization. MT Mesh 2.0 introduces the new Time-Space Division feature, which supports hardware virtualization (SR-IOV), secure physical partitioning, up to 32 virtualization paths, Windows cloud desktop GPU virtualization, and GPU passthrough for Unity and Kylin operating systems.

Android Cloud Acceleration Technology ACX

Using Android GPU container pass-through and integrated rendering and encoding technologies can accelerate Android cloud mobile solutions, reducing latency and increasing concurrent paths. With native support for the OpenGL ES and OpenGL rendering frameworks, as well as hardware acceleration for Android game texture compression algorithms like ETC and ETC2, ACX provides improved rendering effects and compatibility features for Android cloud gaming.

MTVERSE

Moore Threads Metaverse platform provides a full-stack solution from hardware clusters and software infrastructure to SDKs, encompassing multiple core elements of the metaverse

MTVERSE builds on Moore Threads MUSA GPU clusters to provide users with computational infrastructure and services, including platforms for big data, intelligent computing, graphics rendering, and physical simulation. It is a full-stack solution of hardware, software and SDKs toolchains , covering the core of the metaverse—people, scenes, and content. The extensive SDK at the upper layer enables developers and applications to easily access the capabilities of MTVERSE, facilitating functions like digital human creation, document understanding, voice recognition, visual recognition, natural language understanding, conversational interaction, physical simulation, AIGC, and more. This further simplifies the development of applications and solutions, shortening the development cycle.

Cloud Video

Facilitating the transition of video applications to be more ultra-definite and intelligent

MTT S3000 boasts formidable video processing capabilities. It integrates MT Smart Media Engine 2.0, which is Moore Threads' second-generation smart multimedia engine that accelerates a variety of audio and video applications in the cloud. The applications it accelerates include video AI analysis, video conferencing, cloud gaming, cloud desktops, and real-time interaction, with support for up to 8K UHD HDR video real-time processing. It supports mainstream audio and video frameworks including FFmpeg and GStreamer, as well as universal video acceleration APIs across different operating systems.

MetaCreation Platform

The starting point for developers using MUSA, helping developers to quickly learn and use cloud computing capabilities

The MetaCreation Platform is Moore Threads' central platform to assist developers in learning, testing, managing, accessing, and analyzing GPU technology using MTT S3000. With the support of the MetaCreation Platform, developers can fully unleash the computational potential of Moore Threads GPUs in application scenarios such as the metaverse, digital twins, game development, and data centers, further igniting the creative spark within developers.

MTT S3000

Specifications