Skip to main content

Moore Threads MTT S80

Moore Threads' first consumer-grade gaming graphics card, equipped with full "Chunxiao" chip, 4096 MUSA cores, released November 2023, representative work of domestic gaming graphics cards.


Core Specifications

SpecificationValue
ArchitectureMUSA (Chunxiao chip)
Process7 nm
MUSA Cores4096 cores
GPU Frequency1.8 GHz
TDP255 W
Memory16 GB GDDR6
Memory Bus Width256 bit
Memory Bandwidth448 GB/s
FP32 Compute14.4 TFLOPS
FP16 Compute57.6 TFLOPS (estimated)
INT8 Compute115.2 TOPS (estimated)
InterfacePCIe 5.0 x16
Video OutputSupports 8K ultra-HD / 1080P 360Hz
Launch Date2023-11-11
Price¥2,999 (launch price)

Technical Highlights

  • Chunxiao chip: Moore Threads' second-generation GPU chip, 7nm process, 4096 MUSA unified computing cores
  • PCIe 5.0: Industry's first consumer-grade graphics card supporting PCIe 5.0, bidirectional bandwidth up to 128GB/s
  • Full-function GPU: Simultaneously supports graphics rendering, AI computing, and video encoding/decoding, based on the same MUSA architecture
  • DirectX support: Supports DirectX 11/12, can run mainstream 3D games
  • AI inference capability: FP16 compute 57.6 TFLOPS, can accelerate local AI inference (such as LLM inference)

Product Positioning

MTT S80 is Moore Threads' first gaming graphics card targeting the consumer market, positioned as NVIDIA RTX 4060 level product. As the representative of "domestic gaming graphics cards", S80 has unique value in domestic substitution and local AI inference scenarios.


Application Scenarios

  • Domestic 3A gaming (based on DirectX 11/12)
  • Local AI large model inference (LLaMA, ChatGLM, etc.)
  • 8K video playback and processing
  • Graphics rendering and CAD
  • Domestic operating systems (Unity UOS, Kylin) graphics acceleration

Reference Price

ChannelPriceDescription
Official launch¥2,999JD.com launch price November 2023
Current market price¥2,500–3,500Fluctuates with market supply and demand

  • MTT S70 — Streamlined version gaming graphics card
  • MTT S5000 — Training+inference integrated server GPU

References