Diefolio
English

Moore Threads · GPU

Moore Threads MTT S5000

Moore Threads MTT S5000 — Specifications and benchmarks. .

These specifications are supported by the linked sources. Core counts, clocks and memory capacities do not predict workload performance. Check benchmark versions, environments and sample counts.

Evidence review in progress · Search indexing deferred

Compare this product →

Compute

ParameterValue and conditions
Architecture
PH100; PingHu architecture
Source and conditions

View source ↗

{
  "state": "known",
  "raw_label": "GPU Architecture",
  "raw_value": "PH100; PingHu architecture",
  "qualifiers": {
    "configuration_status": "source-listed; family maxima and clock ranges retained as text",
    "evidence_type": "manufacturer-specification",
    "peak_compute_engine": "unspecified; do not mix with dense/sparse Tensor peaks",
    "source_field": "GPU Architecture",
    "source_language": "zh-cn",
    "source_parameter_language": "zh-cn",
    "source_row": 1
  },
  "document_id": "chinese-www-mthreads-com-52186b9832147cdb1428f8cd-1cad82f3210307e1"
}

Connectivity and features

ParameterValue and conditions
MUSA native precision support
FP8 through FP64
Source and conditions

View source ↗

{
  "state": "known",
  "raw_label": "Native AI precision support",
  "raw_value": "FP8 through FP64",
  "qualifiers": {
    "configuration_status": "source-listed; family maxima and clock ranges retained as text",
    "evidence_type": "manufacturer-specification",
    "measurement_type": "manufacturer_peak",
    "peak_compute_engine": "unspecified; do not mix with dense/sparse Tensor peaks",
    "precision": "unknown",
    "source_field": "Native AI precision support",
    "source_language": "zh-cn",
    "source_parameter_language": "zh-cn",
    "source_row": 1,
    "sparsity": "unknown"
  },
  "document_id": "chinese-www-mthreads-com-52186b9832147cdb1428f8cd-1cad82f3210307e1"
}
MUSA AI software support
PyTorch; Megatron-LM; vLLM; SGLang; Torch-MUSA
Source and conditions

View source ↗

{
  "state": "known",
  "raw_label": "AI software support",
  "raw_value": "PyTorch; Megatron-LM; vLLM; SGLang; Torch-MUSA",
  "qualifiers": {
    "configuration_status": "source-listed; family maxima and clock ranges retained as text",
    "evidence_type": "manufacturer-specification",
    "measurement_type": "manufacturer_peak",
    "peak_compute_engine": "unspecified; do not mix with dense/sparse Tensor peaks",
    "precision": "unknown",
    "source_field": "AI software support",
    "source_language": "zh-cn",
    "source_parameter_language": "zh-cn",
    "source_row": 1,
    "sparsity": "unknown"
  },
  "document_id": "chinese-www-mthreads-com-52186b9832147cdb1428f8cd-1cad82f3210307e1"
}
Manufacturer AI workload conditions
Published >=4000 prefill and >=1000 decode tokens/s per-card-normalized claims use a 48-card 2P4D cluster, FP8, MTP on; not a single-card isolated run; exact model/checksum unspecified
Source and conditions

View source ↗

{
  "state": "known",
  "raw_label": "Manufacturer AI workload scope",
  "raw_value": "Published >=4000 prefill and >=1000 decode tokens/s per-card-normalized claims use a 48-card 2P4D cluster, FP8, MTP on; not a single-card isolated run; exact model/checksum unspecified",
  "qualifiers": {
    "configuration_status": "source-listed; family maxima and clock ranges retained as text",
    "evidence_type": "manufacturer-specification",
    "measurement_type": "manufacturer_peak",
    "peak_compute_engine": "unspecified; do not mix with dense/sparse Tensor peaks",
    "precision": "unknown",
    "source_field": "Manufacturer AI workload scope",
    "source_language": "zh-cn",
    "source_parameter_language": "zh-cn",
    "source_row": 1,
    "sparsity": "unknown"
  },
  "document_id": "chinese-www-mthreads-com-52186b9832147cdb1428f8cd-1cad82f3210307e1"
}

Public benchmarks

Compare results with matching versions, scenes, settings and statistics. A run count of 0 identifies an aggregate or reported value, not a newly measured run.

No comparable public scores are available. This does not mean a score of zero or poor performance.

Sources

Source terms and retrieval dates are listed below. Published values and conditions may change.