Unverified50% confidenceFactExact time
DeepGEMM Ascend edition supports matrix multiplication in multiple precision combinations including FP4, FP8, and BF16.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/3/2026
First Seen
Valid until: 1/1/2027
Sources
Related Entities
Related Claims
UnverifiedDeepGEMM Ascend was first released on 2026.09.30, supports Ascend 950 devices, is fully compatible with DeepGEMM's API, supports BF16, FP8, and FP4 matrix multiplication, and uses the same package name as the original version.76% similarUnverifiedTensorRT supports multiple precision calibrations including FP32, FP16, INT8, and FP873% similarUnverifiedThe M5 Ultra's FP8 and FP4 floating-point compute performance is identical to its BF16 performance, and the gap with the Spark narrows noticeably when dropping to FP4 precision.73% similarUnverifiedDeepGEMM Ascend edition achieves a hardware utilization rate of 98.3% for FP4 multiplied by FP4.70% similarUnverifiedThe input matrices that the vLLM linear backend must handle may be in FP16, BF16, INT8, or various GPTQ/AWQ quantization formats.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/970644API
curl https://kongchang.com/api/v1/knowledge/claims/970644MCP
get_claim(id=970644)