List your company

Latency Evaluation on NSA Workload Head Dimension 128

0.67Latency (s)LLM-TL
-0.35486.562613.4820.3974Jun 14, 2025

Evaluation results

MethodLinks
LLM-TLSequence Length=512, Head Dimension=128, Hardware=A100 GPU2025.060.671.25
Naive NSASequence Length=512, Head Dimension=128, Hardware=A100 GPU2025.060.84—
LLM-TLSequence Length=1k, Head Dimension=128, Hardware=A100 GPU2025.061.261.33
Naive NSASequence Length=1k, Head Dimension=128, Hardware=A100 GPU2025.061.68—
LLM-TLSequence Length=2k, Head Dimension=128, Hardware=A100 GPU2025.062.591.29
Naive NSASequence Length=2k, Head Dimension=128, Hardware=A100 GPU2025.063.35—
LLM-TLSequence Length=4k, Head Dimension=128, Hardware=A100 GPU2025.065.251.26
Naive NSASequence Length=4k, Head Dimension=128, Hardware=A100 GPU2025.066.61—
LLM-TLSequence Length=8k, Head Dimension=128, Hardware=A100 GPU2025.0610.591.26
Naive NSASequence Length=8k, Head Dimension=128, Hardware=A100 GPU2025.0613.34—
LLM-TLSequence Length=16k, Head Dimension=128, Hardware=A100 GPU2025.0621.271.24
Naive NSASequence Length=16k, Head Dimension=128, Hardware=A100 GPU2025.0626.29—