DEEP DIVES · LONG READS

Deep dives · page 9

325 deep dives · newest first · 30 per page · page 9 of 11

Looking for a title? Use site search.

2026-10-23 14

Infra

Is 2:4 Sparsity Still Worth It? Semi-Structured Sparsity versus Quantization in LLM Inference

An independent researcher compares "add 2:4 sparsity" versus "drop one precision tier" on Llama-3.1-8B-Instruct and Qwen3-8B under matched storage bits (including sparse indices) and matched nominal compute tiers, with measurements on L40S, H100, and RTX PRO 6000. ==As long as a lower precision is still available, adding 2:4 is both less accurate and slower==: 2:4 + INT8 costs the same storage as INT5.

2026-10-22 3

2026-10-21 3

2026-10-20 2

2026-10-19 3

2026-10-16 1

2026-10-15 1

2026-10-13 1

2026-09-30 1

2026-09-10 1