• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
  • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    SemiAnalysis on HBM Stack Height Capacity and Bandwidth

    SemiAnalysis reply explains why lowering HBM stack height from 12-Hi to 8-Hi improves efficiency.

    SE
    1 Source, 25d ago, first seen 25d ago

    TLDR

    SemiAnalysis, a semiconductor research firm, posted that capacity scales with stacks times dies per stack times GB per die, while bandwidth scales with stacks times interface width times pin speed. The reply states that stack height mainly adds capacity, not bandwidth. Cutting from 12-Hi to 8-Hi removes one-third of the DRAM dies, the most constrained part of the bill of materials, while bandwidth stays the same or rises slightly. Bandwidth per GB of capacity therefore increases 50 percent. The post included a dark-themed infographic chart.

    Combined views

    39.5K

    1 Source, first seen 25d ago

    Combined views

    39.5K

    1 Source, first seen 25d ago

    106 likes
    106 likes
    12 comments
    59 saves
    7 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    12 comments
    59 saves
    7 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    1 Source

    @SemiAnalysis_Capacity ≈ stacks × dies per stack × GB per die Bandwidth ≈ stacks × interface width × pin speed Stack height buys capacity, not bandwidth. Cutting 12-Hi to 8-Hi drops a third of the DRAM dies, the most supply-constrained silicon on the BOM, while bandwidth holds or ticks up. Bandwidth per GB of capacity rises 50%, and $/bandwidth gets meaningfully better right as HBM prices hike. Inference is bandwidth bound: every decoded token re-reads weights and KV cache. (2/2)

    1 Source

    @SemiAnalysis_Capacity ≈ stacks × dies per stack × GB per die Bandwidth ≈ stacks × interface width × pin speed Stack height buys capacity, not bandwidth. Cutting 12-Hi to 8-Hi drops a third of the DRAM dies, the most supply-constrained silicon on the BOM, while bandwidth holds or ticks up. Bandwidth per GB of capacity rises 50%, and $/bandwidth gets meaningfully better right as HBM prices hike. Inference is bandwidth bound: every decoded token re-reads weights and KV cache. (2/2)