Brute-force vector search is memory-bound

How it behaves

One input on the horizontal axis, the others held fixed.

Worked examples, recomputed

Every number the mechanisms quote for this relationship is recomputed from it, so none of them disagrees with it.

• 10 million × 768 dims × 4 B ≈ 30 GB
• at 50 GB/s → ≈ 0.6 s per query
• 96-byte PQ codes → scans read about 30× less

Where it appears

Each mechanism states it in its own names. The general symbol each name stands for is shown with it.

• Specialized Processing Models

  What do you build when a query predicate cannot be reduced to a range over an ordered key?

  $$N\times d\times 4\text{bytes}÷\text{bandwidth}$$

  Read it in place →