PostgreSQL with pgvector extension outperforms dedicated vector databases in benchmark

A recent benchmark comparing pgvector, Qdrant, and Pinecone on 50 million vectors revealed that PostgreSQL with the pgvectorscale extension significantly outperformed dedicated vector databases. On the same AWS hardware with a 99% recall target, PostgreSQL achieved 471 queries per second, while Qdrant managed only 41 queries per second, an 11.5x difference. While PostgreSQL showed superior performance for most current retrieval-augmented generation (RAG) workloads in terms of speed, cost, and operational complexity, the benchmark also indicated a scale threshold where dedicated vector databases like Qdrant could offer advantages. AI

IMPACT Suggests that traditional databases may be sufficient for many RAG applications, potentially reducing reliance on specialized vector databases.

RANK_REASON Benchmark comparison of database technologies for AI workloads. [lever_c_demoted from research: ic=1 ai=0.7]

Read on Towards AI →

• AWS
• Milvus
• pgvector
• Pinecone
• PostgreSQL
• Qdrant
• retrieval-augmented generation
• Weaviate

• infra
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →