Assertion Supported AI assessment confidence: 95% certainty 5/5 debate potential 3/5

Freedman: pgvectorscale is 28 times faster than Pinecone for high recall

Mike Freedman · Turbocharging Postgres for Time Series & Vectors — Timescale CTO Mike Freedman | Data Driven NYC · Sep 25, 2024 · at 9:56

Mike Freedman, CTO of Timescale, presents benchmark data comparing Timescale's open-source PostgreSQL extension (pgvectorscale) against standalone vector database Pinecone.

0:00 / 0:08exact quote · 8.1s

▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark

“Compared to one of the leading vector-only databases, Pinecone, a PG vector scale is 28 times faster for a high recall scenario.”

quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →

The claim, plainly

Timescale's pgvectorscale extension is 28 times faster than Pinecone in high-recall vector search scenarios.

a quote alone can be ambiguous, so this restates it with the references and context of the conversation around it resolved. AI-written; the tape below the player is the source.

Assessment

Timescale's published benchmarks and reproducible test suite (released in June 2024 based on 50 million Cohere 768-dimension embeddings) demonstrated that pgvectorscale achieved a p95 query latency of 28 ms at 99% recall, compared to 784 ms for Pinecone's s1 pod tier. This confirms a 28x latency speedup in high-recall vector search scenarios on a 50M dataset, satisfying the resolution criteria.

On June 11, 2024, Timescale published benchmarks showing pgvectorscale achieved a p95 latency of 28 ms at 99% recall on a 50M embedding dataset, compared to 784 ms for Pinecone's s1 tier (28x lower p95 latency).
Independent technical reviews confirmed that Timescale's open-source benchmark suite reproduces the 28x p95 latency difference against Pinecone s1 under these high-recall conditions.

Checked against

[1] prnewswire.com [2] tigerdata.com [3] github.com [4] dev.to [5] medium.com [6] prnewswire.com [7] vecstore.app [8] crail.ai

"Pinecone" response to Timescale pgvectorscale benchmark 28x "pgvectorscale" "Pinecone" "s1" benchmark details 28x lower latency Timescale pgvectorscale Pinecone benchmark 28x 50M embeddings Timescale pgvectorscale Pinecone 28 times faster 50M embeddings

assessed as of 2026-07-30

This assessment is an AI reading of a public statement, made against the cited references with the confidence shown. It can be wrong. The quote, the tape and the references are all on this page; if something is off, use report an error below. full disclaimer →

Mentioned

pgvectorscale Pinecone

⚑ Spot an error? Report it

More from Mike Freedman

Assertion Not checkable as stated

Freedman: Timescale outperforms vanilla Postgres, InfluxDB, and AWS Timestream

“In fact, not only is Timescale a better database for time series than vanilla Postgres, including compared to both Amazon's RDS and Aurora, but it is also better than custom-built proprietary time series databases like InfluxDB and AWS Timestream, even with th…”

Mike Freedman Sep 25, 2024 ▶ 6:08 Turbocharging Postgres for Time Series & Vectors — Timescale CTO Mike Freedman | Data Driven NYC

Assertion Supported

Freedman: pgvectorscale on AWS is 75% cheaper than Pinecone

“The monthly costs of running such a PG vector scale deployment on AWS is 75% less expensive than PyIncome.”

Mike Freedman Sep 25, 2024 ▶ 10:22 Turbocharging Postgres for Time Series & Vectors — Timescale CTO Mike Freedman | Data Driven NYC

Assertion Not checkable as stated

Freedman: Nobody runs RAG applications at hundreds of thousands of RPS

“Nobody is taking rag apps to that scale today.”

Mike Freedman Sep 25, 2024 ▶ 20:00 Turbocharging Postgres for Time Series & Vectors — Timescale CTO Mike Freedman | Data Driven NYC

Insight

Freedman: New data workloads do not require new database architectures

“A new data workload does not require new database architecture.”

Mike Freedman Sep 25, 2024 ▶ 15:24 Turbocharging Postgres for Time Series & Vectors — Timescale CTO Mike Freedman | Data Driven NYC

Assertion Supported

Freedman: Amazon Redshift is a fork of PostgreSQL 8

“Amazon Redshift is actually a fork of Postgres eight.”

Mike Freedman Sep 25, 2024 ▶ 16:31 Turbocharging Postgres for Time Series & Vectors — Timescale CTO Mike Freedman | Data Driven NYC

Disclosure

Freedman: Timescale's insights service ingests 800 billion records daily

“That insights product is actually backed by a timescale service, much the same as any of our customer could do, that has more than a petabyte of data and currently is ingesting, I think last week I saw, eight hundred billion records a day.”

Mike Freedman Sep 25, 2024 ▶ 25:22 Turbocharging Postgres for Time Series & Vectors — Timescale CTO Mike Freedman | Data Driven NYC

Freedman: pgvectorscale is 28 times faster than Pinecone for high recall

More from Mike Freedman

Turn any episode into a week of clips.