G3 Sequencer: Getting to 60,000 TPS with 10 ms Latency [Conduit Workshops]
Most sequencer benchmarks are a number on a slide. This session is a live run.
Rez, Core Team Lead at Conduit, will walk through how we rebuilt the sequencer on top of Reth, keeping Reth’s validation pipeline while replacing ingestion and block building, and then put G3 under sustained payment-style load in a production-like environment with the metrics dashboard on screen.
What we’ll cover
Durable ~10 ms flashblocks: confirmations that survive an unsafe shutdown or failover, with monitoring for any preconfirmation reorg
Pipelined block building: starting the next block early via Engine API attribute hints to cut builder idle time and tail latency
Slipstream: native batch RPC for authorized backends that bypasses the vanilla mempool, with synchronous inclusion or rejection
Workload-specific prewarming: pulling the right storage slots into memory before execution
Scaling work we upstreamed to Reth: configurable peering for very large blocks and an RLPx sub-protocol for chunking block bodies past the 16 MiB limit
Why it matters
Offchain applications can react to onchain events several times within a single block, without trading durability for speed.
Agenda (30 min)
The use case and how G3 is built (10 min)
Live load test (10 min)
Q&A (10 min)
About the speaker
Rez is the Core Team Lead at Conduit, where he leads the G3 sequencer.
About Conduit
Conduit is the infrastructure provider for production and enterprise custom blockchains. Conduit designs and operates the systems required to run high-performance blockchains reliably at scale, from sequencing and RPC infrastructure to upgrades, security, and operational coordination.
Conduit powers over 55% of blockchains on Ethereum, as well as many of the most demanding chains in crypto and beyond, including Blackbird, Polygon, Zama and Tempo. Backed by Paradigm and Haun Ventures, Conduit gives institutions enterprise-grade infrastructure so they can focus on building.
![Cover Image for G3 Sequencer: Getting to 60,000 TPS with 10 ms Latency [Conduit Workshops]](https://images.lumacdn.com/cdn-cgi/image/format=auto,fit=cover,dpr=2,background=white,quality=75,width=400,height=400/uploads/30/a5bd4b63-5a13-49c5-ba43-5a9f8b0f12a2.png)