- Agave briefly reached 1.1 million transactions per second in a synthetic single-node test, matching Firedancer’s reported lab peak.
- The run used simple transfers and relaxed throughput limits, indicating a short-lived burst rather than sustainable mainnet capacity.
Solana’s Rust validator client Agave has matched Firedancer’s laboratory peak by briefly reaching 1.1 million transactions per second in a synthetic single-node benchmark. The run used simple transfers on a branch that combines several performance changes not yet merged into the main codebase. The developers noted that block and shred limits were disabled for the test and that the figure reflects a short burst rather than a steady-state throughput number.
Hit a burst of 1.1m TPS on agave this morning. Single-node synthetic test with simple transfers. On a branch with several changes not yet merged:
– PoH recording improvements
– Status-cache performance improvements
– scheduler-bindings
– block/shred limits disabled@solana— Andrew Fitzgerald (@apfitzge) August 28, 2025
Test setup and immediate caveats
The measurement isolates execution and scheduling under controlled conditions. With block and shred limits disabled, the benchmark reveals headroom on a single machine rather than the constraints of a globally distributed network. The disclosure emphasised that the result should not be read as mainnet capacity. It instead reflects ongoing work on Proof-of-History recording, status-cache behaviour and a mechanism dubbed “scheduler-bindings.” That mechanism allows custom block-packing logic to be attached without forking core, which is intended to accelerate iteration on how transactions are ordered and packed.
Engineers also referenced changes adjacent to the execution path. Recent Agave updates describe a revamped TPU client, reduced AccountsDB input and output, a greedy scheduler and improvements to snapshot handling and gossip. These adjustments target contention and coordination costs that influence user-perceived latency during periods of heavy load. They are designed to improve predictability even if they do not translate one-for-one into higher synthetic peak numbers.
Competitive context and product focus
The 1.1 million TPS data point feeds into a wider narrative about client competition within Solana. Firedancer and Agave now share the same synthetic peak, which indicates that multiple implementations are pushing similar ceilings under laboratory conditions. Market participants view this through the lens of resilience and operational redundancy since distinct codebases stress different parts of the system and reduce single-client dependency.
Attention is also shifting from headline throughput to confirmation times. Community discussion around SIMD-0326, also known as Alpenglow, centres on reducing consensus timers towards approximately 150 milliseconds. That target addresses latency for payments, trading and real-time applications where consistent finality intervals define user experience more than sporadic peaks. At press time, SOL traded at 202.65 US dollars.
