Short version: a transaction on Solana pays a tiny fixed base fee and an optional priority fee. Base fee is a fixed amount of lamports per signature and it goes to the network. Priority fee goes to the leader and is a price per compute unit that you pay voluntarily, but it is practically mandatory to secure ordering in a block where the accounts you work with are overloaded. Most of the questions like "why are my fees high" and "why didn't my transaction land" in fact boil down to the priority fee. The priority fee market is local to the accounts a transaction touches, so a global average fee has literally zero value here. This guide covers both fees, priority market specifics, and how to land on Solana cheaply and reliably.

Key Takeaways

  • Base fee is a fixed amount per signature, with half of it burned. Priority fee is compute units requested multiplied by price per unit, paid to the leader, optional, but practically mandatory during congestion.
  • The priority fee market is local to the accounts that a transaction writes. The optimal fee for a hot pool is completely unrelated to the network average.
  • Over-requesting the compute budget wastes money and makes the transaction unattractive for the leader. Set the limit precisely.
  • Priority fee provides ordering between transactions already received by the leader, not a delivery guarantee. The latter is ensured by a staked submission path.

The base fee

Each transaction on Solana is charged a fixed base fee set in lamports per signature (currently 5,000 lamports, or 0.000005 SOL, per signature). Most transactions have one signature, so the base fee is a fraction of a cent. Half of the base fee is burned, and the other half is distributed to the validator that produced the block.

The base fee is the same in both idle and busy states of the network. It has no effect on the ordering or inclusion. If the base fee was the only type of fee, every transaction would be identical to the leader, hence the necessity of introducing priority fees.

The priority fee

Priority fee is the product of the compute unit limit and the compute unit price. Two transactions with the same price but different limits will end up with different fees, and it is the price per unit, and not total fee, that the leader's scheduler uses to rank transactions.

Priority fee is paid to the leader and serves as compensation for the validators' and, via MEV and fee sharing, delegators' efforts to order transactions under load.

Why the fee market is local

Solana's scheduler has limitations on the compute that may be written to any single account in the block. So, when a pool, token, or oracle account is busy, the competition is around the write budget of that account and only transactions working with the busy account will compete and will need a high fee. Other transactions will not compete and do not need a high fee.

In practice, it means that the commonly provided network average priority fee metric is basically useless. The relevant metric is the fee distribution on transactions writing the same accounts you will be writing. Providers usually provide access to this metric via methods like getRecentPrioritizationFees with the relevant accounts as arguments. It will return a per-slot floor (the lowest fee paid by the transactions that wrote the accounts in question in each of up to 150 recent slots), and savvy clients estimate the fee per transaction from it or from provider estimators reporting percentiles.

Solana TPS explained helps to understand why this per-account budget, rather than a transaction count, is the true limit of the throughput.

Setting the fee correctly

  1. First, simulate your transaction to learn how many compute units it consumes.
  2. Set the compute unit limit to the number of units consumed plus some buffer. Over-requesting increases the total fee and wastefully consumes block budget; the scheduler also prefers transactions that fit.
  3. Then estimate the price per unit based on the fees recently paid by transactions touching the accounts you touch, using the percentile matching your urgency: median for the regular traffic and a high percentile for something that must land in this slot.
  4. And recompute per transaction. Fees on a busy pool may change every second, and the fixed number is usually wrong.
  5. Cap it. Decide on the maximum you are willing to pay for this transaction category and do not exceed it; otherwise you risk running out of budget because of the runaway bidding during the launch.

A transaction well-tuned for a quiet account will pay close to the base fee. A transaction well-tuned for a hot pool during the launch may fetch any price the local market can demand, and this is simply the cost of doing business in this trade.

Priority fees versus Jito tips

They serve different purposes and apply to different parts of the system.

Priority fees versus Jito tips

A bundle does not require priority fees on its transactions for them to be valid: they execute in the order they are listed and the tip decides the auction. For the tip side, check Jito bundles and tips explained.

What fees do not buy: delivery

Priority fee only affects the ordering of the transactions that the leader received. It does nothing for a transaction that never got to the leader. During congestion, the leader's inbound connections are saturated and the protocol services staked connections first, so a high-fee transaction sent via an unstaked connection can be simply dropped before the scheduler sees the fee.

This is the situation described as "I paid a huge fee and it still did not land". The solution is not increasing the fee; it is choosing a staked submission path that reaches the leader, such as a stake-weighted QoS lane or a bundle. Solana MEV relayers compared has both covered, and the OrbitFlare RPC plans feature staked forwarding.

Reading fees after the fact

When the transaction lands, its fee is written in the transaction metadata: the base fee plus the priority fee actually paid. Monitor it. There are two patterns in the data that you should react to:

  • Fees increase while the landing rate decreases: you are bidding against yourself via an unstaked path. Fix the path, not the fee.
  • Fee paid significantly above the median fee in the landed block for the same accounts: your percentile is too aggressive, or your estimator observes the wrong accounts.

Conclusion

Solana transaction fees are a tiny fixed base fee plus an optional priority fee that buys ordering in a market local to the accounts you write. Set the compute unit limit precisely, estimate the price per unit from the fees recently paid by the transactions touching your specific accounts, recompute per transaction, and cap it. Then remember what the fee cannot do: it simply ranks the transactions received by the leader; reaching it under load is about the staked submission path, not about paying more.

For the submission path, the RPC plans feature staked transaction landing and OrbitFlare Apex is the dedicated landing service. For the broader understanding of where the milliseconds go, see network infrastructure for low-latency Solana trading.

Originally published at orbitflare.com.