Scheduler Reward Density Mechanics
A persistent misconception among Solana developers is that reducing compute budget below the default 200,000 CU ceiling provides no economic benefit when base transaction fees are flat at 5,000 lamports.
Empirical measurements across both Agave and Frankendancer validators demonstrate the opposite.
The Priority Score Formula
In modern validator scheduling architectures, incoming transactions are prioritized by Reward Density per requested compute unit:
$$\text{Priority Score} \propto \frac{\text{Base Fee (5,000 L)} + \text{Priority Fee}}{\text{Requested Compute Unit Limit}}$$
Because the requested Compute Unit limit sits directly in the denominator, decreasing the budget increases scheduler priority score at zero incremental financial expenditure.
Density Scaling Table (Flat 5,000 Lamport Base Fee)
| Program Architecture | Requested CU Budget | Total Base Fee | Priority Tip | Reward Density | Relative Priority Multiplier |
|---|---|---|---|---|---|
| Standard Anchor Default | 200,000 CU | 5,000 L | 0 L | 0.0250 L/CU | 1.0× (Baseline) |
| Optimized Borsh Contract | 50,000 CU | 5,000 L | 0 L | 0.1000 L/CU | 4.0× |
| Constrained Micro-Program | 30,000 CU | 5,000 L | 0 L | 0.1667 L/CU | 6.67× |
| u64se-seu-solana Wire | 2,000 CU | 5,000 L | 0 L | 2.5000 L/CU | 100.0× (Dominant) |
To achieve the same $2.50\text{ L/CU}$ reward density with a bloated 200,000 CU transaction, a sender would be forced to pay an additional 495,000 lamports (~0.0005 SOL) priority tip on every transaction.
Experimental Isolation: Write-Lock Contention Model
To isolate scheduling priority from arrival order and eliminate concurrency noise, u64se-block-order enforces two experimental constraints:
- Single-Payer Write-Lock Contention: All probe transactions write-lock the identical fee-payer authority (
6MpbQcu...). Because Solana’s Sealevel runtime serializes transactions that write-lock the same account, the validator is forced to establish a strict total ordering among our probes when they enter the same Banking Stage window. - Inverted Dispatch Order: Probes are constructed and dispatched from
Rank 0(lowest priority density) toRank 4(highest priority density). WhenRank 4lands ahead ofRank 3inside the same block, pure arrival FIFO is falsified in favor of priority-density sorting.