Half of Solana's Transactions Are Not Transactions
Solana is usually said to do thousands of transactions per second and the XRP Ledger about fifteen. Both numbers are real and neither means what people think. We measured nearly six hours of Solana, 218,728 of its transactions and 1,000 XRPL ledgers, then took out the things that are not user transactions. Solana still does far more work — but roughly a third of what its headline claims, and it takes nearly four times as long to make a payment permanent.
What gets counted, and what should not
A transaction counter should measure work someone asked for and got. Two things on Solana fail that test, and they are large.
Votes. Every Solana validator publishes a vote transaction for nearly every block, and those votes are ordinary transactions on the chain — so they land in the count. Across 5.8 hours of measurement they were 1,621 of 3,366transactions per second, 48% of the total. They are how the network agrees with itself. The XRP Ledger does the same job without writing anything to the ledger, so it has no equivalent to remove.
Failures. Pulling 150 real Solana blocks apart transaction by transaction — 218,728 transactions in all, 36% of the non-vote transactions failed — they were included, charged a fee, and changed nothing. Most are trading bots racing for the same opportunity, where only the first one can win.
The same treatment on the XRP Ledger is shorter, because there are no votes to remove — but it is not flattering either. Of 69,664 transactions across 1,000 ledgers, 23% failed, mostly trade offers that could not fill.
The fair comparison
Counting only transactions that a user asked for and that actually worked: Solana ~1,117 per second, the XRP Ledger 13.9. Solana does about 80 times more.
That is a very large gap and it should be stated plainly rather than explained away. Solana carries far more activity than the XRP Ledger does, and no amount of adjusting the definition changes that. What the adjustment does change is the size of the claim: the honest figure is roughly a third of the headline one, because two out of every three transactions Solana reports are either a validator voting or an attempt that failed.
It is worth being just as blunt about the XRP Ledger. Its 23% failure rate is not something to be proud of, and its throughput is a fraction of Solana's. Most of what it does carry is trading: 47% of successful transactions are offers being placed and another 10% are offers being cancelled, against 28% payments.
But “fast” means two different things
Speed is usually quoted as how quickly a transaction appears. The number that matters for a payment is different: how long until it cannot be reversed. Those are the same thing on the XRP Ledger and very different on Solana.
A Solana block appears in about 422 milliseconds, which is where the “sub-second” reputation comes from. But it is not irreversible until it is finalized, and we measured that lag directly by asking the network for its latest block twice — once accepting anything, once accepting only finalized blocks. The gap was exactly 32 blocks on all twenty attempts, which at 422ms each is 13.5 seconds.
An XRP Ledger ledger closes every 3.86 seconds, and when it closes it is done. There are no confirmation levels to choose between and no possibility of the chain reorganising underneath you. So the ledger that looks nine times slower per block is 3.5 times faster to a payment you can rely on.
Which number should you use?
It depends entirely on what you are doing. If you are a trading bot, 422ms to appear is what matters and 13.5s to finality is irrelevant, because you will have traded again long before then. If you are settling a payment, taking money for goods, or moving funds between institutions, the only number that counts is time-to-irreversible — and there the ordering flips.
This is why comparing the two chains on a single figure is close to meaningless. They are not fast and slow versions of the same machine. Solana optimises for volume and speed of appearance and accepts a waiting period for certainty. The XRP Ledger optimises for certainty and accepts far lower volume. Both trades are defensible; neither is the “winner” the comparison implies.
Method
Measured 3 August 2026, both chains at the same time of day. Solana throughput from getRecentPerformanceSamples over 346 samples of 60 seconds — 5.8 hours of network time — using the numNonVoteTransactions field the RPC returns alongside the total; raw TPS came out at 3366 with a 95% confidence interval of ±33, so the spread is small. The failure rate comes from 150 blocks pulled with getBlock at full transaction detail — 218,728 transactions — strided across roughly two hours of slots rather than taken consecutively, separating votes by looking for the Vote program in each transaction's account list. Finality lag from getSlot at processed versus finalized commitment, 20 times: the answer was 32 slots on every attempt, with no variance at all. XRPL figures from 1,000 consecutive ledgers (69,664 transactions, 64 minutes) read off our own node, counting only tesSUCCESS as successful.
Two notes on honesty. First, an earlier run of this article used 6 Solana blocks and 200 XRPL ledgers and understated both failure rates — Solana 30.6% to 36%, XRPL 16.8% to 23% — which is why the sample was grown tenfold before publishing. Second, the failure figures above are pooled: every transaction counted equally. Averaging per block instead gives 33.1% (±2) on Solana and 12.4% on XRPL, both lower — because failures cluster in the busiest blocks. Pooled is the right answer to “what share of transactions failed”; the per-block number is the right answer to “what does a typical block look like”. Throughput on either chain varies with demand, so treat all of this as a well-sampled snapshot rather than a constant.