All chains compared

Nominal block time is what each protocol targets. The measured median is what our nodes recorded over the sample window. Times to confirmation are measured from the block that first included a transaction. Every cell carries its raw value in seconds in a data-seconds attribute.

Chain	Nominal block time	Measured median interval	1 confirmation	3 confirmations	6 confirmations
Bitcoin BTC	10 min	7 min 16 s	7 min 16 s (p90 23 min 58 s)	27 min 10 s (p90 57 min)	58 min 5 s (p90 1 h 38 min)
Litecoin LTC	2 min 30 s	1 min 45 s	1 min 45 s (p90 5 min 44 s)	6 min 43 s (p90 13 min 15 s)	14 min 17 s (p90 23 min 15 s)
Dogecoin DOGE	1 min	43 s	43 s (p90 2 min 25 s)	2 min 48 s (p90 5 min 36 s)	6 min 1 s (p90 9 min 34 s)
Dash DASH	2 min 30 s	1 min 46 s	1 min 46 s (p90 6 min 2 s)	6 min 55 s (p90 13 min 58 s)	14 min 48 s (p90 24 min 31 s)
Ethereum ETH	12 s	12 s	12 s (p90 12 s)	36 s (p90 36 s)	1 min 12 s (p90 1 min 12 s)

Per chain detail

Bitcoin (BTC)

4,175 blocks sampled Heights 955,954 to 960,128 Nominal 10 min 14.5% of blocks took more than twice the nominal time

Block interval	p10	Median	Mean	p90	p99
Seconds between blocks	1 min	7 min 16 s	10 min 21 s	23 min 58 s	48 min 54 s

Confirmations	Median time	p90 time
1	7 min 16 s	23 min 58 s
3	27 min 10 s	57 min
6	58 min 5 s	1 h 38 min

Proof-of-work block discovery is a Poisson process, so intervals are approximately exponentially distributed: the mean sits near the protocol target while the median lands near 0.693x it, and the tail is long. The gap between the nominal block time and the measured median is a property of that distribution, not a sign the chain is running fast. 112 of 4174 intervals (2.68%) run backwards and 1 are exactly zero; the most negative is -422s. Block timestamps are miner-declared and consensus only requires them to exceed the median of the previous 11 blocks, so a miner with a lagging clock can stamp a block earlier than its parent. Raw values are reported; nothing is clamped or reordered.

Litecoin (LTC)

17,234 blocks sampled Heights 3,133,537 to 3,150,770 Nominal 2 min 30 s 13.6% of blocks took more than twice the nominal time

Block interval	p10	Median	Mean	p90	p99
Seconds between blocks	15 s	1 min 45 s	2 min 30 s	5 min 44 s	11 min 39 s

Confirmations	Median time	p90 time
1	1 min 45 s	5 min 44 s
3	6 min 43 s	13 min 15 s
6	14 min 17 s	23 min 15 s

Proof-of-work block discovery is a Poisson process, so intervals are approximately exponentially distributed: the mean sits near the protocol target while the median lands near 0.693x it, and the tail is long. The gap between the nominal block time and the measured median is a property of that distribution, not a sign the chain is running fast. No negative intervals were observed; 20 of 17233 intervals are exactly zero (consecutive blocks sharing a timestamp).

Dogecoin (DOGE)

40,933 blocks sampled Heights 6,269,119 to 6,310,051 Nominal 1 min 14.8% of blocks took more than twice the nominal time

Block interval	p10	Median	Mean	p90	p99
Seconds between blocks	7 s	43 s	1 min 3 s	2 min 25 s	5 min 15 s

Confirmations	Median time	p90 time
1	43 s	2 min 25 s
3	2 min 48 s	5 min 36 s
6	6 min 1 s	9 min 34 s

Proof-of-work block discovery is a Poisson process, so intervals are approximately exponentially distributed: the mean sits near the protocol target while the median lands near 0.693x it, and the tail is long. The gap between the nominal block time and the measured median is a property of that distribution, not a sign the chain is running fast. No negative intervals were observed; 116 of 40932 intervals are exactly zero (consecutive blocks sharing a timestamp).

Dash (DASH)

16,446 blocks sampled Heights 2,496,584 to 2,513,029 Nominal 2 min 30 s 14.4% of blocks took more than twice the nominal time

Block interval	p10	Median	Mean	p90	p99
Seconds between blocks	17 s	1 min 46 s	2 min 38 s	6 min 2 s	12 min 38 s

Confirmations	Median time	p90 time
1	1 min 46 s	6 min 2 s
3	6 min 55 s	13 min 58 s
6	14 min 48 s	24 min 31 s

Proof-of-work block discovery is a Poisson process, so intervals are approximately exponentially distributed: the mean sits near the protocol target while the median lands near 0.693x it, and the tail is long. The gap between the nominal block time and the measured median is a property of that distribution, not a sign the chain is running fast. No negative intervals were observed; 22 of 16445 intervals are exactly zero (consecutive blocks sharing a timestamp). Dash Core's consensus target (nPowTargetSpacing) is 150s; Dash's own documentation commonly quotes 2.6 minutes (156s). We use the consensus value as nominal.

Ethereum (ETH)

215,172 blocks sampled Heights 25,424,496 to 25,639,667 Nominal 12 s 0.0% of blocks took more than twice the nominal time

Block interval	p10	Median	Mean	p90	p99
Seconds between blocks	12 s	12 s	12 s	12 s	12 s

Confirmations	Median time	p90 time
1	12 s	12 s
3	36 s	36 s
6	1 min 12 s	1 min 12 s
12	2 min 24 s	2 min 24 s

Post-merge Ethereum blocks land on fixed 12-second slots, so the interval distribution is discrete, not exponential -- every observed gap is a whole multiple of 12s. The full observed histogram is 12s: 214346 (99.6166%), 24s: 821 (0.3816%), 36s: 4 (0.0019%). A gap above 12s means a slot was missed: its proposer was offline or failed to publish in time. Because every percentile from p10 to p99 lands on 12s, the percentile table understates nothing -- it is genuinely that tight. pct_blocks_over_2x_nominal is near zero here for a mechanical reason and should not be read as 'Ethereum never stalls': one missed slot produces a gap of exactly 24s, which is not strictly greater than 2x nominal and so is excluded. Only gaps of 36s or more count. The honest worst-case figure for Ethereum is the 0.38% of intervals that exceeded one slot at all, and a maximum observed gap of 36s across the whole window.

Methodology

GriffNode operates its own full archival-index nodes for Bitcoin, Litecoin, Dogecoin and Dash, and a full Ethereum execution node (Erigon). For this dataset we read the block header of every block those nodes accepted over the trailing 30 days and used the header timestamps directly. Inter-block intervals are the difference between consecutive block timestamps. Time to N confirmations is measured as the timestamp difference between a block and the block N positions later, computed across every block in the window and then summarised -- it is a direct measurement of how long a payment first seen in one block waited to be buried under N blocks, not a nominal block time multiplied by N. Percentiles use linear interpolation. No smoothing, outlier removal or clamping is applied. The collection script is published alongside this dataset and re-runnable against any full node.

Sample window: last 30 days of blocks June 29, 2026 to July 29, 2026 (UTC) 293,960 blocks in total

Chain	Blocks sampled	First height	Last height
Bitcoin	4,175	955,954	960,128
Litecoin	17,234	3,133,537	3,150,770
Dogecoin	40,933	6,269,119	6,310,051
Dash	16,446	2,496,584	2,513,029
Ethereum	215,172	25,424,496	25,639,667

Caveats

These limits apply to every figure above. They are published because a number without its limits is not evidence.

Block timestamps on proof-of-work chains are declared by the miner, not observed by us. Consensus only requires a block's timestamp to exceed the median of the previous 11 blocks and to be within two hours of network-adjusted time, so individual intervals can be negative or inflated. We report raw values and disclose the negative count per chain rather than clamping them.
This measures time-to-confirmation from the moment a transaction is first included in a block. It does not include the time a transaction waits in the mempool before inclusion, which depends on the fee paid and can dominate the total wait during congestion.
A single trailing window is a snapshot of one hashrate and difficulty regime. Proof-of-work interval distributions shift when hashrate moves between difficulty adjustments, so a different 30-day window will produce somewhat different tail percentiles.
Ethereum after the Merge is not comparable in kind to the proof-of-work chains here. Its blocks land on fixed 12-second slots, so the interval distribution is discrete (multiples of 12s) rather than exponential, and its variance comes almost entirely from missed slots rather than from mining luck.
Measurements come from a single vantage point: our nodes. A node in a different network location may observe blocks marginally earlier or later, though header timestamps themselves are consensus data and identical everywhere.
Confirmation counts here are positional depth only. They do not model chain reorganisation risk, which is what confirmation requirements actually exist to bound.
pct_blocks_over_2x_nominal uses a strict greater-than. On the proof-of-work chains, where intervals are effectively continuous, this makes no practical difference. On Ethereum it does: one missed slot is a gap of exactly 24s, which is not strictly greater than 2x12s and is therefore excluded, driving the figure to near zero. Read Ethereum's per-chain note for the meaningful worst-case number instead.
The four proof-of-work chains carry different amounts of evidence: at 30 days Dogecoin contributes roughly ten times as many blocks as Bitcoin simply because its blocks are ten times more frequent. Bitcoin's tail percentiles (p99 in particular) rest on the fewest observations of any chain here and should be treated as the least stable figure in the dataset.

Data and citation

Download the full dataset as CSV ↓

This dataset is published under a Creative Commons Attribution 4.0 licence. You may republish the figures, including in news coverage and in AI answers, provided GriffNode is credited and the report is linked.

Cite this report

GriffNode (2026). "Measured crypto payment confirmation times". Measured 29 June 2026 to 29 July 2026. Retrieved from https://learn.griffnode.com/research/crypto-confirmation-times

Measured crypto payment confirmation times