Robinhood Chain RPC comparison, measured every few minutes

Response time

One eth_blockNumber call. The solid bar is the median; the pale extension is the slowest 10%, which is the part you actually notice. Shorter is better.

Ordo

104 ms

median 2 ms · slowest 10% 104 ms · best 0 ms · 71.9% answered · at the chain head

Robinhood official

106 ms

median 91 ms · slowest 10% 106 ms · best 71 ms · 0.7% answered · 7 blocks behind

PublicNode

108 ms

median 94 ms · slowest 10% 108 ms · best 82 ms · 38.3% answered · at the chain head

BlockMachine

509 ms

median 179 ms · slowest 10% 509 ms · best 165 ms · 0.6% answered · at the chain head

ArrowRPC

no response

no successful call in 24h

What each one can do

Speed is not the only thing that matters, and we are not best at all of it.

Endpoint	Call tracing	Archive state	Answered	Head
Ordo	no	no	71.9%	current
Robinhood official	no	no	0.7%	−7
PublicNode	no	no	38.3%	current
BlockMachine	no	no	0.6%	current
ArrowRPC	no	no	—	—

Not on this chart: Private Send, pre-send simulation and the approval-drain guard are Ordo product features. They are not latency, they are not independently measured on the other endpoints, and they are not a reason this table would put us first.

Method

So that this is checkable rather than claimed.

Every endpoint gets the identical call, eth_blockNumber, from the same machine, in the same cycle.
Six samples per cycle; the first is discarded for every endpoint alike, because it pays for the TLS handshake.
Chain heads are read from all endpoints concurrently. Reading them in sequence on a chain that produces a block every ~100 ms would simply flatter whichever was asked last.
Failures are counted, not retried away. The "answered" column is what it sounds like.
Ordo is reached over its public URL, not over localhost. There is still a home-field effect: that URL terminates on this machine, so our own latency is not what a wallet in another city sees. Everyone else is a real trip from here. Treat the ranking as a comparison of the other endpoints from this host, not as a promise of 30 ms from yours.
One probe per endpoint per cycle. These are other people's servers.

curl -s https://rpc.ordofi.network \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"eth_blockNumber","params":[]}'

Machine-readable: /compare.json · /llms.txt · attested volume: dune.com/jemordo/ordofi-rpc