A Solana node that is yours alone — no shared rate limit, because there is nobody to share it with. Yellowstone gRPC configured before handover, your choice of Agave, Jito-Solana or Frankendancer, and the option to sit on DoubleZero rather than the public internet.
In the last 24 hours we took 580,831 measurements of public endpoints across 3 chains from 4 regions, against a method we publish in full. Your node is measured by the same probes.
The evidence
Not a claim about our own hardware — we have not sold one yet, so we have no such figure and will not invent it. This is what our probes measured on the public Solana endpoints in the last day.
Fastest measured
53 ms
publicnode
Slowest measured
354 ms
solana-foundation
Spread
6.6×
between the two
Median response and availability over 24h, across the public Solana endpoints we measure, from 3 providers. Taken by our own probes against a published method — the same figures behind the Solana rankings, not a number written for this page.
The problem
The reasons a shared endpoint disappoints on this chain specifically, in the order they bite.
Solana's accounts database is read on every transaction and it does not fit in a cache you can hand-wave. A node that keeps it in RAM answers getProgramAccounts in milliseconds; one that reads it from disk answers in seconds, and the difference is memory rather than clock speed. Every plan here is sized for the accounts database to stay resident.
The ledger is written continuously while the accounts database is read continuously. Put them on the same NVMe and they compete for IOPS, the node drifts a few slots behind, and then a few hundred. Two drives is not a luxury on Solana; it is the difference between a node that keeps up and one that does not.
On a shared plan the 429 you get at 11am is not about your usage — it is about the aggregate, and you have no visibility into it and no lever on it. A dedicated node has no rate limit because there is nobody to ration against.
The advantage
An application on a shared endpoint crosses a balancer, the public internet and a queue it shares with strangers before its request reaches a node. Put it on a machine beside its own node and the path is a cable inside the rack — and that traffic is not charged.
What you get
Configured before handover unless it says otherwise. Where something is optional, the page says optional.
The Geyser plugin is compiled in and running when you receive the machine, with account, slot, transaction and block subscriptions on your private link and a token we issue you. Not a thing to ask for later and wait a week for — the difference between “available on request” and “configured” is a product difference, and we state which one you are getting.
We can attach the node to the DoubleZero fabric instead of routing over the public internet. Vote and block propagation are network problems before they are CPU problems, so this is where the milliseconds actually are. We will tell you honestly whether it helps your workload rather than selling it as a default.
Have an indexer that needs its own plugin? We will build it and run it — or tell you why it will not keep up at your subscription volume, which is more useful than discovering it in production.
A rebuild restores from a recent snapshot rather than replaying the ledger, so recovery is hours instead of days. This matters more on Solana than anywhere else.
How it is built
Your choice
A dedicated node is chosen on what software runs on it. This is the part a shared endpoint cannot offer at all.
Agave
The reference validator client, maintained by Anza. What we install unless you ask for something else, and the right answer for most RPC workloads.
Jito-Solana
Agave with the Jito block engine. For MEV-aware workloads and validators that want tips — which go to an account you nominate, with no cut taken by us.
Frankendancer
Firedancer's networking layer on the Agave runtime, for lower jitter. Available on request; we will talk through whether it suits what you are running first.
Pricing
Per month, exclusive of tax. The machine each one runs on is named, so the hardware can be checked against the price.
A dedicated Solana RPC node, yours alone, with the accounts database and the ledger on separate NVMe drives.
$1,290.00
per month · $12,900.00 a year, 16% less
For indexers and traders: twice the memory, more cores, and headroom for getProgramAccounts without starving the ledger.
$1,890.00
per month · $18,900.00 a year, 16% less
A dedicated Solana validator, operated for you. Vote account and stake remain entirely yours.
$3,490.00
per month · $34,900.00 a year, 16% less
Honestly
Every row is something you can check without asking us — including the one we lose.
| What you can check | RPCNode.IO | The rest of the market |
|---|---|---|
| Price published on the pageNot “contact sales”, and not behind a login. | Yes | Some publish, some quote on request |
| The exact machine namedProvider and SKU, so the hardware can be checked against the price. | Yes | Usually a tier name only |
| Independent measurement of rivalsWe rank every public endpoint on three chains, including our own. | Yes | None publish measurements of competitors |
| Latency claims you can checkEvery figure links to the method and the data behind it. | Yes | Numbers stated without a method |
| Yellowstone gRPC configured on handoverRunning when you receive the machine, not a later request. | Solana plans | Often an add-on or on request |
| Choose your clientAgave, Jito, Erigon, Nethermind, Reth and the rest. | Yes | Varies; often fixed |
| Pay in crypto against an on-chain oracleFeed address and round id printed on the invoice. | Yes | Card, or crypto at an unpublished rate |
| Track record on dedicated nodesWe are new at this. They are not. | Not yet | Years of production customers |
Included
Whichever chain you came for.
This site exists because we measure public RPC endpoints continuously and publish the method. Your dedicated node is measured by the same probes, from the same regions, the same way. Every other provider asks you to believe their latency figures; ours are produced by the apparatus we already point at their endpoints.
Requests to your node appear alongside your API-key traffic — volume, errors, p50 and p95, by endpoint. Not a separate dashboard bolted on, and no per-request charge behind it, because the node is yours and there is nothing to meter.
A private link between your machines, so an application never crosses the public internet to reach its own node. Put an indexer or a bot on an application server in the same rack and the hop is a cable.
It is your machine. Full root from the first minute, any operating system, and nothing of ours that has to stay installed. Monitoring is offered rather than imposed.
We place nodes in the cities our probes already run in, so the latency from that region is a figure we have been publishing rather than a number a datacentre sales page gave us. New York, Dallas, Frankfurt, Amsterdam, Singapore.
Every plan's price is published, along with the machine it runs on. The industry norm is “contact sales”, which exists so the number can vary by how much you look like you can pay.
Questions
A physical machine running a Solana validator client in RPC mode, allocated to one customer. Nobody else sends requests to it, so there is no shared rate limit, no noisy-neighbour latency and no credit metering. You choose the client, the region and the plugins.
Included and configured before handover on every Solana RPC plan. Account, slot, transaction and block subscriptions are exposed on your private link with a token we issue. There is no separate charge and no waiting list for it.
Yes, where the region supports it. DoubleZero is a dedicated fibre network for Solana traffic; attaching a node to it takes propagation off the public internet, which is where the latency actually is for vote and block delivery. Ask and we will set it up.
Jito-Solana is a standard option — pick it at order and tips route to an account you nominate. Frankendancer, which is Firedancer's networking layer on the Agave runtime, is available on request.
Hardware is allocated within hours of payment settling. The node then syncs, which is days on Solana — we restore from a recent snapshot rather than replaying the ledger to keep that short. Your node page shows the sync progress rather than telling you it is live before it is.
No. For a validator, the vote account and the stake are yours and we never hold the withdrawer key. For any machine, your own SSH public key is installed at provisioning and there is no private key of yours anywhere in our system.
The same probe fleet that measures every public Solana endpoint on this site measures yours, from the same regions, with the same published method. You are not taking our word for it — you are reading the same measurements anyone can read about anyone else.
Three reasons that are not marketing. There is no rate limit, because there is nobody to ration against. There is no noisy neighbour, so your p99 is your own workload rather than somebody else's. And you choose the client and the plugins, which on Solana and Ethereum changes what the node can do rather than only how fast it does it.
From what the machine costs us, published alongside what it is. We buy from Latitude, Terraswitch, CherryServers and 23m, and every plan names the provider SKU so you can check the hardware is what we say it is.
Yes — ETH, BNB or SOL, priced in dollars against an on-chain Chainlink feed. The invoice records the feed address and the round id, and the payment dialog publishes both, so the conversion can be checked against the chain rather than taken on trust.
The node keeps running until the end of the period you have already paid for, and the confirmation names the date. Nothing is destroyed early, and an unpaid invoice on a cancelled order is voided rather than left owing.
Not yet, and we are not going to invent one. This is a new product. What we do have is millions of measurements of public endpoints across three chains, taken continuously from four regions against a published method — and the same probes pointed at your node from day one. If that is not enough assurance for your workload, it should not be. Talk to us before you buy rather than after.
Prices are on the page and the specs are real. If your workload is unusual, say so before you buy rather than after — we would rather talk you out of the wrong machine than sell you one.
Solana rankings
New York
Frankfurt
Singapore
