Everything you need.A direct line to Solana.
RPC, Geyser gRPC, and Direct Shreds run on the network right next to Solana. You talk to the chain directly.
Trusted infrastructure
Speed is determined by distance to Solana.
The region matters. But a city name alone is not a network path. Even in the same city, external transit and extra hops can add latency. ERPC selects data centers close to Solana servers, then keeps routes short with no external transit and sustained turbo boost.
ERPC premium route
no external transit
minimum RTT
City-name selection
external transit / 7 hops
70x slower
- 01Same data centerChoose the region first, then the data center closest to Solana.
- 02No external transitAvoid external AS paths to keep RTT and jitter tight.
- 03Full throttleERPC excludes power-saving profiles and keeps resources in sustained turbo boost.
Measured benchmark
Same-class machine. Not the same speed.
AMD Turin, 4 vCPU, Amsterdam, Ubuntu 24.04 — the same on both boxes, one major cloud and ours. The spec sheet says they’re equals. The bench says otherwise.
Same silicon. The gap is our tuning — the box dialed in and seated right next to Solana.
node_bench · same run, same region
ERPC vs a major cloud · same spec
- CPU computesysbench · 4 threads · higher is better
- more CPU throughput1.9×
- Memory bandwidthSTREAM Triad · 4 GiB · higher is better
- more memory bandwidth3.2×
- Disk IOPSfio · 4K randread · QD32 · higher is better
- more disk IOPS16.6×
- Disk latency · p99fio · 4K randread · QD32 · lower is better
- faster at the tail25.7×
events/s
MB/s
IOPS
µs
Stake-weighted priority (SWQoS)
If your transactions keep failing, you’re stuck in the spam lane.
Solana’s leaders split priority bandwidth in two. Connections backed by stake — SOL committed to a validator — get 80% of it. Everyone else fights over the other 20% — the lane that’s packed with spam.
Firing straight at the leader feels like the fast move — but with no stake, that’s the crowded 20% lane, and under load your transaction never makes the block.
So the real answer is a staked validator. We run a top-tier one wired into high-quality RPC lines — hardware placed right next to Solana — so your transactions get through the wide lane.
A top-tier validator in our Shinobi Performance Pool
How Solana’s leaders split priority bandwidth
Stake rides the wide 80% lane into the validator — before any fee. No stake, you’re crammed in the 20% spam lane.
80 / 20 split set by Solana’s leaders, not by us
Core Node Locations
Built where Solana runs fastest.
Across our global footprint, we select only premium data centers where Solana runs at its best, then run RPCs, streams, and dedicated nodes on top-tier hardware at full throttle. We also offer bare metal servers and VPS options so you can deploy apps on the same network and stay on the fastest paths.
The ERPC stack
Every layer runs right next to Solana.
- 01300+ edge locationsRPC EndpointHTTP & WebSocket, multi-region
- 02Push, no pollingGeyser gRPCAccount, tx, slot & block streams
- 03One layer before gRPCDirect ShredsData before the block is assembled
- 0480% priority bandwidthStaked ConnectionSWQoS priority lane
- 051.9× CPU vs Google CloudDirect VPS ServersEPYC compute right next to Solana
- 06Snapshot 2–4h → ~5minDirect Bare Metal ServersExclusive physical resources
- 07~2ms · −20ms TLSUnlimited EndpointsDedicated HTTP, no TLS overhead
- 08CLI agent, local & remoteSLV AINatural-language Solana ops
You're already next to Solana.
RPC, Geyser gRPC, Direct Shreds — the whole stack, racked on the network beside the chain. Start free and your key is live in a minute.