Ethereum RPC Provider Benchmark: Latency & Uptime (September 2026)

By Chainstack Labs Research Team · Research Note No. 045 · September 2026 edition · Published Sep 24, 2026 · Last updated Sep 24, 2026 · Node Infrastructure

Data source: live Grafana compare dashboards, refreshed every 3 minutes, 7-day rolling window across DE, US, and SG.

---

Abstract

Ethereum has the deepest bench of RPC providers of any chain, which makes it the hardest place to separate marketing claims from measured performance — every vendor can point to a region or a percentile where it wins. This report ranks four providers — Chainstack, Alchemy, Quicknode, and dRPC — on Ethereum RPC latency and uptime, using a composite score that weights response time against success rate independently in three regions (Frankfurt, US, Singapore) and then averages the result, so a provider cannot post a strong global number by being fast in one region while degraded in another. Chainstack ranks first with a score of 0.086 and the lowest global P95 (86ms) in the set. The more consequential finding is a single-region tail event: dRPC’s US p99 spikes to 1,370ms — more than 6x its own DE p99 — pulling its composite score to 0.541, over six times Chainstack’s. A blended global average would report this as “dRPC is somewhat slower”; the regional breakdown shows it is Chainstack-competitive in two regions and badly degraded in the third. For a look at how RPC method choice compounds this on Ethereum specifically, see our companion report, RPC Method Latency Across Chains.

Keywords: Ethereum RPC provider, Ethereum RPC latency, RPC benchmark, regional latency, availability, composite scoring, tail latency, provider comparison

---

Contents

1. Introduction
2. Methodology and scoring
3. Providers and regions
4. Results
5. Regional variance and outliers
6. Interpretation notes and known confounders
7. Conclusion
8. Appendix A: measurement parameters

---

1. Introduction

Ethereum mainnet has more RPC vendors competing for traffic than any other chain, and a decade of that competition has narrowed median latency between top providers to single-digit milliseconds in a well-served region. That narrowing is exactly why a single global-average number is least useful here: on a young chain, one provider’s global average is a fair proxy for “is this vendor competent at all”; on Ethereum, the vendors that survive are all competent on average, and the number that actually separates them is what happens on their worst day, in their worst region. This report ranks four Ethereum RPC providers — Chainstack, Alchemy, Quicknode, and dRPC — on exactly that basis, publishing the regional and tail-latency detail that a single blended figure would average away.

Ethereum’s size also raises the cost of getting this wrong. A tail-latency event on a thin chain affects a handful of early integrators; a tail-latency event on Ethereum can sit behind a production trading system, a wallet backend, or an indexer serving thousands of users, which is why this report treats p99 and single-region spikes as first-class findings rather than footnotes to the median.

2. Methodology and scoring

Each provider is evaluated on speed (response time) and reliability (success rate) independently in each monitored region, then combined as:

Score = ResponseTime / SuccessRate³, averaged across regions. Lower is better.

Cubing the success-rate term makes reliability the dominant factor once availability drops meaningfully below 100% — a provider that is fast but flaky is penalized far more than the raw percentage-point gap in its availability figure would suggest. Response time itself is not immune to this either: because the score averages across regions rather than blending raw request counts, a single region with a severe latency spike — as with dRPC’s US result in this round — pulls the average up sharply rather than being smoothed out by two well-behaved regions. Per-method p95 latencies are combined with equal weights to produce the response-time term. Scores are read directly from the underlying Grafana compare dashboards, which update every 3 minutes, rather than recomputed downstream from a static export — the numbers in this report are a snapshot of that live view, not a separately batched dataset.

3. Providers and regions

Four providers offering public Ethereum RPC endpoints were included: Chainstack, Alchemy, Quicknode, and dRPC. Each was measured concurrently from three regions — Frankfurt (DE), a US test node, and Singapore (SG) — against the same Ethereum mainnet RPC methods, over a rolling 7-day window refreshed every 3 minutes. All figures reflect each provider’s standard shared endpoint, not a dedicated or archive-node tier — the tier most teams evaluate first before committing to paid dedicated infrastructure. Chainstack’s Ethereum node deployment options include dedicated and archive tiers that were out of scope for this round.

4. Results

Table 1. Composite score, availability, and global percentile latency by provider, 7-day window. Chainstack posts the lowest composite score (0.086) and the lowest global P95 (86ms); dRPC’s score is more than six times Chainstack’s, driven by a single-region P99 spike rather than a uniformly slower average.

Provider	Score	Availability	P50	P95	P99
Chainstack	0.086	99.93%	41ms	86ms	140ms
Alchemy	0.123	99.96%	37ms	123ms	209ms
Quicknode	0.178	99.94%	81ms	173ms	253ms
dRPC	0.541	99.90%	52ms	576ms	1,370ms

Table 2. P95 latency by region, milliseconds. Chainstack and Alchemy hold every region under 190ms; dRPC’s US column is the standout anomaly in this report, over an order of magnitude worse than its own DE result.

Provider	DE	US	SG
Chainstack	54ms	160ms	43ms
Alchemy	71ms	113ms	186ms
Quicknode	123ms	171ms	224ms
dRPC	127ms	1,360ms	241ms

5. Regional variance and outliers

Table 1 alone would describe dRPC as the clear last-place provider by a wide margin, and Table 2 shows why that framing is incomplete: dRPC’s DE (127ms) and SG (241ms) P95 figures are within range of Quicknode’s, not in a different tier. The entire gap is concentrated in one region — dRPC’s US P95 of 1,360ms is more than ten times its DE figure, and its global P99 of 1,370ms in Table 1 is almost entirely explained by that single regional result rather than a general tendency toward slow responses. This is the clearest illustration in this report series of why regional detail matters more than a global average: a team running exclusively from Frankfurt or Singapore would see dRPC performing roughly in line with Quicknode: a team routing any US traffic through it would see a materially different, and much worse, product.

Chainstack’s own regional spread is not perfectly flat either — its US P95 (160ms) is nearly 4x its SG figure (43ms) — but the absolute numbers stay within a range that keeps its composite score lowest in the set. Alchemy shows the inverse pattern to Chainstack, with its best region in DE (71ms) and its worst in SG (186ms), while posting the second-lowest composite score overall; its regional spread is real but far narrower than dRPC’s.

6. Interpretation notes and known confounders

Shared-endpoint tier only. All figures reflect each provider’s standard shared RPC endpoint. A dedicated node in the affected region would be the direct mitigation for a regional outlier like dRPC’s US result, and this comparison does not capture whether that gap persists on a paid dedicated tier.

A single-region spike can dominate the composite score. dRPC’s result in this round is a useful worked example of the scoring formula’s behavior: one severely degraded region pulled its average score to 0.541 even though two of its three regions were competitive with Quicknode. Readers should check Table 2 before drawing conclusions from Table 1 alone, on any provider, in any round of this report.

Method mix is not disclosed. The P95 figures combine multiple RPC methods with equal weights. As our RPC Method Latency Across Chains benchmark shows, method choice can move Ethereum latency by an order of magnitude on its own — eth_getLogs in particular is a known outlier method — so a workload concentrated in one heavy method may see different relative rankings than this blended figure suggests.

Single measurement window. Figures reflect one rolling 7-day window. Whether dRPC’s US spike is a persistent regional issue or a transient incident inside this window is not something a single snapshot can distinguish; a follow-up edition of this report will show whether it recurs.

7. Conclusion

On Ethereum, the providers in this comparison are closer on average than the composite score alone suggests — the real story this round is a single-region tail event, not a uniform quality gap. Chainstack holds the lowest score and the tightest overall regional spread; Alchemy is a close second with its own moderate regional variance; Quicknode is consistent but not fast in any region; dRPC is competitive in two of three regions and severely degraded in the third. Teams choosing an RPC provider for Ethereum should weight the region their traffic actually originates from more heavily than the global ranking, and specifically check the US column in Table 2 before routing production US traffic through any provider based on a global average alone.

Appendix A: measurement parameters

Parameter	Value
Chain	Ethereum mainnet
Providers	Chainstack, Alchemy, Quicknode, dRPC (shared endpoint tier)
Regions	DE (Frankfurt), US, SG (Singapore)
Measurement window	Rolling 7 days, refreshed every 3 minutes
Score formula	ResponseTime / SuccessRate³, averaged across regions
Latency basis	Per-method P95, equal-weighted
Source	Live Grafana compare dashboards