CompareNodes provides independent and transparent RPC benchmarking services for blockchain nodes and endpoints. Our default regions are North America, Europe and Asia.
On February 3rd, 2025 we benchmarked AWS’ Managed Blockchain service for Ethereum mainnet. AWS offers 6 sizes and our goal was to determine the peak RPC capacity for the 4 larger instance types: c5.4xlarge, m5.4xlarge, c5.2xlarge and m5.2xlarge. Instances range in size from 4 vCPU & 16GB to 16 vCPU & 64 GB which is why we have arranged their results in pairs by the count of vCPUs (see each table below).
Our definition for Peak RPC Capacity = max RPC requests per second and 100% successes (no fails, no timeouts).
In total, we dispatched ~8 million RPC requests. You can learn more about our methodology and use cases for RPC benchmarking-as-a-service.
Benchmarking Configuration
Configs were almost identical for all runs:
- k6 running on EC2 c5.9xlarge;
- Ethereum mainnet in AWS US-East-1;
- Target RPC-scenarios/second: 3000 or 5000.
- 10 stages: 10% increase until 100% (of target RPC-scenarios/second);
- Each stage consist of 2 phases: 30 seconds ramp-up and 60 seconds sustain;
- Total duration: 15 minutes;
- Timeouts: 10 seconds.
Our assumption was that a “8 vCPU” instance might have a similar capacity as another “8 vCPU” instance and both might have much lower capacity than a “16 vCPU” instance. Therefore, we used different values for the Target RPC-scenarios/second depending on the instance vCPU count.
RPC Scenarios & RPC Requests
Our “RPC-scenarios” correspond to 23 popular RPC methods. A target config of “1000” will run 1000 RPC-scenarios per second but will dispatch more than 1000 RPC requests per second because some RPC scenarios dispatch additional 1 or 2 RPC requests on each cycle. Additional RPC requests are for eth_blockNumber/eth_getBlockByNumber because scenarios “follow” the top of the chain during the course of the benchmark run and send “auxiliary” RPC requests to obtain new values for their dynamic parameters.
For this analysis we used 23 RPC scenarios like eth_call and eth_getLogs. In each case, we assigned them a weight to mimic a realistic usage distribution (example). Since Amazon Managed Blockchain does not support “batching” at this time, we set it to”0.0” and increased eth_getLogs to “0.0535” instead.
Each benchmarking run computes various metrics for every scenario as well as all scenarios and requests as a whole. In the tables below, we show a subset of latencies for a subset of scenarios with highest weight values. However, you can navigate to corresponding dashboards in Grafana to explore complete results.
Showdown Results
One of our goals was to determine the approximate peak RPC capacity-without-fails for each instance type. These values are shown as “Breaking points” in the tables below. Also, we show a latencies for 5 of the most frequent RPC scenarios: eth_call, eth_blockNumber, eth_getBalance, eth_getTransactionReceipt and eth_getLogs. Complete latencies by method are in the screenshots on the bottom of the page.
C5.4xlarge vs. M5.4xlarge
Our findings are that C5’s greater clock speed beats M5’s larger memory for RPC processing. Over the course of ~2.5 million requests, M5 “fails” about twice as much as the C5 instance. Moreover, the C5 is able to run at 100% successes up to about 4.65k requests/sec: see “breaking point” and screenshots below or open the Grafana dashboard link to 23 panels.
Amazon Managed Blockchain C5.4xlarge | Amazon Managed Blockchain M5.4xlarge | |
Grafana Dashboard | 23 panels | 23 panels |
vCPU | 16 | 16 |
RAM GB | 32 | 64 |
Target RPC-scenarios/sec | 5000 | 5000 |
Peak RPC requests/sec | 5.17 k | 5.21 k |
Total RPC requests | 2.51 mil | 2.52 mil |
Fails Pct | 3.21% | 6.33% |
Timeouts Pct | 0% | 0% |
Successes Pct | 96.79% | 93.67% |
“Breaking” point: Peak RPC requests/sec | 4.64 k RPC/sec | 3.87 k RPC/sec |
Overall Latencies 23 methods Median p95 | 22.3 ms 153 ms | 29.4 ms 191 ms |
eth_call: 33% Median p95 | 20.7 ms 117 ms | 25 ms 152 ms |
eth_blockNumber: 21% Median p95 | 20.5 ms 114 ms | 24.5 ms 147 ms |
eth_getBalance: 8.15% Median p95 | 20.6 ms 118 ms | 24.7 ms 153 ms |
eth_getTransactionReceipt: 7% Median p95 | 21.4 ms 116 ms | 26 ms 150 ms |
eth_getLogs: 5.35% Median p95 | 132 ms 512 ms | 155 ms 678 ms |
C5.4xlarge breaking-point is the peak requests/sec before fails/timeouts: ~4.64 k

While M5.4xlarge starts failing ~3.87 k

C5.2xlarge vs. M5.2xlarge
In the case of the 2 instance types with 8 vCPUs, we see that C5 holds a performance edge over the M5 again. Also, we can compare them to their respective 4xlarge “siblings” and see that scaling works well. Doubling vCPUs from 8 to 16 achieves an increase from 2.53 k breaking-point to 4.64 k breaking-point which is ~1.83x for C5 and ~1.9x for M5 series.
Amazon Managed Blockchain C5.2xlarge | Amazon Managed Blockchain M5.2xlarge | |
Grafana Dashboard | 23 panels | 23 panels |
vCPU | 8 | 8 |
RAM GB | 16 | 32 |
Target RPC-scenarios/sec | 3000 | 3000 |
Peak RPC requests/sec | 3.19 k | 3.21 k |
Total RPC requests | 1.55 mil | 1.56 mil |
Fails Pct | 10.06% | 16.15% |
Timeouts Pct | 0% | 0% |
Successes Pct | 89.94% | 83.85% |
“Breaking” point: Peak RPC requests/sec | 2.53 k RPC/sec | 2.04 k RPC/sec |
Overall Latencies 23 methods Median p95 | 27.7 ms 333 ms | 41.1 ms 448 ms |
eth_call: 33% Median p95 | 11.6 ms 287 ms | 33.7 ms 370 ms |
eth_blockNumber: 21% Median p95 | 23.6 ms 278 ms | 32.8 ms 355 ms |
eth_getBalance: 8.15% Median p95 | 23.7 ms 288 ms | 33.6 ms 376 ms |
eth_getTransactionReceipt: 7% Median p95 | 24.3 ms 281 ms | 34.3 ms 363 ms |
eth_getLogs: 5.35% Median p95 | 133 ms 1.36 s | 207 ms 1.82 s |
C5.2xlarge peaks without fails at ~2.53 k RPC requests / sec

… and M5.2xlarge achieves ~2.04 k RPC requests/sec before any fails:

Schedule Your Benchmark
This article is an example of our RPC benchmarking service. Additional data is available in private Grafana and we can publish it on multiple platforms, upon your request.
Please get in touch with our CTO Simon to schedule a consultation about your own private or public benchmark via this Calendly link. Alternatively, learn more about our setup process & steps.
Scenarios For 23 RPC Methods
Amazon Managed Blockchain C5.4xlarge

Amazon Managed Blockchain M5.4xlarge

Amazon Managed Blockchain C5.2xlarge

Amazon Managed Blockchain M5.2xlarge
