A 2026 benchmark shows self-hosted Postgres on Hetzner beats AWS RDS on price/performance

A detailed benchmark compares AWS RDS, self-hosted Postgres on Hetzner, and a new managed provider. For cost-sensitive teams, the results show a clear winner outside the major cloud ecosystems.

The Answer Up Front

For bootstrapped teams with Linux administration skills, this benchmark makes a clear case: self-hosting PostgreSQL on a dedicated Hetzner server offers dramatically better performance per dollar than a comparable AWS RDS instance. If you have the operational capacity, the cost savings are significant. Teams who need the operational simplicity of a managed database but are priced out of RDS should consider newer providers building on bare metal. Skip self-hosting if you have no operations staff, are deeply integrated into the AWS ecosystem, or require complex multi-region failover capabilities that only a hyperscaler can provide. The bottom line is that for raw compute and storage, you pay a steep premium for the AWS brand and integration.

Methodology

This v0 review analyzes a single, public benchmark comparing three PostgreSQL hosting solutions, published in July 2026. The source is a blog post by the founder of Hostim, one of the services included in the comparison. We have not independently verified the results. Update cadence: this review will be updated if new, independent benchmarks emerge that contradict these findings.

• Subject: PostgreSQL Hosting Benchmark (July 2026)
• Source URL: https://hostim.dev/blog/postgres-benchmark-rds-vs-hostim-vs-self-hosted/
• What's Covered: The author's reported pgbench results (transactions per second, latency) and cost analysis for AWS RDS, a self-hosted Hetzner server, and the author's own managed service, Hostim.
• What's Not Covered: This review does not include our own independent testing. It does not evaluate long-term stability, security posture, backup and recovery processes, or the time-cost of system administration for the self-hosted option. The potential for author bias in the benchmark design is noted but not quantified.

The Benchmark Breakdown

The author, 'pv1337', compared three distinct environments for running a PostgreSQL database.

The contenders

The test compares three specific infrastructure configurations:

1. AWS RDS: A db.m5.large instance with 2 vCPUs, 8 GB RAM, and a 100 GB gp2 General Purpose SSD.
2. Self-Hosted on Hetzner: An AX41-NVMe dedicated server with an AMD Ryzen 5 3600 CPU (6 cores), 64 GB RAM, and two 512 GB NVMe SSDs in a RAID-1 configuration.
3. Hostim: The author's managed PostgreSQL service, running on the same Hetzner AX41-NVMe hardware as the self-hosted option.

The hardware mismatch is significant and central to the benchmark's findings. The Hetzner dedicated server is substantially more powerful than the RDS instance chosen for a similar price point.

The test conditions

The benchmark used pgbench, the standard load-testing tool included with PostgreSQL. The author reports using a scale factor of 100 (-s 100), running the test for 5 minutes with 8 clients and 8 threads. This simulates a typical online transaction processing (OLTP) workload.

The reported results

The performance differences reported are stark. The self-hosted Hetzner server achieved the highest throughput, followed closely by the managed Hostim service on identical hardware. The AWS RDS instance's performance was substantially lower.

• Transactions Per Second (TPS): Hetzner (self-hosted) reportedly achieved ~7,800 TPS. Hostim was close behind at ~7,650 TPS. AWS RDS lagged significantly at ~1,500 TPS.
• Average Latency: The Hetzner and Hostim setups both had reported latencies around 1.02-1.04 ms. AWS RDS latency was reported at 5.31 ms.

What's Interesting / What's Not

The most interesting takeaway is the quantification of the price/performance gap. While developers have long suspected that dedicated hardware from providers like Hetzner or OVH offers more raw power for the money, this benchmark puts concrete numbers to the claim. A 5x difference in throughput for roughly one-third the monthly cost is a compelling statistic for any founder watching their burn rate.

The inclusion of Hostim provides a useful data point for the 'managed bare metal' category. The small performance drop (~2%) compared to the self-hosted option is a reasonable trade-off for offloading server management, assuming the service is reliable.

What's not surprising is that a powerful dedicated server beats a small, virtualized database instance. The core trade-off has never been just about raw performance, but about total cost of ownership. The benchmark effectively ignores the operator cost: the hours spent on initial setup, security hardening, OS patching, database upgrades, and troubleshooting. For a solo founder or a team without a dedicated ops person, these hidden costs can easily exceed the monthly savings on hosting. The author, who sells a service to solve this exact problem, has a vested interest in highlighting the performance gap while minimizing the complexity of self-hosting.

Pricing

Pricing is based on the figures provided in the article. The author uses an exchange rate of €1 = $1.10.

• AWS RDS (db.m5.large): ~$135 / month (plus data transfer and IOPS costs)
• Hetzner (AX41-NVMe): €40 / month ($44 / month)
• Hostim (Managed on AX41-NVMe): Not explicitly stated in the signal, but competitive services are typically 2-3x the bare hardware cost.

Pricing snapshot from July 2026.

Verdict

This benchmark provides valuable data for founders evaluating their database hosting strategy. If your team possesses strong Linux and PostgreSQL administration skills and your primary driver is minimizing infrastructure costs, self-hosting on a provider like Hetzner is the clear winner on price/performance. The reported 5x performance at one-third the cost of AWS RDS is too large to ignore. However, this path requires a commitment to active, ongoing operational management. AWS RDS remains the default choice for enterprises and teams that prioritize developer velocity, ecosystem integration, and managed reliability above all else, and are willing to pay the associated premium. The analysis does not account for the labor cost of self-hosting, which remains the most critical variable in the equation.

What We'd Test Next

A v2 of this analysis would require independent verification and a broader scope. First, we would reproduce the pgbench results on identical hardware to verify the author's claims. Second, we would test failure and recovery scenarios: How long does it take to restore from a backup on each platform? What is the process for a point-in-time recovery? Third, we would expand the comparison to include other popular managed PostgreSQL providers like Neon, Crunchy Data, and Aiven to better map the entire market. Finally, we would model the total cost of ownership for the self-hosted option by tracking estimated hours for setup, maintenance, and emergency response over a six-month period.

The investor read

This benchmark highlights a persistent vulnerability for hyperscalers: the high margins on commodity infrastructure-as-a-service. The price/performance gap between AWS RDS and dedicated hardware from providers like Hetzner creates a durable market for budget-conscious startups and bootstrappers. Companies like Hostim, and the broader ecosystem of managed service providers built on cheaper bare metal, represent the continued unbundling of the cloud. While they won't capture the enterprise market, they can build profitable businesses by offering a 'good enough' managed experience at a fraction of the cost. This trend suggests a ceiling on hyperscaler pricing power for services that are not deeply integrated with their proprietary ecosystem. The key risk for investors in this space is the thin moat; it's a market defined by price competition and operational excellence, not deep IP.

Pull quote: “A 5x difference in throughput for roughly one-third the monthly cost is a compelling statistic for any founder watching their burn rate.”