OpenSSL Performance Benchmark

The dramatic improvement from OpenSSL 3.1.x to 3.2.x (~2-2.5× for RSA, ~1.5× for ECDSA) is real and expected. OpenSSL 3.2 (released November 2023) included significant optimizations:

• Provider Architecture Overhead Reduction: OpenSSL 3.0/3.1 introduced a provider-based architecture with significant per-operation dispatch overhead. Version 3.2 dramatically reduced this overhead.
• RSA Operations: RSA signing (the bottleneck in RSA handshakes) benefited from reduced context setup and provider lookups per operation.
• ECDSA/ECDH Operations: Elliptic curve operations saw improvements in both the provider layer and underlying implementations.
• TLS Path Optimizations: Streamlined the full handshake code path, reducing CPU cycles per connection.

Note: The 3.0/3.1 performance regression vs 1.1.1w was a known issue. OpenSSL 3.2 marked the beginning of recovery, with 3.4+ often matching or exceeding 1.1.1w performance.

👉 See full analysis of the 3.1 → 3.2 performance improvement →

The dramatic performance improvement from OpenSSL 3.1.x to 3.2.x (~2-2.5× in some configurations) is real and expected. OpenSSL 3.2 (released November 2023) included major optimizations:

• Provider Architecture Optimizations: Reduced per-operation dispatch overhead that accumulated in the 3.0/3.1 transition from the legacy API
• Context Caching Improvements: Better reuse of cryptographic contexts, reducing repeated initialization costs during handshakes
• TLS Session Handling: Streamlined session ticket processing and key derivation paths
• General Performance Work: Accumulated fixes addressing performance regressions identified since 3.0

Bottom Line: The 3.2 release marked a turning point where OpenSSL 3.x handshake performance began approaching—and in some cases exceeding—1.1.1w levels. This improvement is consistent across independent benchmarks.

👉 See full analysis of the 3.1 → 3.2 performance improvement →