Why test without caching plugins on Apache and Nginx?

Two reasons. First, LiteSpeed's LSCache is built into the server itself, not bolted on as a plugin. Comparing 'plain Apache + WordPress' against 'plain LiteSpeed + WordPress (LSCache enabled by default)' is the fair out-of-the-box comparison most agencies actually deploy. Second, the second test phase added WP Fastest Cache to Apache and Nginx specifically to close that gap and see what cached performance looks like on each.

Why Hetzner CPX11 specifically?

It's a representative entry-level VPS: 2 vCPU, 2GB RAM, NVMe storage, on a 10Gbit network. The specs that most agencies actually deploy WordPress on for small-to-mid-size client sites. Bigger machines would shift the absolute numbers up but the ratios between the three servers stay similar.

Wouldn't a tuned Nginx beat default LiteSpeed?

Possibly, with significant manual work: tuning PHP-FPM workers, configuring fastcgi_cache, adding a reverse-proxy cache layer, optimizing keepalive. The point of the benchmarks isn't 'no Nginx setup can match LiteSpeed'; it's 'on the default WordPress install most agencies actually run, LiteSpeed wins by a wide margin without per-site tuning'.

How does WP Fastest Cache compare to LSCache?

WP Fastest Cache is a PHP-level cache: it generates static HTML files for cached pages. LSCache is a server-level cache: the cache lookup happens before WordPress's PHP even starts. The architectural difference shows up in the numbers: WP Fastest Cache improves Apache and Nginx by 10-20x on cached requests, but LiteSpeed's server-level cache is still 2-5x faster on top of that.

Should I always pick LiteSpeed based on these numbers?

For WordPress and PHP-heavy workloads, the case is strong. For static-only sites, API gateways, or reverse-proxy deployments, Nginx is the right answer. The benchmarks reflect WordPress workloads specifically; don't generalize the LiteSpeed advantage to every web-server scenario.

Web Server Performance Benchmarks: LiteSpeed Enterprise vs Apache vs Nginx

Web server choice is one of those decisions agencies usually inherit instead of make. You set up Apache because that’s what cPanel ships with. You set up Nginx because that’s what every “10x your performance” tutorial recommends. Whether either of them was the right pick for the workload nobody actually re-tested.

This is the benchmark I ran to settle that question for myself, on identical hardware, on a vanilla WordPress site. The setup, the numbers, and the conclusions.

Test environment

The same hardware and configuration across every test:

Server: Hetzner CPX11 (2 vCPU, 2GB RAM, 40GB NVMe SSD, 10Gbit network).
Control panel: Enhance.
WordPress: latest release, default theme, PHP 8.1, no Redis, no extra plugins, out-of-the-box configuration.
Web servers tested: Apache and Nginx without their native caching plugins, LiteSpeed Enterprise with the default LiteSpeed Cache plugin enabled at default settings.
Load tester: loader.io.
Test type: Maintain Client Load. Sustained traffic from 1 client/second up to a maximum specified rate over 5 minutes.
Failure threshold: 1%.

The “Maintain Client Load” test ramps from 1 client/second to the maximum configured rate continuously over 5 minutes. It’s the realistic shape of a traffic spike (sustained, increasing, with the system getting hotter as it goes), not a stress test that fires N requests instantly.

Phase 1: no caching plugin

The first phase tests how each server handles increasing load without any caching layer in front of WordPress. The exception is LiteSpeed, where LSCache ships built into the server and is enabled by default. That’s the honest “out of the box” comparison: nobody adds a caching plugin to LiteSpeed because it doesn’t need one.

Test 1: 1 to 350 clients/second over 5 minutes

Apache.

Apache benchmark Test 1 with no cache showing average response time of 4083ms and CPU saturating at 100%

Average response time: 4,083ms
Min/max response time: 134ms / 9,237ms
Successful responses: 12,797
CPU usage: 100%

Nginx.

Nginx benchmark Test 1 with no cache showing average response time of 4422ms and CPU saturating at 100%

Average response time: 4,422ms
Min/max response time: 136ms / 9,050ms
Successful responses: 11,798
CPU usage: 100%

LiteSpeed Enterprise.

LiteSpeed Enterprise benchmark Test 1 with default LSCache showing average response time of 95ms and CPU usage of 25-35%

Average response time: 95ms
Min/max response time: 95ms / 494ms
Successful responses: 542,049
CPU usage: 25-35%

The gap is roughly 40x in successful responses (542,049 vs 12,797) and roughly 40x in average response time (95ms vs 4,083ms). Apache and Nginx both saturate CPU at 100%; LiteSpeed sits comfortably under 35% CPU on the same workload.

Test 2: 1 to 700 clients/second over 5 minutes

Pushing the load to 700 clients/second:

Apache: 5,244ms average, 8,118 successful responses, 162 timeouts, started timing out at 390 clients/second, CPU 100%.
Nginx: 5,254ms average, 7,115 successful responses, 363 timeouts, started timing out at 410 clients/second, CPU 100%.
LiteSpeed Enterprise: 95ms average, 1,114,793 successful responses, 0 timeouts, CPU 35-50%.

Apache and Nginx both fail well below the requested load. LiteSpeed serves over a million requests successfully without a single timeout.

Test 3: 1 to 2,400 clients/second over 5 minutes (LiteSpeed only)

LiteSpeed Enterprise benchmark Test 3 at 2400 clients per second showing 1.6 million successful responses and zero timeouts

Apache and Nginx couldn’t reach this load level on this hardware. LiteSpeed:

Average response time: 213ms
Min/max response time: 95ms / 2,151ms
Successful responses: 1,675,476
Timeouts: 0
CPU usage: 35-100% (with automatic throttling at around 1,000 clients/second to prevent timeouts)

The throttling behavior is the interesting part. LiteSpeed proactively rate-limits incoming requests at the server level when CPU pressure rises, instead of letting requests queue and time out. Apache and Nginx don’t have this built in.

Phase 1 takeaways

40-100x performance gap in raw throughput between LiteSpeed and Apache/Nginx without caching.
CPU efficiency: LiteSpeed handles roughly 100x the workload at 50% the CPU cost.
Reliability under stress: LiteSpeed has zero timeouts at loads where Apache and Nginx are dropping hundreds of requests.
Automatic throttling: LiteSpeed’s request management prevents the cascade-failure shape you see in Apache and Nginx under sustained load.

Phase 2: with caching plugin (Apache and Nginx)

Adding WP Fastest Cache to Apache and Nginx to close the gap.

Test 1: 1 to 350 clients/second over 5 minutes

Apache + WP Fastest Cache.

Apache with WP Fastest Cache Test 1 showing 184ms average response time and 286,741 successful responses

Average response time: 184ms
Successful responses: 286,741
Timeouts: 60 (started timing out at 260 clients/second)
CPU: 5-100% (hit 30% at 50 clients/sec, 100% at 275 clients/sec)

Nginx + WP Fastest Cache.

Nginx with WP Fastest Cache Test 1 showing 526ms average response time and 101,305 successful responses

Average response time: 526ms
Successful responses: 101,305
CPU: 20-100% (hit 100% at just 50 clients/sec)

Test 2: 1 to 700 clients/second over 5 minutes (with cache)

Apache: 184ms average, 286,741 successful responses, 328 timeouts (started at 275 clients/sec), CPU 100% by 275 clients/sec.
Nginx: 1,050ms average, 100,968 successful responses, 7 timeouts (started at 690 clients/sec), CPU 100% by 50 clients/sec.

Phase 2 takeaways

Caching helps a lot. Apache went from 12,797 successful responses (no cache) to 286,741 (with WP Fastest Cache) on Test 1. That’s a 22x improvement.
Apache wins on response time with caching enabled, but Nginx scales further before timing out.
Neither catches LiteSpeed. LiteSpeed without any plugin still serves 1.5x the requests of Apache + WP Fastest Cache, with response times an order of magnitude lower.

What this means for real WordPress sites

The honest reading of these numbers:

For high-traffic WordPress, LiteSpeed wins by enough to matter. The 40-100x throughput gap is unusually large for benchmark tests. Most “X is faster than Y” benchmarks land in the 1.2-2x range; this is a different category of difference.
CPU efficiency matters more than raw throughput. LiteSpeed handling 1.6M requests at 35-100% CPU means a smaller, cheaper VPS can handle the same workload. Hardware savings compound with the license fee.
Caching plugins help, but they don’t fix architecture. WP Fastest Cache makes Apache and Nginx survive moderate load. They still don’t reach LiteSpeed’s numbers, because LSCache is integrated at the server level instead of running as a PHP plugin.
These numbers reflect default configurations. A heavily-tuned Nginx with PHP-FPM workers, fastcgi_cache, and a Cloudflare layer would close some of the gap, at the cost of significant ongoing maintenance. Most agencies don’t have that maintenance bandwidth.

For more on the architecture behind these numbers, see my LiteSpeed Enterprise writeup, which covers the LSCache plugin, QUIC.cloud integration, and where LiteSpeed isn’t the right answer.

Closing the loop

For WordPress workloads on entry-level hardware, LiteSpeed Enterprise is in a different performance category than Apache or Nginx. The numbers are reproducible on identical hardware, and the architectural reasons (server-level caching, event-driven worker model, LSAPI for PHP) explain why.

If you’re running WordPress at scale and the performance numbers aren’t where you’d like them, the Cloud Infrastructure Audit & Hardening engagement covers the LiteSpeed migration as part of the managed setup. For more on the cloud infrastructure side, the cloud infrastructure category has the rest, and the LiteSpeed Enterprise overview covers the why behind these numbers.