The finding that held
Enabling PHP OPcache on the Apache stack, with nothing else changed, moved empty-cart latency from 1,183 ms to 172 ms. Opcode caching mattered far more than the web server.
Ahosting Technical Operations test
Apache vs LiteSpeed for WooCommerce
Re-measured with both stacks configured the same way, the picture changed. Cached pages were tied, and the biggest factor on cart and checkout was not the web server but whether PHP OPcache was enabled.
First measured 24 August 2026, re-measured 28 August 2026 after we found the stacks were not comparable.
24 Years
Trusted Since 2002
99.9% Uptime
Guaranteed SLA
30-Day
Money-Back Guarantee
24/7
Expert Helpdesk
Executive summary
What we found, and what we took back
We tested the same WordPress 7.1, WooCommerce 11.0.1 and 50-product catalog on controlled Apache and LiteSpeed sites, both under the published WooStart limits. After the first publication we found the two stacks were not configured comparably, corrected it, and measured again.
Cached storefront
Home and product pages were tied on both stacks. On cache-eligible pages there is no difference worth claiming, so we do not claim one.
The original number
Our first version reported LiteSpeed cutting cart and checkout latency by 55 to 64 percent. It did not survive re-measurement. The tables below are the corrected ones.
Corrected results
Cart and checkout, with both stacks configured the same
With OPcache enabled on both, the dynamic store paths landed between 171 ms and 318 ms, and the Apache node was ahead on absolute numbers. That ordering reflects hardware, not the web server: the Apache host indexes at 34 ms of PHP compute against 137 ms on the LiteSpeed host.
Median response time by store path
Lower is better
Apache
LiteSpeed
| Scenario | Apache | LiteSpeed | Outcome |
|---|---|---|---|
| Cached home | 13 ms | 12 ms | Effectively tied |
| Cached product | 12 ms | 12 ms | Tied |
| Empty cart | 172 ms | 286 ms | Apache ahead |
| Filled cart | 171 ms | 290 ms | Apache ahead |
| Filled checkout | 174 ms | 294 ms | Apache ahead |
| Add to cart | 191 ms | 318 ms | Apache ahead |
| Cart fragments (wc-ajax) | 86 ms | 125 ms | Apache ahead |
| Store API cart | 90 ms | 136 ms | Apache ahead |
The correction
One missing extension explained the whole gap
The first version of this study reported that LiteSpeed reduced cart and checkout latency by 55 to 64 percent. That result did not survive re-measurement. The Apache lab simply had no PHP OPcache installed, so every request recompiled the full WooCommerce codebase. Installing it, with no other change, produced this.
Apache stack before and after enabling OPcache
Lower is better
No OPcache
OPcache enabled
| Scenario | First run, no OPcache | Re-measured, OPcache on | Improvement |
|---|---|---|---|
| Empty cart | 1,183 ms | 172 ms | 6.9x |
| Filled cart | 722 ms | 171 ms | 4.2x |
| Filled checkout | 696 ms | 174 ms | 4.0x |
| Add to cart | 683 ms | 191 ms | 3.6x |
Equal account limits
Both accounts ran under the same WooStart ceiling
Neither test account was given more resource than a paying WooStart customer gets. The hosts underneath them, however, were not the same machine.
- CPU
- Up to 3 cores
- Memory
- 3 GiB
- Entry processes
- 40
- Disk I/O
- 20 MiB/s
- Web server
- Apache 2.4.67
- PHP
- 8.2.33 PHP-FPM
- Page cache
- WP Super Cache
- Database
- Local MariaDB
- Web server
- LiteSpeed Ent 6.3.6
- PHP
- 8.2.33 LSAPI
- Page cache
- LiteSpeed Cache
- Database
- Remote MySQL
- Apache CPU
- Xeon Silver 4214R
- LiteSpeed CPU
- Xeon Silver 4114
- PHP index
- 34 ms vs 137 ms
That last row is the one that matters. The Apache host completes the same PHP work in roughly a quarter of the time. Any absolute comparison between these two nodes is measuring the processors as much as the web servers.
Method and limits
What this study does not prove
Published so you can judge how much weight the numbers deserve.
- The two nodes are not hardware-identical, so absolute latencies describe these environments, not the web servers in isolation.
- Cart, checkout and account routes must stay outside full-page cache, so no caching result applies to them.
- A synthetic 50-product catalog is not a real store: themes, extensions and payment gateways all add time.
- The Michigan source is origin-proximate, 0.2 ms from both nodes, so these are HTTP response times rather than worldwide visitor latency or Core Web Vitals.
- The PDF edition still carries the superseded 24 August figures and is being reissued. The tables on this page are current.
Looking for general WordPress performance instead? The WordPress study measures dynamic, cached and concurrent requests on the same platform.
Measured, then sold
The stack in this study is the one you get
WooStart runs the same LiteSpeed and LSCache configuration measured above. It starts at $3.79 a month and renews at $3.79, with a free dedicated IP, daily offsite backups and a migration our team runs for you.
- LSCache tuned so cart and checkout stay dynamic
- Free dedicated IP included
- Free expert migration
- 30-day money-back period
WooStart starts from
$3.79 /month
36-month term. Renews at $3.79/mo, no increase and no setup fee.