Settings beyond the standard defaults in the WP-Rocket caching plugin were tested for their effect on site speed on Ancient Greek Keyboard. The control consisted of Siteground’s Static Cache running in the background.
Methods
Each test was performed using Pingdom.com’s Website Speed Test. The San Jose, California location option was used initially. (For the second half of the tests, Pingdom.com changed its system & the San Francisco, California location was the only similar option). Each test was conducted with the HTTP URL rather than the HTTPS URL.
For each setting, 9 tests were run during a single day. (The number of tests per setting was deliberately kept low to reduce the total testing time.)
Load time results were recorded, and averaged at the end of the study. While typically, only settings which improved speed would be kept, in these tests the small sample size created random error & it was arbitrarily decided that even settings which led to a reduction of speed as low as -0.05s would still be used. (Since reduction of speed was essentially close to zero, anyways.)
Results
The following HTML tables were generated by the Tableizer Excel to HTML conversion site:
Table 1. Page Load Speed with Setting: Cache > Separate Cache Files for Mobile Devices
Date |
Time |
Load time of control (sec) |
Load time of setting (sec) |
9/20/2018 |
16:16 |
1.210 |
1.450 |
|
|
1.200 |
1.240 |
|
|
1.360 |
1.370 |
|
21:15 |
1.210 |
1.520 |
|
|
1.490 |
1.450 |
|
|
1.410 |
1.430 |
|
22:39 |
1.360 |
1.230 |
|
|
1.270 |
1.250 |
|
|
1.330 |
1.310 |
n |
|
9 |
9 |
Average |
|
1.316 |
1.361 |
Std Dev |
|
0.101 |
0.108 |
Analysis: The Separate Cache Files for Mobile Devices setting harmed load speed by 0.046 seconds (1.316 seconds for the control vs 1.361 seconds for the setting).
While this reduction of performance was less than the -0.05 second threshold, this setting WILL NOT be used in the future because the site serves little content on mobile devices that is not served to desktop devices.
Table 2. Page Load Speed with Setting: File Optimization > Minify HTML
Date |
Time |
Load Time of control (sec) |
Load Time of setting (sec) |
9/21/2018 |
12:54 |
1.360 |
1.360 |
|
|
1.410 |
1.390 |
|
|
1.400 |
1.360 |
|
16:52 |
1.490 |
1.210 |
|
|
1.210 |
1.220 |
|
|
1.280 |
1.390 |
|
19:46 |
1.370 |
1.430 |
|
|
1.210 |
1.050 |
|
|
1.250 |
1.140 |
n |
|
9 |
9 |
Average |
|
1.331 |
1.283 |
Std Dev |
|
0.098 |
0.132 |
Analysis: The Minify HTML setting improved load speed by 0.048 seconds (1.331 seconds for the control vs 1.283 seconds for the setting).
This setting WILL be used in the future. While it doesn’t seem to improve page load speed much, it doesn’t appear to harm it, either.
Table 3. Page Load Speed with Setting: File Optimization > Remove Query Strings
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
9/22/2018 |
13:03 |
1.540 |
1.280 |
|
|
1.420 |
1.450 |
|
|
1.280 |
1.270 |
|
16:59 |
1.120 |
1.420 |
|
|
1.250 |
1.420 |
|
|
1.370 |
1.240 |
|
22:39 |
1.350 |
1.290 |
|
|
1.220 |
1.260 |
|
|
1.500 |
1.270 |
n |
|
9 |
9 |
Average |
|
1.339 |
1.322 |
Std Dev |
|
0.136 |
0.082 |
Analysis: The Remove Query Strings setting improved load speed by 0.017 seconds (1.339 seconds for the control vs 1.322 seconds for the setting).
This setting WILL be used in the future. While it doesn’t seem to improve page load speed much, it doesn’t appear to harm it, either.
Table 4. Page Load Speed with Setting: File Optimization > Minify CSS
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
9/23/2018 |
13:08 |
1.220 |
1.530 |
|
|
1.210 |
1.480 |
|
|
1.480 |
1.570 |
|
15:55 |
1.360 |
1.770 |
|
|
1.350 |
1.500 |
|
|
1.220 |
1.740 |
|
21:16 |
1.500 |
1.630 |
|
|
1.230 |
1.080 |
|
|
1.240 |
1.840 |
n |
|
9 |
9 |
Average |
|
1.312 |
1.571 |
Std Dev |
|
0.115 |
0.224 |
Analysis: The Minify CSS setting harmed load speed by 0.259 seconds (1.312 seconds for the control vs 1.571 seconds for the setting).
This setting WILL NOT be used in the future since it causes a significant loss of page load speed.
Table 5. Page Load Speed with Setting: File Optimization > Combine CSS
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
9/24/2018 |
12:16 |
1.400 |
1.550 |
|
|
1.340 |
1.430 |
|
|
1.340 |
1.420 |
|
14:34 |
1.220 |
1.490 |
|
|
1.350 |
1.110 |
|
|
1.360 |
1.330 |
|
16:36 |
1.320 |
1.310 |
|
|
1.230 |
1.190 |
|
|
1.440 |
1.450 |
n |
|
9 |
9 |
Average |
|
1.333 |
1.364 |
Std Dev |
|
0.071 |
0.143 |
Analysis: The Combine CSS setting harmed load speed by 0.031 seconds (1.333 seconds for the control vs 1.364 seconds for the setting).
This setting WILL NOT be used in the future, despite its negative impact on page speed being insignificant (ie: it fell below the -0.05 second threshold).
The reason for rejecting this setting is because it breaks some of the Child Theme styles.
Table 6. Page Load Speed with Setting: File Optimization > Optimize CSS delivery
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
9/25/2018 |
19:40 |
1.430 |
1.250 |
|
|
1.160 |
1.120 |
|
|
1.140 |
1.110 |
|
22:09 |
1.260 |
1.030 |
|
|
1.290 |
1.320 |
|
|
1.280 |
1.150 |
|
|
1.040 |
1.190 |
|
|
1.070 |
1.280 |
|
|
1.050 |
1.120 |
n |
|
9 |
9 |
Average |
|
1.191 |
1.174 |
Std Dev |
|
0.133 |
0.093 |
Analysis: The Optimize CSS delivery setting improved load speed by 0.017 seconds (1.191 seconds for the control vs 1.174 seconds for the setting).
This setting WILL NOT be used in the future. Although it had a minor positive impact on page speed, it required 2 IPs to be whitelisted.
Table 7. Page Load Speed with Setting: File Optimization > Minify JS
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
9/26/2018 |
18:33 |
1.140 |
1.170 |
|
|
1.110 |
1.260 |
|
|
1.290 |
1.070 |
|
21:12 |
1.320 |
1.090 |
|
|
1.280 |
1.120 |
|
|
1.350 |
1.110 |
|
22:47 |
1.180 |
1.300 |
|
|
1.330 |
1.160 |
|
|
1.190 |
1.190 |
n |
|
9 |
9 |
Average |
|
1.243 |
1.163 |
Std Dev |
|
0.089 |
0.077 |
Analysis: The Minify JS setting improved load speed by 0.080 seconds (1.243 seconds for the control vs 1.163 seconds for the setting).
This setting WILL be used in the future because it has a major positive impact on page speed.
Table 8. Page Load Speed with Setting: File Optimization > Combine JS
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
9/27/2018 |
12:12 |
1.240 |
1.200 |
|
|
1.240 |
1.150 |
|
|
1.240 |
1.310 |
|
15:32 |
1.280 |
0.991 |
|
|
1.180 |
1.110 |
|
|
1.290 |
1.100 |
|
18:02 |
1.070 |
1.310 |
|
|
1.110 |
1.200 |
|
|
1.110 |
1.220 |
n |
|
9 |
9 |
Average |
|
1.196 |
1.177 |
Std Dev |
|
0.081 |
0.102 |
Analysis: The Combine JS setting improved load speed by 0.019 seconds (1.196 seconds for the control vs 1.177 seconds for the setting).
This setting WILL be used in the future. While it doesn’t seem to improve page load speed much, it doesn’t appear to harm it, either.
Table 9. Page Load Speed with Setting: File Optimization > Load JS Deferred
Date |
Letter Grade |
Load Time of control (sec) |
Load Time with setting (sec) |
9/28/2018 |
86 |
1.200 |
1.210 |
|
86 |
1.300 |
1.460 |
|
86 |
1.110 |
1.160 |
|
86 |
1.300 |
1.160 |
|
84 |
1.460 |
1.310 |
|
86 |
1.330 |
1.230 |
|
86 |
1.310 |
1.310 |
|
86 |
1.330 |
1.210 |
|
86 |
1.130 |
1.190 |
n |
|
9 |
9 |
Average |
|
1.274 |
1.249 |
Std Dev |
|
0.110 |
0.097 |
Analysis: The Load JS Deferred setting improved load speed by 0.026 seconds (1.274 seconds for the control vs 1.249 seconds for the setting).
This setting WILL be used in the future. While it doesn’t seem to improve page load speed much, it doesn’t appear to harm it, either.
Table 10. Page Load Speed with Setting: File Optimization > Load JS deferred + Safe mode for Jquery
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
9/29/2018 |
10:47 |
1.340 |
1.310 |
|
|
1.240 |
1.490 |
|
|
1.480 |
1.410 |
|
13:03 |
1.240 |
1.230 |
|
|
1.320 |
1.340 |
|
|
1.430 |
1.250 |
|
15:00 |
1.220 |
1.450 |
|
|
0.804 |
1.480 |
|
|
1.600 |
1.230 |
n |
|
9 |
9 |
Average |
|
1.297 |
1.354 |
Std Dev |
|
0.224 |
0.106 |
Analysis: The Load JS deferred + Safe mode for Jquery setting harmed load speed by 0.057 seconds (1.297 seconds for the control vs 1.354 seconds for the setting).
This setting WILL NOT be used in the future, since it harmed page load speed beyond the -0.05 second threshold.
Table 11. Page Load Speed with Setting: Media > Lazy Load for Images
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
9/30/2018 |
10:18 |
1.500 |
1.380 |
|
|
1.520 |
1.360 |
|
|
1.460 |
1.440 |
|
12:46 |
1.300 |
1.200 |
|
|
1.470 |
1.160 |
|
|
1.250 |
1.310 |
|
15:31 |
1.440 |
1.260 |
|
|
1.300 |
1.540 |
|
|
1.630 |
1.490 |
n |
|
9 |
9 |
Average |
|
1.430 |
1.349 |
Std Dev |
|
0.123 |
0.129 |
Analysis: The Lazy Load for Images setting improved load speed by 0.081 seconds (1.430 seconds for the control vs 1.349 seconds for the setting).
This setting WILL be used in the future because it has a major positive impact on page speed.
Table 12. Page Load Speed with Setting: Media > Lazy Load for iFrames / Videos
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
10/1/2018 |
10:59 |
1.370 |
1.320 |
|
|
1.300 |
1.550 |
|
|
1.470 |
1.360 |
|
13:56 |
1.190 |
1.360 |
|
|
1.210 |
1.270 |
|
|
1.220 |
1.440 |
|
16:29 |
1.360 |
1.370 |
|
|
1.350 |
1.300 |
|
|
1.490 |
1.250 |
n |
|
9 |
9 |
Average |
|
1.329 |
1.358 |
Std Dev |
|
0.109 |
0.092 |
Analysis: The Lazy Load for iFrames / Videos setting harmed load speed by 0.029 seconds (1.329 seconds for the control vs 1.358 seconds for the setting).
This setting WILL be used in the future, since the reduction of performance was less than the -0.05 second threshold.
Table 13. Page Load Speed with Setting: Media > Replace YouTube iframe with preview image
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
10/2/2018 |
10:57 |
1.380 |
1.290 |
|
|
1.220 |
1.470 |
|
|
1.240 |
1.240 |
|
15:47 |
1.450 |
1.480 |
|
|
1.210 |
1.400 |
|
|
1.240 |
1.220 |
|
17:44 |
1.240 |
1.190 |
|
|
1.380 |
1.390 |
|
|
1.110 |
1.380 |
n |
|
9 |
9 |
Average |
|
1.274 |
1.340 |
Std Dev |
|
0.107 |
0.108 |
Analysis: The Replace YouTube iframe with preview image setting harmed load speed by 0.066 seconds (1.274 seconds for the control vs 1.340 seconds for the setting).
This setting WILL NOT be used in the future, since the reduction of page load speed exceeds the -0.05 second threshold and the site does not currently use any YouTube videos or iFrames.
Table 14. Page Load Speed with Setting: Preload > Sitemap-based cache preloading
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
10/3/2018 |
10:57 |
1.390 |
1.220 |
|
|
1.170 |
1.420 |
|
|
1.200 |
1.270 |
|
13:15 |
1.420 |
1.510 |
|
|
1.280 |
1.430 |
|
|
1.390 |
1.210 |
|
15:55 |
1.360 |
1.210 |
|
|
1.210 |
1.340 |
|
|
1.200 |
1.300 |
n |
|
9 |
9 |
Average |
|
1.291 |
1.323 |
Std Dev |
|
0.099 |
0.109 |
Analysis: The Sitemap-based cache preloading setting harmed load speed by 0.032 seconds (1.291 seconds for the control vs 1.323 seconds for the setting).
This setting WILL NOT be used in the future. Although the reduction of performance was less than the -0.05 second threshold, this setting caused cron jobs to “crawl” the site every 12 hours (cluttering up the site’s statistics).
Table 15. Page Load Speed with Setting: Preload > Preload bot (automatic)
Date |
Time |
Load Time of control (sec) |
Load Time with setting (sec) |
10/4/2018 |
12:01 |
1.460 |
1.440 |
|
|
1.250 |
1.450 |
|
|
1.230 |
1.250 |
|
14:11 |
1.380 |
1.240 |
|
|
1.600 |
1.530 |
|
|
1.270 |
1.300 |
|
17:11 |
1.270 |
1.600 |
|
|
1.390 |
1.210 |
|
|
1.290 |
1.300 |
n |
|
9 |
9 |
Average |
|
1.349 |
1.369 |
Std Dev |
|
0.121 |
0.140 |
Analysis: The Preload bot (automatic) setting harmed load speed by 0.020 seconds (1.349 seconds for the control vs 1.369 seconds for the setting).
This setting WILL be used in the future, since the reduction of performance was less than the -0.05 second threshold.
Conclusions
Unusable Settings
The following settings for WP-Rocket will not be used on Ancient Greek Keyboard because they either cause a reduction of page load speed more than the -0.05 second threshold, or require outside IPs to be whitelisted, or initiate undesirable cron jobs which clutter up the site’s statistics, or break the site’s CSS:
- Cache > Separate Cache Files for Mobile Devices
- File Optimization > Minify CSS
- File Optimization > Combine CSS
- File Optimization > Optimize CSS delivery
- File Optimization > Load JS deferred + Safe mode for Jquery
- Media > Replace YouTube iframe with preview image
- Preload > Sitemap-based cache preloading
Usable Settings
The following settings will be used on Ancient Greek Keyboard because they either improve page load speed (or at the very least, do not harm it significantly):
- File Optimization > Minify HTML
- File Optimization > Remove Query Strings
- File Optimization > Minify JS
- File Optimization > Combine JS
- File Optimization > Load JS Deferred
- Media > Lazy Load for Images
- Media > Lazy Load for iFrames / Videos
Preload > Preload bot (automatic) [See Nov 8, 2018 update]
UPDATE (Nov 8, 2018): The recent update of WP Rocket (vers. 3.2) no longer uses a bot for Preload, so this option no longer exists.
The replacement will not be used however, since it tends to clutter up the statistics log.
Other posts in this series:
- Siteground Speed Test (With & Without Static Caching)
- Speed Test of Swift Performance Lite plugin
- Caching Plugin Speed Tests: Swift Performance Lite vs. W3 Total Cache
- Speed Test: W3 Total Cache With Autoptimize
- Speed Test of WP-Rocket plugin (default settings)
- Speed Test of WP-Rocket plugin (additional settings)
- WP Rocket Speed Test (with & without SiteGround Static Caching)