Results of the study on the impact on browser performance of the most popular extensions for Chrome. It has been shown that some extensions can significantly affect performance and create a heavy load on the system, as well as drastically increase memory consumption. The tests evaluated CPU load creation in both active and background modes, memory usage, and the impact on the speed of displaying opened pages. The results are presented in two samples covering 100 and 1000 of the most popular extensions.
Among the 100 most popular extensions, the most CPU-intensive are the 'Evernote Web Clipper' (4 million users) and 'Grammarly' (10 million users), which lead to an additional 500 ms of processor time consumption when opening each page (for comparison, opening the test site without extensions takes 40 ms).
In general, 20 extensions consume more than 100 ms, while 80 consume less than 100 ms. Surprisingly, the Ghostery extension shows relatively high resource consumption, using 120 ms of CPU time. The LastPass password manager consumed 241 ms, and Skype consumed 191 ms. These resource usages do not stop rendering but obstruct the start of interaction with the page and affect the device's energy consumption.
In the sample of 1000 extensions, there are extensions that create a significantly higher load:
In the page rendering delay test, extensions such as Clever, Grammarly, Cash Back for Shopping, LastPass, and AVG delayed loading by 150-300 ms, in some cases introducing delays comparable to the rendering of the page itself. Overall, the situation is normal, as out of 100 extensions, only 6 caused delays of more than 100 ms.
Results of the sample of 1000 extensions:
When assessing the CPU load generated by background operations performed by the extension, the extension demonstrated
Avira Browser Safety, which consumed nearly 3 seconds of CPU time, while the costs of other add-ons did not exceed 200 ms. Since the background mode is usually used to process network requests made during the page opening, the test was repeated on the apple.com website, which makes 50 requests instead of one. The results changed, and Ghostery became the leader in load creation, while Avira Browser Safety dropped to 9th place (analysis showed that the load decreased because the apple.com site was whitelisted).
Test results for 1000 add-ons:
The bottom 20 results in testing 1000 add-ons:
Since users often blame low performance and delays on the browser rather than on installed add-ons, Google experiments to inform about problematic add-ons. In the stable release of Chrome 83, a setting "chrome://flags/#extension-checkup" was introduced, which includes informative messages about the possible impact of add-ons on privacy and performance. When the option is enabled, a warning starts appearing on the new tab page and in the add-ons manager that add-ons may lead to significant resource consumption or have access to personal data and user activity.
A separate comparison was conducted of add-ons for ad blocking and privacy protection, focusing on resource savings by blocking external scripts and ad inserts. All add-ons demonstrated a reduction in load of at least three times when processing a test article from one of the news websites. The leader was the DuckDuckGo Privacy Essentials add-on, which reduced the load time for opening the test page from 31 seconds to 1.6 seconds of CPU time by decreasing the number of network requests by 95% and the amount of data downloaded by 80%. uBlock Origin showed a similar result.
DuckDuckGo Privacy Essentials and uBlock Origin also excelled in resource consumption evaluation during background operations.
In testing memory consumption, DuckDuckGo Privacy Essentials and uBlock Origin reduced memory usage from 536 MB during full processing of the test page to ~140 MB.
Similar testing was conducted for web developer add-ons. CPU load:
CPU load during background operations
Rendering delays:
Memory consumption:
Source: opennet.ru
