Page load time extension results for popular websites

[SchumannFrequency]

I measured the load times of five of the most popular websites through an extension.
You can find this extension in the chrome web store.

Page load time

Displays page load time in the toolbar

chrome.google.com

I think it's nice to measure what the results are in reality instead of doing synthetic tests that very often measure things that aren't bottlenecks in reality.
I've taken screenshots of the specific results.

Instagram

493 ms

TikTok

3104 ms

facebook

489 ms

YouTube

885 ms

Twitter

540 ms

As hardware I use an 11 year old Intel i3 CPU. As software I use FreeBSD, bspwm and Chromium.
I am connected via an ethernet cable and I use the cheapest internet plan available in my country.

What result do you get and on which hardware, browser and connection?

 

[SchumannFrequency]

facebook

350 ms

Instagram

408 ms

These last two results are with different hardware.
Here I am using a thirteen year old Core i3 M 370 with a very slow HDD.

I don't really know why this system scores better than the faster system. I can only think of two reasons.
1. Brave is faster for these websites than Chromium with the extensions I use.
2. The thirteen year old laptop is much closer to the initial router. (the ethernet cable goes through far fewer rooms and also has fewer switching points)

I can improve my first 5 results with the FreeBSD system anyway by making a more direct connection to our router that passes through fewer rooms and fewer additional routers.

I think few people will know that a Core i3 M 370 CPU in 2023 can fully load Facebook in 350 ms which is only one third of a second.
These (seven) results use a connection that only scores 96 MB/s download speed on speedtest.net.

 

[SchumannFrequency]

AMD R5 PRO 3400G, dual-channel DDR4 @2666MHz, windows11 and Chrome 114.0.5735.110 (64-bits)

The ThinkCentre-M75s with the AMD R5 PRO 3400G has more powerful hardware and loads any website faster than the less powerful systems.

My suspicion is that this ThinkCentre-M75s has a very fast network card. The Realtek RTL8111EPV may be faster than the older network cards.

 

[SchumannFrequency]

I retested the i3-3240 system from the first 5 results today but with NixOS + Brave instead of FreeBSD + Chromium. These were the results.

When you add it up, it seems that FreeBSD will win the equation by being much faster at loading TikTok.

I chose Brave on NixOS because I know Chromium is slower on NixOS than on FreeBSD.
In synthetic benchmarks, Chromium is faster on FreeBSD than on NixOS.

This test with the most realistic results seems to say that FreeBSD is possibly the fastest system to explore the web with.
But a server is not under the same load every day, so you should still take these results with a grain of salt.

 

[SchumannFrequency]

The operating system, hardware, network card driver and browser have an impact on how fast websites load.

The biggest factors seem to be the number of routers to the server, speed of the server, load on the server and optimization/size of website itself.

 

[SchumannFrequency]

System that I use for the four following tests.
Hardware: Intel i3-3240 + 8GB DDR3 @1600MHz dual-channel + NVIDIA GTX 650 1GB + EVO 850 500GB
Software: Clear Linux + i3-wm + Brave browser + nouveau open-source GPU driver

These (four) results use a connection that only scores 96 MB/s download speed on speedtest.net.

 

[SchumannFrequency]

Today I did the test again with much faster hardware. The connection speed was the same as in all previous tests which makes all the results fairly comparable and shows the impact of software and hardware.

Software: ALT Sisyphus -- LXQt -- Chrome (beta version) / Chromium version 120 / Brave Nightly / Vivaldi (recent) snapshot -- proprietary Nvidia driver
Hardware: Intel 12600KF (stock) -- Kingston 6000 MHz CL40 -- GTX 650 1GB -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB

The faster hardware allowed me to get "a highest score" in three of the four results. If you have old hardware then websites load slower.

Since I tested four different browsers, it was also a good browser comparison. I noted the following observations.
-Chromium 120 achieved the best result in three of the four websites (and the website where it did not score the highest it was very close with the fastest browser for that website)
-Brave Nightly was the slowest on YouTube
-For Facebook, Brave Nightly was faster than Chrome Beta and Vivaldi snapshot
-Vivaldi snapshot was the slowest in Instagram
-For TikTok, it was hard to notice differences between browsers.

 

[SchumannFrequency]

I used the same extension to evaluate Firefox's speed. Here you can see the results for some websites that run faster on Firefox than on Chrome.

Software: ROSA Fresh Desktop 12.4 -- LXQt -- Nvidia proprietary driver -- XFS as root -- Firefox 122.0a1
Hardware: Intel 12600KF (stock) -- Kingston 6000 MHz CL40 -- GTX 650 1GB -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB
I am connected via an ethernet cable and I use the cheapest internet plan available in my country.

 

[lemonadesoda]

My load times:
instagram 0ms
Facebook 0ms
Tiktok 0ms

… because like any sensible expert, DNS is first resolved by my firewall before asking the external Cloudfare DNS 1.1.1.1 for the address. And my firewall knows very well that any attempts to those websites needs to land at 127.0.0.1

Job done

 

[SchumannFrequency]

This is my result for the popular website https://gamegpu.com/

Hardware: Intel 12600KF (stock) -- Kingston 6200 MHz CL36 -- Sapphire RX 7600 -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB
Software: Alpine Linux, sway wm, Mesa open-source driver, Firefox 124.0.1

I then reloaded the page 30 times, first on Firefox, and then on Chromium.
My observation is that Chromium is on average about 35% slower than Firefox for this specific website.

 

[SchumannFrequency]

Hardware: Intel 12600KF (stock) -- Kingston 6200 MHz CL36 -- Sapphire RX 7600 -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB
Software: Alpine Linux, sway wm, Mesa open-source driver, Firefox 124.0.1

These results use a connection that only scores 98 MB/s download speed on speedtest.net.

The two websites are in the top 40 most popular websites. Both seem to load faster on Firefox. Chromium seems to be a little slower.

The maximum speed I get for ebay is about 0.7 seconds with Firefox.

The fastest load speed for VK is about 0.45 seconds with Firefox.

I am in a location quite far from urban areas and I also have a slow internet subscription and I am not directly connected to the router, there are some unnecessary extra switching points.

It would be interesting if I were to take the fastest possible internet subscription and see if this has much effect on the speeds. But the averages reported for 2023 are strangely very slow.

If industry-wise averages are checked
On average, automotive retail websites take six seconds to load.
Consumer packaged goods websites load in an average of 6.1 seconds.
On average, financial websites require 5.1 seconds to load.
On average, healthcare websites require 5.6 seconds to load.
On average, media websites require 5.5 seconds to load.
On average, retail websites take six seconds to load.
On average, technology websites take 6.8 seconds to load.
On average, travel websites take 6.7 seconds to load.

The average web page load time is 2.5 seconds on desktop and 8.6 seconds on mobile, based on our analysis of the top 100 webpages worldwide.

 

[SchumannFrequency]

AccuWeather: 225 ms

Fox News: 257 ms

BBC: 131 ms

The New York Times: 254 ms

Hardware: Intel 12600KF (stock) -- Kingston 6200 MHz CL36 -- Sapphire RX 7600 -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB
Software: Alpine Linux, sway wm, Mesa open-source driver, Firefox 124.0.1

These are the exact nPerf connection statistics on which the above results are based.

 

[bonehead123]

Nice work with some interesting comparisons, but as demonstrated, any number of variables & combo's of hardware, software, connections, time of day etc can have either minor or major effects on the results, so they are just that...comparisons

 

[cvaldes]

One thing I've noticed since the emergence of the smartphone is that the smartphone version of a web service may actually be faster than the PC-focused browser-based version.

Clearly, based on the number of client devices, a lot of these services are accessed far more frequently via smartphone than PC so web developers are more focused on optimizing mobile device performance than PC browser performance.

This has been blatantly evident since the mid-2010s.

Of course, a lot of the basic framework is already in the smartphone app itself. It doesn't need to load the entirety of its code every time you access the site in question.

So when you install YouTube on your iPhone (320+ MB), you're pre-emptively downloading a bunch of code that a typical web browser must load on a frequent basis, especially if you're like me and browse the web in incognito mode.

Using web browser based benchmarks is far less pertinent if you actually own and use a smartphone.

Even at home, often I access many commonly used services via a smartphone or tablet rather than a PC. It is far quicker logging into my bank using FaceID or TouchID than typing in a password on a PC-based web browser; the device is well credentialed. And for location-based services, you will get better accuracy using a smartphone anyhow due to GPS circuitry, GPS triangulation services, Wifi triangulation services much more than a static Ethernet connection which may have no verifiable location status.

I don't even need to use my primary iPhone. I have an older iPhone that I use around the home for this sort of thing as well as an iPod touch.

A lot of social media services are heavily optimized for mobile users. We saw this transition about ten years ago with wider adoption of smartphones. Remember that Twitter was primarily accessed via PC-based web browsers in its earliest years. Twitter let PC desktop client support fall off in subsequent years including its own first-party PC desktop client as well as third-party solutions like Tweetdeck.

In 2024 there are precious few web services that actually run faster on desktop PC than mobile apps. Using desktop web browser benchmarks to measure this stuff has limited applicability in this era.

Remember that even on something like a banking site, on a phone you just tap the icon and it normally starts the user authentication process. On a web browser, you need to mouse over to the bookmark then find the "login" button (where ever it may be today) *before* starting the authentication process which may require another hand movement (like a fingerprint scan). It all adds up to the access times. It's not just about a page loading.

 

[SchumannFrequency]

Hardware: Intel 12600KF (stock) -- Kingston 6200 MHz CL36 -- Sapphire RX 7600 -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB
Software: Calculate Linux, KDE Plasma, Mesa open-source driver, XFS file system, Firefox 126.0

This is a test of Firefox 126 released four days ago. I see higher performance for the above four websites.
It could also be due to lower load on the server or acceleration of the server.
There are also some websites for which I measure lower performance than before in Firefox 126, especially the website The New York Times.

 

[SchumannFrequency]

Hardware: Intel 12600KF (stock) -- Kingston 6200 MHz CL36 -- Sapphire RX 7600 -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB
Software: OpenBSD, FVWM, open-source GPU driver, FFS file system, Chromium 124.0.6367.207, Firefox 126.0

The above websites are all four in the top 10 most popular websites. I see no significant difference between OpenBSD and Linux for these four websites.
I feel that the results of these websites depend more on how heavily loaded the server is than on the operating system you are using.

There are some websites where I did see that OpenBSD was significantly slower than Linux, but this seems to be only +- 20% of all websites.

As far as I am concerned, OpenBSD has better audio quality than windows10/11 and Linux.
If you watch a lot of YouTube/Spotify/TikTok/Netflix/SoundCloud and similar websites then OpenBSD may be a finer browsing experience than windows10/11 and Linux.