Serious Discussion Chrome vs. Chromium: Differences in Benchmarks

[Miravi]

I've repeatedly encountered assertions and evidence that Google Chrome outperforms open-source Chromium and its various forks, but I hadn't really done comparative testing on my own. Now that I'm running a Fedora Linux desktop, I have copies of both Google Chrome and a vanilla Chromium build maintained by the Fedora Project itself, so I explored this subject a little further.

Here are the optimizations—the "secret sauce"—that Google applies to their own binaries derived from Chromium:

• PGO (Profile-Guided Optimization): This is the single biggest factor for Speedometer scores. Google runs internal "traces" of common user workflows (like the tasks in Speedometer) and feeds that data back into the compiler. This allows the compiler to optimize the "hot" paths of the code specifically for those real-world tasks.
• ThinLTO (Thin Link-Time Optimization): Chrome is built with ThinLTO enabled across the entire browser. This allows for cross-module optimizations, enabling the compiler to inline functions and eliminate dead code across different parts of the browser that are usually siloed during a standard build.
• Advanced V8 Tiers: Chrome often ships with the latest "Maglev" and "Turbofan" JIT compiler configurations tuned for specific hardware. While Chromium has the same code, Google’s builds are often compiled with specific flags that prioritize execution speed over binary size or build time.
• Binary Size & Layout: Google uses proprietary tools to optimize the physical layout of the binary on your disk and in memory (reordering functions so that frequently used ones are physically close together), reducing instruction cache misses.

And for the sake of completeness, here are some unique advantages of Chrome on Linux:

• Clang vs. GCC: Google builds Chrome for Linux using a very specific, highly patched version of the Clang compiler. Many Linux distros build Chromium using GCC to match their system's toolchain, which often lacks the specific LLVM-based optimizations (like certain ThinLTO implementations) that V8 favors.
• Sandboxing & Overhead: Chrome’s multi-process architecture on Linux is tuned for the specific kernel namespaces Google targets. Third-party Chromium builds (especially Flatpaks or Snaps) may introduce slight overhead due to additional containerization layers that Chrome's native .deb or .rpm packages avoid.
• Hardware Acceleration (VA-API): Chrome has historically had better "out of the box" integration for GPU-accelerated video decoding on Linux. While Chromium supports this, it often requires manual flag toggling (like --enable-features=VaapiVideoDecoder) or specific patches that vary by distribution.
• Memory Management (PartitionAlloc): Chrome on Linux uses a highly tuned version of PartitionAlloc (Google’s custom memory allocator). While this is in Chromium, Google’s production builds are often tuned with specific bucket sizes and "StarScan" (an algorithm to mitigate use-after-free) that are optimized for the memory layouts of modern Linux kernels.

I chose to compare them using Speedometer, which is intended as a web app responsiveness benchmark by simulating real-world user interactions. Set up with NVIDIA's proprietary Linux drivers and zero extensions (guest profile), I ran three Speedometer 3.1 tests in succession with each and calculated an average.

Chrome produced a 37.44% higher score. (Chrome: ~31.57, Chromium: ~22.97)

 

[Parkinsond]

Chrome is the standard-of-industry; each other browser is evaluated compared to.

 

[Sorrento]

IHO I still don't feel that benchmarks such as Speedometer relate at least here relate to every day browsing, secondly using Chrome you have to jump through hoops to even get it to delete browsing data on exit (why? because because google just don't want you too) & if after going through menus & sub menus it still isn't effective - Seems the advantages in privacy etc of using Linux are wiped out by using Chrome in the first place when there are browsers out there that at least respect our wishes? at least to some degree - Just my 10 pence worth?

 

[Sampei.Nihira]

I've repeatedly encountered assertions and evidence that Google Chrome outperforms open-source Chromium and its various forks, but I hadn't really done comparative testing on my own. Now that I'm running a Fedora Linux desktop, I have copies of both Google Chrome and a vanilla Chromium build maintained by the Fedora Project itself, so I explored this subject a little further.

Here are the optimizations—the "secret sauce"—that Google applies to their own binaries derived from Chromium:


And for the sake of completeness, here are some unique advantages of Chrome on Linux:


I chose to compare them using Speedometer, which is intended as a web app responsiveness benchmark by simulating real-world user interactions. Set up with NVIDIA's proprietary Linux drivers and zero extensions (guest profile), I ran three Speedometer 3.1 tests in succession with each and calculated an average.

Chrome produced a 37.44% higher score. (Chrome: ~31.57, Chromium: ~22.97)

Yes, when I stopped using Edge, I chose Chrome because it was the fastest Chromium-based browser.

For a more comprehensive test, you can also check the DOMContentLoaded value (in seconds) in Chrome’s developer tools (Network tab). This essentially represents the time it took for the browser to load the structure of the web page you’re on (DOM), at which point you can interact with it.

The results should always be directly proportional to those from Speedometer 3.1.

 

[rashmi]

I don't know about benchmarks, but I find Vivaldi faster in displaying pages/browsing than Chrome, Edge, Brave, and Firefox. Vivaldi takes a little time to launch/settle on my old HDD laptop.

 

[Marko :)]

IHO I still don't feel that benchmarks such as Speedometer relate at least here relate to every day browsing...

Because they don't. These tests are made by browser developers to test capabilities of their web browsers and to see in which areas they can improve. They aren't meant for users like us nor do they represent reality.

For example: I get low score for Firefox, yet it's faster in rendering websites than Chromium on my Lenovo Legion laptop, despite Chromium having way larger score.

 

[Miravi]

Speedometer doesn't serve as a 1:1 representation of subjective speed, especially overall UX and responsiveness across multiple tabs. However, Apple, Google, and Mozilla were the main engineers behind Speedometer with contributions from Microsoft and Intel, and according to their collective expertise, it's far from meaningless.

Engineers at Apple (who originally created Speedometer) argue that its relevance comes from the TodoMVC and NewsSite components. By using actual code from React, Vue, and Angular, Speedometer demonstrates browsers' abilities to optimize for the exact patterns modern web developers use. If a browser performs poorly here, it will almost certainly suffer on sites like GitHub, Airbnb, or LinkedIn, which rely heavily on these frameworks.

Mozilla themselves have publicly stated that Speedometer 3.1 is the "new standard" for measuring speed, and they specifically praised the inclusion of complex news sites (Next.js/Nuxt) and rich text editing (CodeMirror).

Based on the informed opinions of all the engineers involved, Speedometer obviously has some value to offer.

The fact that Google takes the very same engine and applies objective optimizations just through sophisticated compilation techniques isn't trivial. DNS-level adblocking will do a lot to improve subjective page loading/responsiveness as well.

Chrome's privacy can be tweaked to a fairly satisfactory level. With the right configuration, it won't be Degoogled Chromium, but it will be close. It's true that you'll need an extension (or to use a policy) if you want the browsing history automatically wiped every exit on a normal profile. Chrome does allow you to automatically wipe cookies and site data from the settings UI. The other options are incognito mode or a guest profile.

The original creator of Linux and Benevolent Dictator for Life, Linus Torvalds, uses Google Chrome on his Linux machines precisely because of its reliable engineering.

 

[Sampei.Nihira]

Speedometer doesn't serve as a 1:1 representation of subjective speed, especially overall UX and responsiveness across multiple tabs. However, Apple, Google, and Mozilla were the main engineers behind Speedometer with contributions from Microsoft and Intel, and according to their collective expertise, it's far from meaningless.

Engineers at Apple (who originally created Speedometer) argue that its relevance comes from the TodoMVC and NewsSite components. By using actual code from React, Vue, and Angular, Speedometer demonstrates browsers' abilities to optimize for the exact patterns modern web developers use. If a browser performs poorly here, it will almost certainly suffer on sites like GitHub, Airbnb, or LinkedIn, which rely heavily on these frameworks.

Mozilla themselves have publicly stated that Speedometer 3.1 is the "new standard" for measuring speed, and they specifically praised the inclusion of complex news sites (Next.js/Nuxt) and rich text editing (CodeMirror).

Based on the informed opinions of all the engineers involved, Speedometer obviously has some value to offer.

The fact that Google takes the very same engine and applies objective optimizations just through sophisticated compilation techniques isn't trivial. DNS-level adblocking will do a lot to improve subjective page loading/responsiveness as well.

Chrome's privacy can be tweaked to a fairly satisfactory level. With the right configuration, it won't be Degoogled Chromium, but it will be close. It's true that you'll need an extension if you want the browsing history automatically wiped every exit on a normal profile. Chrome does automatically wipe cookies and site data if you wish. The other options are incognito mode or a guest profile.

The original creator of Linux and Benevolent Dictator for Life, Linus Torvalds, uses Google Chrome on his Linux machines precisely because of its reliable engineering.

The Microsoft team has also made significant improvements to Edge's code using the Speedometer test.
Microsoft contributed to the test just like the others you mentioned.

Contributing to Speedometer 3.0: Capturing real-world challenges on the web



P.S.

If you want to delete data from Chrome, you need to use the corresponding policy.
I don't save my browsing history again, thanks to a policy.

 

[oldschool]

It's true that you'll need an extension if you want the browsing history automatically wiped every exit on a normal profile. Chrome does automatically wipe cookies and site data if you wish. The other options are incognito mode or a guest profile.

Users can also add the "Delete browsing data" trash can icon to the toolbar via Settings > Appearance > Customize your toolbar, or use the tried and true Ctrl + Shft + Delete command.

 

[Miravi]

The Microsoft team has also made significant improvements to Edge's code using the Speedometer test.
Microsoft contributed to the test just like the others you mentioned.

Contributing to Speedometer 3.0: Capturing real-world challenges on the web



P.S.

If you want to delete data from Chrome, you need to use the corresponding policy.
I don't save my browsing history again, thanks to a policy.

Ah, good point. There is the SavingBrowserHistoryDisabled policy to keep away browsing history, although it's not included in the UI for ease of access.

As a general rule, group policies/enterprise policies are the best way to keep a Chromium browser configured to your liking. If you want to tweak Chrome for better privacy, for example, you're better off using these policies to make sure settings don't get reverted.

 

[Parkinsond]

There is the SavingBrowserHistoryDisabled policy to keep away browsing history, although it's not included in the UI for ease of access

But it need to keep cookies for certain websites to avoid repeated log in with each browser launch.

 

[LinuxFan58]

For example: I get low score for Firefox, yet it's faster in rendering websites than Chromium on my Lenovo Legion laptop, despite Chromium having way larger score.

How do you observe or measure that?

 

[Miravi]

But it need to keep cookies for certain websites to avoid repeated log in with each browser launch.

If you want to exclude cookies from being cleared on exit, you can use the "Allowed to save data on your device" list in settings or the CookiesAllowedForUrls policy.

 

[Parkinsond]

If you want to exclude cookies from being cleared on exit, you can use the "Allowed to save data on your device" list in settings or the CookiesAllowedForUrls policy.

I use this option in Chrome settings, inspite being less granular compared to the detailed one of Edge (allowing to select which item to be excluded from clearing on browser exit).

 

[Marko :)]

How do you observe or measure that?

Exactly how @Sampei.Nihira said; open console, load websites without cache and compare DOMContentLoaded value. It's lower for me on Firefox than it is on Brave. I also provided screenshots to @Sampei.Nihira in a topic I cannot remember anymore.

 

[Digmor Crusher]

Unless I can see a speed difference with my eyes then it doesn't exist.

 

[LinuxFan58]

Exactly how @Sampei.Nihira said; open console, load websites without cache and compare DOMContentLoaded value. It's lower for me on Firefox than it is on Brave. I also provided screenshots to @Sampei.Nihira in a topic I cannot remember anymore.

I used PageSpeed Insights on firefox and brave with same websites and brave was faster, funny how experiences can differ

 

[rashmi]

Unless I can see a speed difference with my eyes then it doesn't exist.

I can see a speed difference; Vivaldi is the fastest for me, Chrome, Edge, and Brave give me a similar (middling) experience, and Firefox is the slowest for me. Brave is the fastest to launch, and Vivaldi the slowest on my old HDD system. Simply sharing my personal experience on my system!

 

[Marko :)]

I used PageSpeed Insights on firefox and brave with same websites and brave was faster, funny how experiences can differ

This test indicates how fast website will open from Google's side, not by your side. It's used by developers to see how they can improve load time of their website. It doesn't show you how fast website is in your web browser.

 

[Miravi]

I started this thread intending to examine (in brief) important optimizations that set browsers with the same underlying engine apart. I never intended to declare which browser is best, and I'm not obsessed with chasing milliseconds. You're free to believe that Speedometer 3.1 is irrelevant, but the engineers who made the browser you're currently using disagree. No, Speedometer doesn't tell the whole story as a web app responsiveness benchmark, but it was both convenient and impactful for my purposes here.

The primary reason that Chrome outperforms vanilla Chromium, PGO (Profile-Guided Optimization), is also an important reason that other browsers perform as well as they do. Google's implementation of PGO just happens to be world-class. LTO (Link-Time Optimization) is a simpler, practical step more often applied to browser binaries. Mozilla applies both PGO and LTO to their binaries on all platforms. Microsoft Edge uses PGO heavily as a key performance enhancer. Brave and Vivaldi simply rely on the upstream Chromium build configuration for PGO because of the size of their teams.

Implementing PGO effectively requires large build server farms that can run the browser through thousands of simulated tasks. Google, Mozilla, and Microsoft all have enough resources to do so. Red Hat itself is essentially the engine room for Firefox on Linux, and they employ full-time engineers who work around the clock to make Firefox high-performance and modern on Linux—including fine-tuned PGO.

Chromium on Linux doesn't get almost any PGO love at all, which was reflected in Speedometer results. Google Chrome is currently the only Linux Chromium browser widely shipped with PGO. Meanwhile, Red Hat delivers Firefox builds for Fedora that are extremely optimized and could certainly outperform Chromium browsers in a number of ways.

On Fedora Linux, I wrote a script to measure the cold start times of Chrome, Chromium, and Firefox—the very moment the browser reached the display server. I cleared all relevant kernel caches (for complete cold starts) and loaded temporary browser profiles in a RAM disk to avoid distorted latency. I performed the test three times to calculate an average.

Google Chrome: ~170.33 ms​

Chromium: 582 ms​

Firefox: 227 ms​

Red Hat's special Firefox with PGO + LTO + other Fedora optimizations cold starts 2.56x faster than vanilla Chromium. In fact, Firefox demonstrates incredible performance on Fedora. You're likely to find it to be one of the fastest browsing experiences ever if you don't linger on sites explicitly optimized for Blink/V8 at the expense of other engines.

For Mozilla, Speedometer 3.1 is essentially their North Star for modernizing Gecko and SpiderMonkey because it reflects the most advanced (but also real) web apps today. Speedometer is, in fact, the primary point of reference for Mozilla's PGO, but they also pull in other data on real user behavior. Mozilla's philosophy and techniques are fundamentally different from Google's or Apple's, but the engines they ship are quite robust and insist on adhering to open standards.

Improving Performance in Firefox and Across the Web with Speedometer 3

And we’re seeing results: Firefox got faster for real users in 2023 as a direct result of optimizing for Speedometer 3. This took a coordinated effort from many teams: understanding real-world websites, building new tools to drive optimizations, and making a huge number of improvements inside Gecko to make web pages run more smoothly for Firefox users. In the process, we’ve shipped hundreds of bug fixes across JS, DOM, Layout, CSS, Graphics, frontend, memory allocation, profile-guided optimization, and more.

Is Mozilla just keeping busy by bolting AI and VPN onto aging technology? Not quite. Mozilla and engineers from other companies have been aggressively upgrading every part of Firefox on every platform—Windows, macOS, Linux, Android, and iOS are all being treated as first-class citizens. They progressively rewrite how Firefox works at the lowest levels every year with specialized optimizations for their environments.