Windows Web Browser Performance

toggle-button
Some interesting conclusions about web browser performance in Windows

I recently took over the category Best Free Web Browser for Windows here at Gizmo's Freeware. We have three categories of Windows web browser, as shown below, and I have been running many tests and benchmarks which can be used to rate each web browser. The purpose of this article is to keep you updated with some of the key results and conclusions that will help you to understand the differences between those web browsers.

 Mega Browsers 

 

 Lightweight Browsers 

 

 Other Browsers 

The purpose of this article is not to recommend particular web browsers or to provide detailed results of benchmarks and other tests. I am instead trying to describe more general principles and trends that can be used to help you to make your own choices.

These tests were run on a low-end system because I was primarily testing for lightweight browsers. To get an idea of how each browser is likely to perform on your system, you can either run benchmarks yourself or you can refer to comparative reviews on systems that have closer specifications to your own system.

The price-performance curve for Windows web browsers

The typical price-performance curve shows how the price or cost of an item increases to achieve higher performance. Here's a link to a real-life example of a classic price-performance chart for high-end audio speakers. As you get better performance (on the vertical y axis) the price increases (on the horizontal x axis). Initially, there is a wide range of performance within a narrow band of prices. Then additional performance becomes more expensive to provide and the price increases at a higher rate. This illustrates a common feature of such curves. The cost of an incremental improvement in performance keeps increasing to the point where it does not make any economic sense to continue.

You might be saying "How does this relate to free web browsers? They are free so they don't have a price!" Actually there is a price. But is not the purchase price of the software. Instead the price that I am considering is the cost of the computer resources (disk, memory, CPU) needed to run the web browser.

I have found that there is a clear trade-off between the browser performance, as measured by various benchmarks, and the resources required to achieve it. This relationship is illustrated in Diagram 1. There are many benchmarks that can be used so I also tried this with different groups of benchmarks and the underlying relationship did not change.

What this means is that the worst performing browsers are usually the least resource-hungry and the best performing browsers are usually the most resource-hungry. The browsers also fall into bands of performance consistent with the underlying web (layout and rendering) engine that they are built on.

Diagram 1 Web browser performance versus resources used to achieve that performance

 

How this curve was constructed

The performance that I was measuring consists of an average of the speed of rendering and the browser's conformance to web standards such as HTML5 and CSS3. The resources that I measured are primarily CPU time and memory used (working set private memory). The trend line is drawn in to make it easier to observe the relationship and compare the three charts.

For each benchmark score that I recorded, I normalized all data in the range 0% to 100% where the best score represented 100% and all other values were calculated in relation to that. The normalized scores went down as low as zero if the benchmark failed to run to the end.

  • Performance is the average of the normalised benchmark results. None of the scores reach 100% because they are the averages of the HTML5/CSS3 scores and the throughput benchmarks which include JavaScript rendering speed.
  • Resource use is the average of the normalised CPU and memory usage. For example, resource usage of 4 means four times the lowest usage. This is shown on the x-axis at the bottom of the chart, where "1.0" is the least amount of CPU and memory used by any of the web browsers and "10.0" is ten times that minimum. No web browsers have a resource usage of one because two different browsers had the lowest memory usage and the lowest CPU time.

These price-performance curves only cover 32-bit web browsers running in Windows. Similar curves could also be produced for 64-bit web browsers but there are far fewer so the trends are not as clear.

There are also a couple of unusual results. I removed one outlier for Maxthon which normally uses the Blink engine but can use the Trident engine for compatibility. I was unable to run the Trident engine by itself so I removed that result. The Sleipnir result is also unusual but I have not yet identified exactly why that is.
 

Note 1: These are examples not general recommendations

You need to be aware that my results are specific to the low-end computer they were tested on. Different combinations of CPU and graphics processor can produce much different results and the relative performance of browsers can change. While the engine averages and the overall trends should be consistent, the results for each specific browser will change.

 

Note 2: There are other factors to consider when choosing a web browser

The results that I am illustrating here do not say anything about many browser features such as security, add-ons and extensions, bookmarking, menus, and settings. As an example of this you can see In Diagram 2 that Opera uses less resources than Chrome to provide similar performance. One reason for this is that Opera lacks many important features. So my preference would be to use Chrome because some of those features are really important to me.

 

Comparing the major web engines and web browsers

If you are considering a major web browser then other browsers using the same engine are more likely to be perform better by running faster and being compatible with more up-to-date website features. This is the case for Firefox and Chrome, at least, as Diagram 2 shows. I have indicated the more popular browsers using each engine and the average for each engine. Much of the reason for the difference appears to be that the other browsers do not implement the same features.

Diagram 2 Web engine averages and the major web browsers

 

Some other browsers stand out

Outside of the major web browsers, four other web browsers stood out because they provided greater performance using fewer resources than their companions. While this result is likely to apply more generally, you would have to run benchmarks on your own system to confirm that they would be faster for you too. You should also take note that a performance improvement can be achieved by removing features that are not measured by the benchmarks which is another reason to evaluate browsers yourself to ensure that they have the features you rely upon.

  • Maxthon and 360 Browser were the fastest.
  • Lunascape with the Gecko engine uses the least resources.
  • For the Trident engines, SlimBrowser is almost competitive with Internet Explorer.

Diagram 3 Some stand-out web browsers

 

Should I use a 64-bit web browser?

Do you have Windows 64-bit?

If you don't have a 64-bit version of Windows then here are some articles to help you understand the issues: 32-bit and 64-bit explained and How Windows 64-bit supports 32-bit applications.

"Should I use a 64-bit web browser?" depends upon what is most important to you.

What is your priority?

  • Maximize security? 64-bit web browsers can use enhanced security features available on your 64-bit CPU that are not available to 32-bit processes.
  • Minimize CPU processing? 64-bit web browsers can perform better on your 64-bit CPU for two reasons. First, some native 64-bit features are faster. The more important reason is that your programs don't have to run under emulation which is slower. But of course, 64-bit values take up more space on disk and in memory than 32-bit values so 64-bit programs start at a significant disadvantage.
  • Minimize memory used? 64-bit web browsers have larger programs and data simply because 64-bit values are twice the size of 32-bit values. However, Windows can more easily manage memory because 64-bit memory pointers are not limited to 32-bit addresses. Plus Windows allows larger 64-bit processes, 8GB instead of 2GB. This is particularly important for Firefox which does not run separate processes for each tab or group of tabs.

 

Comparing the major web browsers

The 64-bit version of Chrome was released in August 2014. One month later Guy McDowell at makeuseof.com has published some useful comparisons between the 32-bit and 64-bit versions of Chrome. Although I have run the same benchmarks and others, the startup time is meaning less to me because my web browser runs all the time. What is more useful to me is how much memory and CPU time is used to load web pages. So I compared the 32-bit and 64-bit versions of Chrome, Internet Explorer, and Firefox (Waterfox is the 64-bit version).

32-bit web browsers use less memory

The following graphs show memory use for 1, 20 and 50 tabs open to the top website list at Alexa - I did exclude sites that were red in WOT, flagged by my antivirus, or not good for kids. The lower numbers are after all but one tab are closed and the web browser has had time to release the memory that no longer needs to be used.

Diagram 4 32-bit web browser memory use

 

If you compare Diagrams 4 and 5, you will notice that the 32-bit versions use less memory both when the tabs are all open and after they have closed. If I had continued testing to a 100 tabs then Chrome and Firefox might have their 64-bit versions using less memory.

Diagram 5 64-bit web browser memory use

 

64-bit web browsers use less CPU time

64-bit web browsers use less CPU time to load the web pages. The improvement is greater for Chrome and least for Internet Explorer which appears to be more efficient with fewer tabs open.

Diagram 6 64-bit web browser memory use

What you use depends upon what is more important to you

Clearly, my results are similar to Guy McDowell's: there are trade-offs in swapping out your 32-bit web browser for a 64-bit version. In my tests, CPU times are less but more memory is used.

Personally, I like 64-bit browsers because they have better security. They also seem more responsive when I use them which makes sense if they require less CPU power.

 

Editor

This article is maintained by volunteer editor Remah. Registered members can contact the editor with any comments or questions they might have by clicking here.

You are welcome to join the discussion in our web and networking forum

Back to the top of the article


 

 Mega Browsers 

 

 Lightweight Browsers 

 

 Other Browsers 

 

Please rate this article: 

Your rating: None
4.615385
Average: 4.6 (13 votes)
toggle-button

Comments

I had a problem with Maxthon 4 a while back but they seem to have fixed it,,
The tabs would not open in the same order that you saved them in.. I like my tabs in a certain order..:)
But it seems to be fixed now.. Great!

I have updated the price-performance section and added a section comparing the three main 32-bit web browsers (Chrome, IE, Firefox) with their 64-bit counterparts (Chrome, IE, Waterfox). Remah - Editor
Thank you Remah for your article I have been using Maxthon for quite a few years now and have always found it easy to use, your comparisons were very interesting thank you for your effort.
I simply find Maxthon fast and very easy to use for my want in and out of the cloud. The paste and go feature is also very helpful.

What about extensions? Is there now a way to use Chrome's with it? Otherwise, living in the minuscule Maxthon extension ghetto won't be pretty.

Thanks AWS. There will be more info to come. I'm interested to hear what you like the most about Maxthon as I'm still trying it out.

Many thanks for advice - I did download it from a third party vendors site so they could indeed have wrapped up malware with it.
I will have another go downloading the file from the Maxathon site directly.

Have just downloaded Maxathon installation file.
AVG unable to accept saying file contained MalSign InstallCore.B06 Adware.
Is this an acceptable install risk please?

You have taken the first and most important step. You are using a powerful and well tested security tool. And, although no defense package is 100% accurate, this time it did its job.

Adware, whether flagged by an AV or not, is never an acceptable risk.

At best it clogs and impedes your system and forfeits your privacy. At worst it provides a major surface of vulnerability, opening your system and your network to significant infestation. When that happens, not only have you been compromised, but you have become a potential vector of compromise for every system your system touches or has the ability to touch, whether through e-mail or directly.

With the recent permitting and/or use of installation wrappers as common practice among sites that have been long trusted as viable sources of freeware downloads, it has become absolutely essential never to opt for Express or Standard installs. Always choose Custom installation. Always watch your installations carefully. Always opt out of any toolbars and/or additional programs. Consider the viability of using the portable (non-installer) version of a program instead. Among the legitimate programs, these are usually free of bundle-ware.

Where possible, try to access your freeware directly from the author/publisher. Be aware that is not a guarantee that the installer will be free of ride-alongs. Learn to consult VirusTotal. There are even extensions, such as VTzilla for Firefox, that can integrate VT into your download process and also give you context menu items and a toggled toolbar with the ability to scan site and file URLs. VT is helpful, but not infallible. Many of the scanners polled by VT detect viruses, trojans and worms within an executable program but not "legitimate" opt-in bundle-ware within the wrapper.

Another useful tool which has become commonplace is the reputational site scanner/immunizer. It can be found in most solid malware defense packages such as AVG, avast!, SuperAntiSpyWare, Malwarebytes, SpyBot Search & Destroy (the Immunizer function of this old standby) and many others. It is also built into many contemporary browsers. There are even free-standing site scanners such as SpyWare Blaster . Of course, in terms of program installation, these are usually going to be detecting the hinky sites you may have been pointed towards by those questionable installs. But you will also be kept away from the worst sites where you might acquire questionable downloads or drive-by infestations in the first place. Another way to help avoid these sites is an add-on such as AdBlock Plus or, my favorite, AdBlock Edge, with a good and frequently updated set of block lists.

So, in answer to your question: What do you consider acceptable? Is the promise of FREE worth the reality of a system compromise? There is no free lunch. I believe in freeware. But I like to know what lunch is going to cost.

There endeth the sermon.

dark matter, For the record, I moderate the comments on my articles. Normally I would delete your comment because it is off the topic of this article, which is the performance of web browsers, and it is quite long. I only left it here because internet safety is such an important issue.

> I only left it here because internet safety is such an important issue.

...which is why I invested the time and effort in answering kthtl's inquiry.

My screed may seem oblique. It is actually on topic. The security of all of us depends upon the security of each of us. And without secure practices, the most efficient and effective browsers are worth nothing.

Speaking of which, thank you for this interesting investigation. The follow-on promises to be fascinating. I have been using Firefox since version 0.8. For the last 10+ years it has been a central piece of kit. Along with many others, the search is now on for a shapable, reliable and lovable replacement for the ruins that the mozilla devs have left us. Your methodical and organized explorations are eagerly anticipated.

LOL, I hope I don't disappoint you as I'm not one of the many who use the "shapable" features of web browsers. I've been a longtime user of Firefox but I use it straight out of the box and my only preference is a WOT add-on. That means that very few developer shennanigans impact me and I regularly swap to use other browsers without any issues. In my plain vanilla pattern of use I find that the current browsers are lot more interchangeable than they were five years ago. It's a very different story if you heavily customize your browser.
Virustotal.com didn't find any problems when I scanned the install files recently: mx4.4.2.2000.exe https://www.virustotal.com/en/file/6eb6531760ac8d9b38e24ccfb5f7fc5544906... mxsetup.exe https://www.virustotal.com/en/file/31c1dc3eabc5b5f6ebf19afe13ed0ca11702e...

AVG is a little over-cautious, it even flags programs that have been in my filesystem for years without any problems as threats to be eliminated. As long as your dealing with a reputable site you should be fine. Yiu can temporarily turn off AVG to allow you to install the program. Just remember to turn it back on!

The download source can indeed be significant. This is often triggered by a wrapped installer. MC - Site Manager. http://superuser.com/questions/722205/does-filezilla-contain-a-virus https://forum.filezilla-project.org/viewtopic.php?t=32672
MC, that's a very important point. kthtl, we check the download links that we provide in the Quick Selection Guide (QSG) at the end of the article. If you are downloading from anywhere else then you need to perform your own checks with your antivirus.
Thanks for the comparison Remah. Could you please tell the system configuration that you are testing these browsers with. Also I believe the results may vary widely when you're having a low-end system.
I really wanted the main focus of the article to be on the relationship between performance and resource usage so I didn't want to publish details about benchmarks and specs. If I do that will go into a more detailed article later on. To make that clearer I've removed the tables. I've also emphasized the note "You need to be aware that my results are specific to the computer they were tested on. ..." and repeated it in the introduction.
Thanks for making it clear, although it wasn't necessary to remove the tables.