Firefox 69 Beta On Linux Bringing Better Performance

[phoronix]

Phoronix: Firefox 69 Beta On Linux Bringing Better Performance

With the recent release of Mozilla Firefox 68 there are some nice WebRender performance improvements that Linux users can enjoy. But with Firefox 69 now in beta there is even better performance, including when enabling WebRender on Linux.

http://www.phoronix.com/vr.php?view=28069

[treba]

While these benchmark are somewhat interesting, I think these are really not good benchmarks for browsers. What actually matters for users are things like page loading times, skipped frames and low resource requirements. And that's where Firefox has had big improvements lately.

[schmidtbag]

While these benchmark are somewhat interesting, I think these are really not good benchmarks for browsers. What actually matters for users are things like page loading times, skipped frames and low resource requirements. And that's where Firefox has had big improvements lately.

You're right that skipped frames and resource requirements are important, but the primary bottleneck of page load times is HEAVILY bottlenecked by your network connection and local storage. If you remove that bottleneck, then these benchmarks will gain relevance.

[numacross]

You're right that skipped frames and resource requirements are important, but the primary bottleneck of page load times is HEAVILY bottlenecked by your network connection and local storage. If you remove that bottleneck, then these benchmarks will gain relevance.

Try loading a "modern" page and you'll get bottlenecked by tons of useless JavaScript. Network and storage are kind of irrelevant because most JS is hosted on shared CDNs which get put into the browser cache and in turn into RAM as system file cache.

Also some of us have gigabit fibre and NVMe storage

[perpetually high]

A couple weeks ago I finally made the switch to Firefox from Chrome and couldn't be happier. WebRender was a big part of it.

I ended up enabling WebRender through about:config by toggling gfx.webrender.all to true and verified it with about:support. With WebRender on, you can leave OpenGL hardware-accelerated off (i.e leaving the setting "layers.acceleration.force-enabled" to false). WebRender makes scrolling buttery smooth as well.

Another thing that always turned me off on Firefox was the default setting for fonts. If you have the Windows fonts installed (e.g. Arial for sans-serif and Times New Roman for serif), and then update the preferences in Firefox to use them, you'll get the same look as Chrome.

Before

After

[andrei_me]

To me, the great advantage of Firefox is the memory consumption, all the memshrink effort along the years really pays off now.

If I leave chrome open for about a week and around 12 tabs opened, I'm practically forced to close it and open again because of it's never ending hunger of RAM

[bash2bash]

I made firefox at least 10x faster than before, by installing uBlock Origin... suddenly all those garbage javascript lines don't load and I can block many more by hand. I've broken some websites in the process but then who cares, I just find an alternative that offers similar content.

[treba]

One funny thing here is that FF should be way faster handling unused JS code. While Chrome (V8) compiles all of it into native code, FFs SpiderMonkey/JägerMonkey does some sophisticated stuff to lazily JIT-compile it and optimize hot paths. That is good for the average web page but bad for many synthetic JS benchmarks, making the comparison so hard :/
It short, well optimized JS code (best if statically typed) is usually faster on V8, real world code is often faster on SpiderMonkey (AFAIK).

[andreano]

Tip: I'm writing this in Firefox 70 (nightly)

[F.Ultra]

One funny thing here is that FF should be way faster handling unused JS code. While Chrome (V8) compiles all of it into native code, FFs SpiderMonkey/JägerMonkey does some sophisticated stuff to lazily JIT-compile it and optimize hot paths. That is good for the average web page but bad for many synthetic JS benchmarks, making the comparison so hard :/
It short, well optimized JS code (best if statically typed) is usually faster on V8, real world code is often faster on SpiderMonkey (AFAIK).

Sounds like V8 does JS better then. lazy-JIT with hotpath optimizing is tech that sounds good on paper but then you realize that #1 if the JS code is loaded by the page then it's probably being used somewhere so the lazy JIT will just add unnecessary overhead and hotpath optimization is the pipe dream that one day the cost of profiling will be less than the optimizations gained.