Firefox 71 Doesn't Do Much For Performance

[phoronix]

Phoronix: Firefox 71 Doesn't Do Much For Performance

Following last week's release of Firefox 70 and Chrome 78 I posted some fresh Linux web browser benchmarks where the Mozilla browser continued to get beat severely by Google on Linux. But is the situation any better with Firefox 71 in beta? Not really...

http://www.phoronix.com/scan.php?page=news_item&px=Firefox-71-Beta-Benchmarks

[HyperDrive]

Yeah, but… Text subpixel positioning! 🤩

[NiLSPACE]

Any chance you could compare Firefox results on Windows and Linux? I often see people saying that Mozilla isn't focussing on optimization Linux as much as Windows, so it would be interesting to actually see some benchmarks. I'm especially interested in the difference in Webrender performance as WebRender is only enabled on some Windows computers.

[aufkrawall]

It seems the AMD D3D11 driver doesn't like Webrender. I basically have 0 stuttering on Linux when scrolling, while on Windows, it runs much worse than old layered D2D rendering.

[NateHubbard]

If may not do much for performance, but it said that it stopped a social network from tracking me here.

[grigi]

Yeah, but… Text subpixel positioning! 🤩

Woah! That's a thing? Nice!

[Scellow]

The should rewrite their code to C++ instead of rust, it doesn't seems to improve performance like they initially hoped

[arQon]

I'm at a loss for any genuine browser "performance" case that's mattered for, hmm, 5 years now, possibly more. Seriously, how many scenarios are about ANYTHING except giving fanboys something to argue about in forums?

*Atom* CPUs can play fullscreen video - and that's on platforms where WebRender doesn't work. How many people are playing Crysis in a web browser? Even crappy pseudo-app websites spend 99.9% of their time waiting for either human input or network IO.

Browser performance just DOESN'T MATTER once you get above a certain threshold, and we passed that point long enough ago that nobody should even be paying attention to it any more. Privacy matters. Not leaking hundreds of MB a day matters. Fixing basic @#$%ing bugs matters, Firefox developers. (CTRL-Q, anyone?).

But artificial benchmarks for contrived scenarios? It's hard to think of anything LESS important about a browser than that.

[Jabberwocky]

It was my understanding that WebRender uses the GPU to draw parts of the page, do we know if the GPU was used druing the tests?

CPU (% per core) and GPU % would be useful.

[Leprechaunius]

I'm at a loss for any genuine browser "performance" case that's mattered for, hmm, 5 years now, possibly more. Seriously, how many scenarios are about ANYTHING except giving fanboys something to argue about in forums?

*Atom* CPUs can play fullscreen video - and that's on platforms where WebRender doesn't work. How many people are playing Crysis in a web browser? Even crappy pseudo-app websites spend 99.9% of their time waiting for either human input or network IO.

Browser performance just DOESN'T MATTER once you get above a certain threshold, and we passed that point long enough ago that nobody should even be paying attention to it any more. Privacy matters. Not leaking hundreds of MB a day matters. Fixing basic @#$%ing bugs matters, Firefox developers. (CTRL-Q, anyone?).

But artificial benchmarks for contrived scenarios? It's hard to think of anything LESS important about a browser than that.

I'm just gonna quote my answer to this question from a couple of months ago (https://www.phoronix.com/forums/foru...nchmarks/page3) :

I work on an e-government web app that due to some stupid legal issues has to be able to print a pdf from the browser directly, without showing it on the screen.
To do that we set print() as an open action to the PDF document and then send it to the browser, where we use a simple hidden iframe and 2 lines of JS to set the PDF data as a source to the hidden iframe. Browsers then parse the PDF document and open a print dialog to print the document, while users see nothing.
With moderate size PDF documents - e.g. 1MB the speed difference is massive. Chrome opens up the print dialog and prints the pdf almost instantaneously (less than 1s), while Firefox may take anywhere up to 15 seconds to process the PDF and display the print dialog.

So yeah. Firefox is slow. Good in other ways, but slow.