Hacker News Reader

Firefox’s biggest problem is rapacious battery draining. Unplug your charger and Safari feels like a dainty butterfly in a soft summer breeze, while Firefox thrashes around like a drunken Godzilla in mid-90s Tokyo.

The energy usage tab on the activity monitor consistently shows Firefox at a ~40 “score” (whatever that unit is), and Safari more around 4. Anecdotally, the difference in battery time is real.

This is the biggest reason to keep using Safari on Mac in battery mode. (Or Always Be Charging :) )

I hope they can ever fix this :( sounds pretty fundamental, but I know little.

I also observe the same problem with Firefox on Windows and Linux. I have an impression that their developers don't have any priority on the battery drain, instead they just measure speed, and don't want to investigate too much on reducing actual work done, even when, from the user perspective, we're "doing nothing" (e.g. just keeping the browser open on some pages). I know there are many sites today that have tracking codes etc that "always run" but it seems that it's an excuse not to investigate what the browser "always does" in all the scenarios when it should just do as little as possible. Also, different kind of "movement" on the pages (videos, animated gifs, sounds) all take more battery than the competition.

My measurements are that on some specific pages, even when I "do nothing" Firefox uses up to twice as much CPU. Luckily, it's not always twice as much battery drain, because even idle uses power. But it's definitely visible and measurable: from 3 hours of other browsers you'd probably have only 2 of Firefox.

> like a drunken Godzilla in mid-90s Tokyo.

You know you're old when you hear people making references to Godzilla in the 90s :-)

It is pretty fundamental. While Safari lacks features, it feels so much lighter and is noticeably less memory/CPU intensive. As a developer who needs every MB of memory, this makes me resort to Safari for pretty much everything. I'll occasionally use Firefox/Chrome when I really really need their superior webdev tools.

What tools in FF do you think are better than Chrome? Honestly, my computer is beefy (and firmly attached to a power source) enough that performance is rarely an issue with any browser, and Chrome dev tools seem to be superior in most ways, with the slight exception that it's harder to see what events are bound to an element, but the software I develop usually has fairly tight couplings so that feature isn't needed often.

I need someone to make a case that FF dev tools are better than Chrome, simply because without performance being an issue, I need confidence that switching would be a net productivity increase despite having to learn the slightly different tooling of the FF dev tools.

Well don't listen to me, my work is mainly backend. The only frontend work I do is restricted to side projects. But most frontend devs I know prefer Chrome. My personal preference to FireFox is paranoid me trying to avoid Google (but that's another topic).

As someone who has worked a bit on energy usage: it's a really, really hard problem, and the tools used to develop/optimize for energy are really, really bad.

Also, being cross-platform complicates things a lot: I'm certain that Safari has access to a number of libraries that are energy-efficient but exist only on macOS/iOS.

I have a vague hope that maybe the Rust community can find a way to somehow develop a standard benchmarking tool that takes into account energy use. This would help considerably on the front of refactoring Firefox towards better energy efficiency.

I switched over to Firefox for about 3 months. The following were dealbreakers for me and I switched back to Chrome:

- Battery draining...it's seriously really bad. Even if the app is just open on my MacBook, I go from ~8hours of battery to like ~2hrs. I travel a lot...it's untenable

- Performance is generally pretty good, but it never felt better than Chrome.

- At first CPU on streaming video (Twitch, YT, etc) was better on Firefox, but over time I got the infamous "fan spinning" and CPU churning issues.

- Site compatibility: about 10-15% of sites I would visit broke.

- Nitpick, but not deal breaker: icons on my bookmark toolbar didn't work consistently. I would save a bookmark and it would stick to some other site and I could never clear the cache to fix it. It also frequently used non-retina images.

It would be useful to indicate whether higher results are better or not for each of the tests. From what I know, Speedometer: higher is better; ARES-6: lower is better.

It's much faster than it was. I don't notice any difference when I switched. I switched just so chrome wouldn't have such a monopoly

FYI, firefox is working on switching to rendering the entire page in opengl which should bring it's motion mark score up substantially. Details on that (and how to enable it in nightly at the bottom of each post) at https://mozillagfx.wordpress.com

They’d better switch to using metal because Apple deprecated OpenGL for MacOS.

That's Apple's problem, not Mozilla's.

Going to affect the LibreOffice guys also.

Apple caused it, but Mozilla (and, well, everybody else) needs to pay for it.

That's a problem that Apple fabricated themselves. All MacOS games would also have to switch and I don't see that happening other than through the MoltenVK library if they support Vulkan at all.

Since Firefox is a cross-platform project it'd be easier for them to target Vulkan and then use the MoltenVK library to do Vulkan on macOS:

https://moltengl.com/moltenvk/

https://www.khronos.org/vulkan/

Work on WebRender started before Vulkan API was widely available (even now there are lots of users, that can't use it) and right now they focus on integrating it in Firefox. Vulkan support is kept to be implemented "later", as it should enable further opportunities for parallelization in future.

Sadly a substantial number of Mac users are on os versions that don't support Metal. And a substantial number of then wouldn't even get support if they updated their OS.

So Metal-only isn't really an option for many projects, it's OpenGL + Metal or just OpenGL (for now).

Apple deprecated a lot of stuff decades ago that still works fine (talking about macOS).

There will be a free OpenGL-over-Metal implementation sooner or later, so I would still go with OpenGL for a cross-platform application.

Webrender (the new renderer maxyme was talking about) will move to gfx-rs later on [1]. That allows the renderer to run on Vulkan, DirectX 11 and 12, Metal and OpenGL.

[1] https://github.com/servo/webrender/issues/407

FWIW:

                    |     Firefox     |     Edge        |    Chrome 
    ----------------|-----------------|-----------------|-----------------
    Speedometer 2.0 |  65.70 ±01.100  |  59.10 ±01.300  |  86.92 ±00.600
    JetStream       | 191.73 ±14.598  | 226.78 ±12.531  | 162.21 ±02.1999
    Motion Mark     | 173.67 ±04.61%  | 432.36 ±13.07%  | 279.61 ±13.67%
    ARES-6          |  63.72 ±01.26ms |  77.97 ±20.19ms |  27.65 ±00.79ms

To compare: i7 6700k / GTX 1080 / Windows 10

                  |     Edge 42       |    Chrome 67     |     Firefox 61
  ----------------|-------------------|------------------|------------------
  Speedometer 2.0 | [3]  76.8 ±1.0    | [1] 118.0 ±1.3   | [2]  85.7 ±2.1
  JetStream       | [1] 283.0 ±21.4   | [3] 235.1 ±2.4   | [2] 240.8 ±2.7
  Motion Mark     | [3] 273.2 ±12.9%  | [2] 305.4 ±0.9%  | [1] 445.7 ±0.8%
  ARES-6          | [3]  65.4 ±16.5ms | [1]  17.8 ±0.4ms | [2]  51.4 ±1.7ms

Thanks for putting the [1] [2] [3], it becomes easier to understand for someone who doesn't know about these test scores

FWIW the major factor driving user experience is content blocking. It is good to know the subtle diffrences, but any browser that ships with content blocking enabled by defult smashes the browser with the fastest rendering speed when it comes to actual browsing.

Moz://a could easily ship the best browser on the market if they actually cared about the users, instead of taking Google money and pretending to be a competition.

If they stopped taking the Google money it would be much harder to pay developers. I think this would be worse overall for the users, and I bet Mozilla thinks the same.

Disagree, there are many ways to financial stability besides taking money from a direct competitor. Are they that cash strapped?

They would be if the revenue from search engine deals fell away. They've been experimenting with different, user-respecting sources of funding (and still are), but also have to face with enormous amounts of backlash every time they do, so finding alternative funding is not as simple as you make it out to be.

> Are they that cash strapped?

It's more the opposite. They get a huge amount of money from Google. It's pretty hard to replace. E.g. see https://en.wikipedia.org/wiki/Mozilla_Foundation#Financing for a mention + source of 300 MUSD per year.

Assuming these are all in ms, how do you conclude that Safari is faster than Chrome or Firefox? It's slower in almost every test.

I know that at least for the Jetstream benchmark, higher scores are better (ie they're not just time measurements). If you make an assumption that makes the parent comment contradict itself, it seems like it would make sense to investigate your assumption instead of thinking the parent comment is contradictory.

I dunno, this _is_ the internet...

Higher is better in Motion Mark, Speedometer, and Jetstream. ARES-6 is elapsed time.

Safari beat the other two browsers in all four benchmarks.

FWIW, I keep coming back to Opera which feels the snappiest on the latest MBP 13". Any real-world numbers?

Strangely, I've also found that Opera's android browser is by far the fastest, very noticeably.

Also Opera is the only major mobile browser left with a decent text reflow implementation. Shame on Chrome for dropping it.

I've also switched to Opera a couple of months ago. I'm running Opera touch on android too. I've found Opera on desktop (Linux) to be faster than Firefox and Chrome and the only functionality I'm missing is "to move the tabs to left and right". Even the chromecast works on Opera, which is fantastic. I've been using the Flow functionality ever since I've started using Opera, and I think that is the second reason (well, after it being fast) I'm going to stick with Opera.

Yeah I agree with your assessment ancedotally. Firefox a few months ago had horrible performance in google maps and strangely facebook. I checked again today, it's improved a lot, but it is still noticeably slower for those two cornerstone websites. It was enough to make me stop using firefox again

I'm glad that the perf issues have been improved. They just need to get a little bit better and most objections for firefox would go away. Maybe as a workaround someone could make an extension that opens specific urls in chrome/safari when you open them. Only use chrome for maps and facebook for example :p

Too bad there is no way to install Safari on Windows.

Safari is heavily optimized for Mac, so porting it doesn't really make sense.