Announcement

Collapse
No announcement yet.

Firefox 105 Now Available - Better Linux Performance Under Memory Pressure

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • arQon
    replied
    Originally posted by bug77 View Post
    It's not subjective, it's very, very objective. Somebody actually measured (a small study or something like that a few years back) and found that we routinely transfer a few MB of data just to convey 2-4kB of useful information (i.e. text). The rest is images, styles and whatnot. That alone puts a hard lower limit on RAM usage.
    IIRC it's actually even worse than that. The "average" web page equivalent of "Hello world" would be something like 10K of CSS, 50K of spyware, and 200K of ads now; or some equally ridiculous set of multipliers.

    You'd think there'd be a sort of natural upper bound to those aspects, but it turns out there isn't really: I keep seeing insane numbers like "37 different trackers per page" etc in studies (and that's mainstream stuff, not pr0n/warez sites!), and given that most of the emails I get from Outlook clients already burn up to 150 chars worth of CSS per char of content there are obviously even worse publishing systems waiting to happen.
    The only thing keeping ads from wasting even more bandwidth is there not being any screen real estate left for them, but they're already getting around that by using animated multi-frame images instead, and once 4K displays become mainstream they'll quadruple the size of the textures "because otherwise they look blocky".

    So yeah, give it a couple more years - especially with the desperation of a recession going on - and that 1000-to-1 noise ratio of yours will need at least one more zero, with a corresponding further increase in RAM usage. The only piece that definitely *won't* expand is the actual signal. (And in fact, most articles on news sites etc are getting shorter, not longer, because there's much more profit to be made by serving up multiple short pieces of clickbait than from a single story that had actual effort put into it).

    It's remarkably informative to install NoScript and uBlock and compare the memory usage of the same page in two different instances of FF with and without them. Don't worry though, because once Google demands Mozilla drop support for Mv2 none of those tools will work any more, so that we get to "experience" the modern web in all its spyware-laden, auto-playing video ads glory. I can't wait...

    Leave a comment:


  • arQon
    replied
    Originally posted by bug77 View Post
    Firefox used to be pretty good wrt memory management. Back then, people were bashing it for not following Chrome and run a JS interpreter per tab. Well, today it does. And people bash it for running a JS interpreter per tab and chewing through RAM, just as fast as Chrome does.
    Yeah, you can't really win (though the complaints were nothing like that advanced, they were just "Why is FF so much slower on this subset of garbage-tier sites?", and "Why does FF crap itself and lose my other 30 tabs when it falls over on this other unrelated tab?", both of which are certainly reasonable questions for users to ask).

    The biggest problem I have with FF these days is with Mozilla's dishonesty, not any technical concerns. BS like the Compact Mode removal only drives home just how much they're willing to twist the facts to get to the end result they want, and that and the Chrome-chasing shows that the UI team not only have absolutely no idea what they're doing but can't even be trusted to act in good faith. FF is still better than the alternatives, but it's dead set on a suicide run down towards that level rather than away from it.

    Leave a comment:


  • yump
    replied
    Originally posted by stormcrow View Post
    A very vocal few still stuck in the early 10s with only 4GB of system RAM?
    Poor people are buying computers with 4GB system RAM in 2022. A whole lot of Chromebooks have that much. Also phones, because Firefox Android exists (and is notorious for losing tabs under memory pressure).

    Leave a comment:


  • darkonix
    replied
    I'm not noticing any difference yet with my 8GB circa 2010 build. Firefox works just fine anyway for me. Having said that any improvements in memory usage are appreciated.

    Leave a comment:


  • carewolf
    replied
    Originally posted by oleid View Post
    Uh, that is what buggs me about Chromium most (I use Chromium only for pages that don't work with Firefox on my work laptop with Debian 11). Very nice. Do we know if these changes are available for Android as well?
    Chrome has also has a memory pressure system. It took a while to implement for Linux too, but they did that two years ago. If I start very aggressive compile jobs and take all the memory in the system, the Chromium based applications all go blank when they clear their caches and kill helper processes to free up memory. Of course under less pressure they just clear caches, which is much harder to spot as it only means slightly slower reloads.

    Leave a comment:


  • Michael_S
    replied
    I hate the amount of memory Firefox uses but I'm not convinced Chrome or other browsers with the same add-ons are any better. Chrome on my work laptop with ~50 tabs open currently has almost 4GB of RAM, Firefox on my home computer with ~50 tabs has something similar. (Measured in `top` on both machines.)

    I blame the websites, not the browsers. If I monitor memory and navigate Sourcehut (no React, no Vue, no Angular, no Bootstrap, no jQuery, no trackers, no ads) my memory usage goes up by a few MB per tab, and most of that is from the images. If I pop open Youtube in any browser my memory usage goes up many hundreds of MB.

    Shared memory between sites and processes has been shown to be a security risk, both in terms of buffer overruns and also information leaks. I think the problem is the web, not the browsers.

    Leave a comment:


  • finalzone
    replied
    Originally posted by Vistaus View Post

    Not sure which one, but I have an AMD Ryzen 9 5900HX and Inxi reports:
    Code:
    [FONT=monospace][COLOR=#5454ff][B]Graphics:[/B][/COLOR]
    [COLOR=#5454ff][B]Device-1:[/B][/COLOR][COLOR=#000000] AMD Cezanne [/COLOR][COLOR=#5454ff][B]driver:[/B][/COLOR][COLOR=#000000] amdgpu [/COLOR][COLOR=#5454ff][B]v:[/B][/COLOR][COLOR=#000000] kernel[/COLOR][/FONT]
    You can also view about:support which may show WebRender (software). I encounter similar issue with AMD Ryzen 7 5825u equipped with updated Vega8 (Barcelo).

    Leave a comment:


  • Vistaus
    replied
    Originally posted by Vistaus View Post

    Yes, I don't have enough RAM. Guess I need 64 GB instead of the 32 GB I have now...
    And for the record: I don't even have that many open tabs. I have about 25-40 tabs open on a daily basis, but only about 10 of them are actually loaded simultaneously. Yet FF manages to use a ridiculous amount of RAM per tab. And I do look at active utilization.

    (And 10 loaded and 25-40 total tabs isn't all that much. Maybe a bit more than the average user, but a lot of people, esp. in tech communities, have over 100 total tabs, so compared to them, my usage is still pretty light.)
    I do have to admit that the situation has gotten better with today's 105 update. I'll continue to monitor it...

    Leave a comment:


  • Vistaus
    replied
    Originally posted by finalzone View Post

    Which AMD graphic card of APU? Hardware acceleration is enabled on Raven Ridge series.
    Not sure which one, but I have an AMD Ryzen 9 5900HX and Inxi reports:
    Code:
    [FONT=monospace][COLOR=#5454ff][B]Graphics:[/B][/COLOR]
    [COLOR=#5454ff][B]Device-1:[/B][/COLOR][COLOR=#000000] AMD Cezanne [/COLOR][COLOR=#5454ff][B]driver:[/B][/COLOR][COLOR=#000000] amdgpu [/COLOR][COLOR=#5454ff][B]v:[/B][/COLOR][COLOR=#000000] kernel[/COLOR][/FONT]

    Leave a comment:


  • finalzone
    replied
    Originally posted by Vistaus View Post

    Same on AMD, at least for me.
    Which AMD graphic card of APU? Hardware acceleration is enabled on Raven Ridge series.

    Leave a comment:

Working...
X