Chrome 89 Increases Desktop Memory Efficiency With PartitionAlloc

Google Chrome version 89 began rolling out to users in the stable channel on March 2 and should be on most people's machines by now. From a report: The new build offers significant memory savings on 64-bit Windows platforms thanks to increased use of Google's PartitionAlloc memory allocator. On macOS, Chrome 89 plays catch-up and gets closer to the performance of the flagship Windows builds. Google says use of RAM in 64-bit Windows is down up to 22 percent in the browser process, 8 percent in the renderer, and 3 percent in the GPU. The company also claims a 9 percent decrease in latency, meaning a more responsive browser. The improvements are largely due to intercepting malloc() calls with PartitionAlloc. Chrome 89 has also gotten significantly more aggressive about discarding unused RAM. When you scroll resources such as large images off-screen in the foreground tab, Chrome discards the memory those resources used. The change impacts background tabs as well, resulting in a savings of as much as 100MiB per tab.

Re: Islam

[pele]

"Civilisation", you say?

Chrome v Chrome

[shanen]

Please don't perpetrate troll Subjects. Or did you have some reason to engage?

Then again, I have to admit I was confused by the story and now I see it as rather a trivial topic unlikely to produce much discussion. At first I thought it was about the Chrome OS rather than the browser. Yesterday was my first attempt to run it from a virtual machine and I already encountered a number of questions... It would have been one of those interesting coincidences if Slashdot had been closer to the topic. As regards th

Re: Still over 1G here

[BanHammer]

Sadly there is no real lightweight browser on PC

Re:

[MightyMartian]

Considering the Javascript, WebAssembly, HTML 5 engine and all that goes along with it, I don't really see how a lightweight browser that actually renders would even be possible. There's always Lynx if you don't give a shit about images and the like, but basically web content these days requires quite a bit of horsepower.

Re:

[Mr. Barky]

I really doubt it is the code itself that is the problem (i.e. the number of features). Yes, it adds something, but not 1GB - probably in the 10s of MB at most. It is really the size of pages (and the amount of local data they want to store) and things like caches (including caches of compiled javascript or whatever) that eat up the memory.

Re:

[MightyMartian]

It's also very likely a lot of preallocation when a new tab or page opens. If you've got a few gig of free RAM, browsers may just go "Well, let's malloc() lots of that, in case we need it." Chrome may lean more heavily towards grabbing the memory it can at initialization, whereas, perhaps Chromium Edge or Firefox might decide to more conservative. As much as people criticize software for these sorts of presumptuous use of resources, there's a logic to it. It usually only becomes a problem for those that lik

Re:

[Bengie]

Firefox is conservative? It consumes nearly 2x the memory as Chrome on my home machine, and at work with very specific cloud services, more like 10x the memory. Literally got FF holding down 50-70% of my physical memory and about 80%(20GiB) of my virtual memory. That with tabs slept. A single tab of Jira eats up 500-1GiB in FF, but only 200--300MiB on Chrome. Been like this for 2 years since we went Jira.

I've tried disabling all extensions, just in case. No change, other than using more memory because of

Noscript

[DrYak]

I really doubt it is the code itself that is the problem (i.e. the number of features). Yes, it adds something, but not 1GB - probably in the 10s of MB at most.

Sweet summer child...

Open the debug tab of your browser.
Reload any random internet page.
Look at the zillion of various random javascript frameworks that get downloaded.

Then install NoScript(*) in Firefox. Go to the same page, and slowly start whitelisting the bare minimum that is needed to get the page functionnal.
- Notice: how much less memory you're consuming.
- The gazillion of frameworks (not even their dependencies yet at this stage) that the page is still begging NoScript to further allow.

Modern intern

Re:

[xack]

I recently tried Firefox on a PC with 256mb ram (yes mega, not giga). It worked with most modern sites apart from Google’s Bloattube.

Re:

[ArchieBunker]

My wristwatch has 1Gb of ram. Get with the times grandpa.

Re: Still over 1G here

[Viol8]

And probably needs charging once a day LOL :) Meanwhile the battery in my cheap casio has been going for 2 years already.
I'm sure PT Barnum would have lived you as one if his visitors.

Re:

[williamyf]

My wristwatch has 1Gb of ram. Get with the times grandpa.

Sometimes, I need a bunch of VMs to simulate certain topologies, or usage patterns. It helps a lot to reduce the footprint of said machines as much as posisble.

To that effect I use AntiX Linux (if I want the minimum footprint without much fuss) or CrunchbangPP (if I want the most modern debian available) with 256MB RAM, so I can create more machines on my 16GB desktop, and/or my 8GB NAS.

Probably, the "grandpa" you are talking about is doing something simlarly advanced, beyond your comprehension.

So, educate

Re:

[Arthur, KBE]

I was banned from Ars Technica after being a subscriber for 14 years. Somewhere along the line it shifted from being a tech site, to a political site (and you can guess *exactly* where on the political spectrum their editorializing lands). They're now essentially the Daily Kos of tech "journalism", and if you call them out for it, they totally don't understand WTF you're talking about. I think Jonathan Gitlan, who does the car reviews, is probably the only writer on that site that seems even remotely non

Re:

[Arthur, KBE]

And BTW, as you might have guessed, I was banned by Aurich Lawson.

Re:

[Arthur, KBE]

Try it and report back.

Re:

[Arthur, KBE]

I was on Ars in the late 90's when Ken Fisher was running the show. He's clearly a classic New England liberal, but he managed to be objective and kept the staff on a tight leash. Obviously, I don't align with him politically, but he was a reasonable person, I think, and the site reflected that. Now, every article on Ars that doesn't come from Gitlin or Eric Berger, has a "tone" to it, that is going to annoy the living fuck out of 50% of the US population.

Re:

[williamyf]

But, I am curious, in the email where they informed you you were banned:

what was the bezel situation?

were very big bezels? What was the bezel to content ratio? Did it have an ugly chin bezel?

Just asking.

Re:

[Arthur, KBE]

But, I am curious, in the email where they informed you you were banned:

It's hilarious to me, that you think they possess that level of courtesy.

And I have no idea what you mean by a "bezel situation".

Re:

[williamyf]

And I have no idea what you mean by a "bezel situation".

If after 14 years on ARS you do not know abut the Bezel Situation running gag, you did not pay enough attention during your 14 years on ARS. ;-) :-D

Re:

[Arthur, KBE]

Well then explain it to me, enlightened one.

Re:

[williamyf]

enter to Ars as anonymous, and read the comments of any Ron Amadeo Article.

Don't see the point

[willoughby]

As long as the "Software Reporter Tool" is installed along with Chrome any browser improvements are pretty much a waste of time. The Chrome browser has never brought my computer to its knees but Software Reporter Tool sure has.

Benchmarking

[rsilvergun]

And I suppose if you're running a cheap Chromebook with only 2 or 4 gigs of ram.

Re:

[thegarbz]

Considering Software Reporter Tool runs once per week, uses less than 20MB of RAM, audits the plugins and then closes itself if it's bringing your computer to its knees you're doing something very VERY wrong. Even if your computer is a potato.

Re:

[random_nb]

Software Reporter Tool and Chrome Cleanup integrate a version of ESET's general Windows antimalware engine and check far more than just the plugins list. It's efficient for what it does, but there's no mistaking when it has decided to run a more complete checkup of things on a less-than-potent computer. It'll chew on the CPU and disk for multiple minutes.

Re:

[thegarbz]

No it doesn't. Not even on a potato. There are the occasional people who have reported bugs, but even on shithouse PCs Software Reporter Tool uses little to no resources and is rarely running.

If you have a problem with it, file a bug report and help the developers help you.

Finally...

[dutt]

Finally a patch to chrome that decreases memory usage, in contrast to all other ones that have been increasing usage since launch.

Re:

[93 Escort Wagon]

Well, as Sundar Pichai famously said - 640T ought to be enough for anybody!

I went to look at 89's release notes

[93 Escort Wagon]

But by the time the web page loaded, they were on release 94.

Now they've moved on to 102.

100MB????

[pele]

I remember being forced by my boss to stick to 128k per page limit - html, images and all. So what's this 100MB malarky?

Re:

[thegarbz]

I remember being forced by my boss to stick to 128k per page limit - html, images and all. So what's this 100MB malarky?

We introduced browsers that could display pictures.

Re: 100MB????

[pele]

What part of "html, images and all" got you confused?

Re:

[thegarbz]

The part where an image is considered more than a 256color GIF that is 128x128 pixels in size.

Re:

[omnichad]

You can do a lot with 128KB. Even a large "hero image" on a page doesn't have to be over 70-90KB. The problem is when you've got video clips being used as background images, loading jQuery and a dozen plugins, maybe 2-3 more JS frameworks, a CSS framework too, scripted banner ads (even MSN.com can sometimes redirect people to a Microsoft Support scam through a malicious ad), web pages just get stupid big for no good reason.

Re:

[xwin]

What a bull crap. Look at this page, it hardly has any pictures. I am viewing it in Firefox, I don't use Chrome on my personal machine. The Firefox shows size as 149,028 bytes. The Firefox task manager shows this page being 11MB which is 100 times more than the actual HTML size. However the smallest Firefox process that size that I can see is 31MB which is 300 times more than the actual HTML size. When I open the same page in Chrome it is 33MB.
I actually run ad blocker and do not have any banner ads and o

Re:

[antdude]

That still can be done today. ;)

Re:

[Mr. Barky]

"Progress"

Right from the manual

[gillbates]

Good to see that Google has read How to Write Unmaintainable Code" [stateslab.org], and has reimplemented the memory allocator, as recommended in the "Roll Your Own" section.

Good for them!

Re:

[bill_mcgonigle]

> reimplemented the memory allocator

Yeah, why would they roll their own when there's a good, secure, efficient memory allocation library [googlesource.com] that they could have used instead?

Can't they see that the community should build one good allocator to help people write more secure maintainable code?

Re:

[EMN13]

To be fair, a modern browser is considerably more complex than a classic OS, and for good reason. Seems pretty reasonably to have all kinds of fundamental stuff re-implemented specifically for it.

Heck, while Firefox may have lost the browser wars, just consider the fact they invented and maintained a revolutionary (depending on your perspective) programming language just for the browser, and at least technically it seems to have even kind of payed off, weirdly enough. And just how many JS compilers/JITs/ba

Re:

[toddestan]

I actually worked on a code base where someone who had written significant portions of it a long time ago had done exactly that. It never actually caused any problems for me, but I guess something like that in commonly used code paths that had been around for a long time must have had the bugs worked out of it. Of course, it was utterly pointless and unnecessary, but we were all afraid to mess with it so it lived on.

Chrome Disk Thrashing vs. Edge

[northerner]

I find I am using Edge more often because Chrome seems to cause a lot of disk activity when I fire it up. What in the world is it doing? Edge doesn't cause the same disk thrashing. It isn't really a big deal other than I find it annoying on PCs with louder hard drives. It isn't this memory efficiency stuff because it happens even before I open a page.

Re:

[SuricouRaven]

I wonder if that just means Windows has sneakily pre-loaded some of Edge's components already.

Re:

[ArchieBunker]

You're still using platter drives to run an OS in 2021? Amazon has 120Gb SSDs for like $20. I put them in low end i3 laptops and they fly.

Now cut down on Youtube’s memory usage.

[xack]

It got so bloated the space bar function broke on it this week. On many of my computers I get “a script is slowing down your browser” error on Youtube.

I down up all the things

[bustinbrains]

"Google says use of RAM in 64-bit Windows is down up to 22 percent in the browser process"

Punctuation. It's important.

thanks, but no thanks

[Pasajero]

I stoped using Chrome a while ago. Firefox keeps getting better and new Edge is actually pretty neat.

Fix the urgent problem

[robi5]

- Open two dozen Google slides, docs etc
- keep the tabs but don't use them
- still, eventually, 100% CPU use

Re:

[ledow]

Stop opening dozens of tabs and then just leaving them open forever?

Though I can confidently say, in an organisation with literally millions of Google docs, sheets and slides, and a Chrome-only browser policy... I have never noticed such an effect.

Re: Fix the urgent problem

[TuballoyThunder]

Probably because Google has harvested all of your personal information and they don't need to update their profile on you. (jj but not really)

Welp, that oughta help with Tumblr

[Wolfrider]

...if anyone is still using it after the censorship started back a couple of years ago :B