Chrome 91 Benchmarks On Linux Showing Off Even Better Performance

[schmidtbag]

Chrome don't works on Wayland natively, but Firefox works. Chrome Wayland support currently work-in-progress, but only with Ozone abstraction layer, that by the way have dozens of issues, during Firefox works on Wayland full natively and I don't see any issues. On benchmarks Chrome works about 2x faster, but on practice I don't see it during every day web browsing. I used Chrome during many years, but switched to Firefox for native Wayland support.

My experience with Chrome+Ozone is fine. Not great, but fine. I don't like that it uses the system borders and changes the mouse pointer on me but that's about the worst of it.

You are right that in real-world usage, the performance difference hardly ever matters.

[CochainComplex]

I don't know why you'd think my cat Spark Plug would help in Chrome benchmarks either. He's much better with compression.

Is this a deb or rpm package?

[Brane215]

WHo cares about CPU cycles allegedly spared.
It leaks user data like a sieve.
How many human years of life were lost because of that ?
And even those gains are done just to coax users into running on-line applications to further surveillance and control.

[ferry]

My experience with Chrome+Ozone is fine. Not great, but fine. I don't like that it uses the system borders and changes the mouse pointer on me but that's about the worst of it.

You are right that in real-world usage, the performance difference hardly ever matters.

Nope, that's wrong. Other performance matters.

On my laptop I have 4GB ram (and no swap), so I care about the number of tabs I can open. And with Chromium that's a hell of a lot less then with Firefox. With Chrome my system will lockup in no time (not even mouse movement) forcing me to reboot.

Also, it's on battery. So I care who will allow longest browsing without recharge.

Unfortunately, none of this is tested.

[ZFKerr]

My experience with Chrome+Ozone is fine. Not great, but fine. I don't like that it uses the system borders and changes the mouse pointer on me but that's about the worst of it.

You are right that in real-world usage, the performance difference hardly ever matters.

The system borders usage can be disabled in the Chrome settings.

[schmidtbag]

Nope, that's wrong. Other performance matters.

I said "hardly", not "never". I'm aware there are circumstances where it matters. At my job, the performance matters a whole lot. For 99% of the pages you browse, your network is most likely going to be a bottleneck before the CPU. This doesn't mean that CPU performance shouldn't be optimized (because it definitely should), but rather, it's not a big deal that Firefox is slower, because in real-world examples, you're not likely to notice.

On my laptop I have 4GB ram (and no swap), so I care about the number of tabs I can open. And with Chromium that's a hell of a lot less then with Firefox. With Chrome my system will lockup in no time (not even mouse movement) forcing me to reboot.

Also, it's on battery. So I care who will allow longest browsing without recharge.

Uh... why are you using a desktop OS with 4GB of RAM and no swap? That's pretty reckless. There's no doubt Chrome uses too much RAM, but it's no wonder you get a hard lockup. In either case, how is this relevant? The context here is the importance of benchmarking CPU performance. Pointing out that Chrome uses up RAM excessively, while true, isn't on topic.

Something that calculates faster usually will mean better battery life, but not always. If it means the code is written more cleanly/efficiently then yes, it will assuredly yield better battery life. If it uses more instruction sets or GPU acceleration, the energy usage becomes a bit more gray, especially when accounting for boost clocks (because wattage doesn't scale proportionately with clock speed).

The system borders usage can be disabled in the Chrome settings.

For Ozone? Because that doesn't seem to work for me.

[skeevy420]

Is this a deb or rpm package?

It was packed, man. It's transitioning into a Flatpak.

[unic0rn]

Chrome don't works on Wayland natively, but Firefox works. Chrome Wayland support currently work-in-progress, but only with Ozone abstraction layer, that by the way have dozens of issues, during Firefox works on Wayland full natively and I don't see any issues. On benchmarks Chrome works about 2x faster, but on practice I don't see it during every day web browsing. I used Chrome during many years, but switched to Firefox for native Wayland support.

Personally I prefer Chromium/Chrome, Stadia being one of the reasons.

That being said:
- hardware video decoding under Linux is disabled by default
- enabling it works in general, at least it's bug-free under Stadia
- ...but try opening main page of something like imgur.com (at least on AMD's Kaveri) and watch amdgpu spam kernel errors

I've switched to macOS on an iMac. All those benchmarks are meaningless when the actual user experience is just, frankly speaking, shit. And yeah, vaapi doesn't even work via Ozone at all - heck, KDE itself still has a ton of issues under Wayland, but that's how Linux on desktop feels. Neverending beta. They rewrote kwin's compositing and still can't get stutter-free animations even under X11, when compiz had that decades ago. The fuck are they doing, moving BACKWARDS?

I got an iMac from 2012 from the company I work for, and using Catalina for a few days was an eye-opening experience. When you realize how far behind quality-wise Linux on the desktop actually is, it just makes you sad. All that tinkering to try to make things work, things that are screwed up by design, is just a waste of time.

And it's not Linux to blame, or KDE, or Wayland. It's the attitude of 3rd party software developers - important software, that people are using daily. Even when it's available, quality bar is usually at the "be happy that it's here and launches at all" level. That is NOT enough.

[unic0rn]

"saves over 17 years of CPU time daily"

Yeah, and at the same time you lose millennia while decoding videos on those CPUs. Any crappy PC or even portables like the PSP could decode YouTube on hardware back in early 2000s, then came google and changed codecs to some crappy vp8, and on Linux hw decoding with chrome(ium) still doesn't work at all.

h264ify + enable vaapi in chrome://flags - it'll max out at 1080p (1440p/4k are vp9 only), but it's better than nothing.

[unic0rn]

My experience with Chrome+Ozone is fine. Not great, but fine. I don't like that it uses the system borders and changes the mouse pointer on me but that's about the worst of it.

You are right that in real-world usage, the performance difference hardly ever matters.

With something like Stadia, Ozone is completely unusable - and that's not even due to lack of hardware video decoding, it's just the way mouse works under it.

Wayland has a LONG road ahead.