The Current Windows 10 vs. Linux Browser Performance For Google Chrome + Mozilla Firefox

[Bsdisbetter]

You are a scary person, you make me feel bad

Yeah, he's not a zealot, that's scary here...

[Bsdisbetter]

And yet, Windows has serious memory leaks when running browsers, and often isn't capable of shutting one down without leaving multiple zombie processes running. Maybe there's something to be said for clean code?

And you believe this... LOL

[Aeder]

Aside from SeleniumBenchmark: MotionMark, Windows wipes the floor with Linux.

Nah, with default settings (which is what the test used) a shit-ton of performance options are disabled on Linux, this has nothing to do with the OS and more to do with Mozilla not prioritizing work to get those functions working under Linux at all.

The fact that it beats the windows version on some tests or even comes close to it with a bunch of accelerated rendering options disabled is already pretty amazing.

[V1tol]

The MotionMark results are really odd. On Windows, both browser use GPU accelerated graphics. In Linux at least Firefox doesn't (and it will stay that way until Webrender gets enabled by default)

WebRender wasn't working on Windows also, unless it was forced (I did not find a statement of this). Because on Windows it is turned on by default only for nVidia GPUs. I have 1070 and on Windows Firefox in terms of CSS effects and other GPU related stuff kicks ass every browser I tried.

[eltomito]

Apart from the obvious, it's probably also the advantage of closed source, no sarcasm. Nobody in their right mind releases in some stupid shit Debug mode (because it leaks source code info), while it's "the default" on some crappy build systems in open source because why not. Open Source devs tend to not give a shit about the end users or end binary quality (how can you "forget" something is in debug build, wtf, they don't even inspect their binaries), just to keep their stupid code clean.

I worked for a commercial software company for a while and my experience was the exact opposite. They actually made it their policy not to care about performance unless the client (i.e. the end user) specifically complained about it. The rationale was that unless the client complains the software is too slow, then trying to make it faster is just a waste of time, i.e., waste of money. I'm serious.

[sireangelus]

something is very wrong here. I get those results on mint 19.04 with a 4770k at 4.3ghz, while windows is usually slower by 2k to 3k points in octane.

[AndyChow]

1. You can do that on Windows too.
2. That is a bit anecdotal. For commercial windows development, I've only ever used VC++.
3. When there are more apps you want to run at the same time, that RAM goes fast Also, you must not be running KDE/Gnome.
4. You can get that with Cygwin or Windows subsystem for Linux

Throw Linux in a docker container and call it a day. Not hat'n, just stat'n.

I agree with you and liked your comment. However:

1. You can cut the bloat way down in linux, and even go headless (although that's getting harder and harder, even if you only SSH into the machine, you need the graphic dependenciers more and more)
2. For portable code dev on C, nothing beats GCC, by a long shot. Sure, you can cygwin that. But you can't build and distribute the entire world, which is easy in linux (and even more in BSD, especially DragonflyBSD with Synth). You can't instantly create instances of Windows without licence headaches
3. The RAM is cached data, so it's not gone, and made free whenever you need it. Unused RAM is wasted RAM. Also, I don't think (might be wrong) that you can enable user-space hugetables in windows. You sure can in linux, and it sure speeds up stuff a lot when page trashing is the bottleneck. It is the bottleneck very often when dealing with a lot of data.
4. Windows subsystem is now POSIX compliant, which is fantastic. Great point.

5? Linux container, using hyper-V, is way slower than on linux. nvidia-docker is not available on windows, and probably never will be. Finally, Hyper-V is way slower than qemu-kvm.

[hubicka]

Thanks to Martin Liska LTO was recently enabled for opensuse build of Firefox. We did not enabled PGO yet because there is an issue with running the training run in the build service environment (Martin knows details). I plan to meet tomorrow also with the maintainer of Chrome package to get LTO+PGO enabled for Tumbleweed builds. So some work is being done on this (though I would wish it progressed faster, too)

But forgot to mention, independently of LTO/PGO issues, the results seems to be affected by something else. Some of benchmarks with shows regressions are almost fully about JIT produced code that should be about the same for all Linux distros at least.
Balloo sounds like an possible explanation. It is unlikely because of btrfs - those benchmarks are CPU bound.

[hubicka]

I tried to bring Tumbleweed's perf to the attention of the community, but everyone turns a blind eye, and even rbrown of opensuse calls this site trash. I don't know what's going on. I have no problems with my installation (though I run ext4 and turn off mitigations), but I'd simply like to know the root cause of the perf delta. Default file system? Kernel? Boot options? Compiler flags? Cpu governor? Firefox options?

Also, it would help to know if this was on an Intel or AMD system.

Thanks to Martin Liska LTO was recently enabled for opensuse build of Firefox. We did not enabled PGO yet because there is an issue with running the training run in the build service environment (Martin knows details). I plan to meet tomorrow also with the maintainer of Chrome package to get LTO+PGO enabled for Tumbleweed builds. So some work is being done on this (though I would wish it progressed faster, too)

[JeansenVaars]

1. You can do that on Windows too.
2. That is a bit anecdotal. For commercial windows development, I've only ever used VC++.
3. When there are more apps you want to run at the same time, that RAM goes fast Also, you must not be running KDE/Gnome.
4. You can get that with Cygwin or Windows subsystem for Linux

Throw Linux in a docker container and call it a day. Not hat'n, just stat'n.

You are a scary person, you make me feel bad