> Not to detract from this very nice development, but it's still the case on many laptops that installing Linux tends to extend battery life substantially, not to mention make graphics more stable (at least, given FOSS-friendly graphics hardware).

I have never heard this and in fact have only ever heard the opposite.

The open source AMD graphics drivers in Mesa are considerably more stable than the equivalent Windows drivers and the proprietary Linux drivers (though those are getting a bit better). In many cases the open source drivers outperform the proprietary ones. The Mesa Intel drivers have been rock solid since about 2010, I've not had an issue with the stable release of Mesa on standard Intel hardware since then. I've run an average of two models from each generation of Intel hardware since 2010 except Broadwell, and not encountered an issue.

Switchable graphics can be a bit of a pain, but with a bit of fiddling around with DRIRC and a PRIME-capable GPU driver it can be had without too much pain.

> The open source AMD graphics drivers in Mesa are considerably more stable than the equivalent Windows drivers

What?? Not in my experience! Windows can even recover from graphic driver crashes.

> In many cases the open source drivers outperform the proprietary ones.

Maybe for OpenGL. But DirectX 11 is nearly always faster (e.g. bytecode shaders) and only available on Windows.

> Maybe for OpenGL. But DirectX 11 is nearly always faster (e.g. bytecode shaders) and only available on Windows.

This statement utterly untrue, there is nothing especially performant about D3D11. FYI, "bytecode shaders" have nothing to do with shader performance. They have some impact on initial parse time, but little else.

> there is nothing especially performant about D3D11.

Not in D3D11 itself, but due to it being the prevalent API, drivers are optimized for it. See https://medium.com/layerth/benchmarks-dota-2-streaming-feat-... for example.

> FYI, "bytecode shaders" have nothing to do with shader performance. They have some impact on initial parse time, but little else.

You're right, it shouldn't make that much of a difference, but I think driver/compiler optimizations for HLSL (and lack thereof for GLSL) is one of the reasons, games are faster on Windows using the AMD driver than Mesa. It also plays a role that most ports are using HLSL to GLSL translators.

There definitely are cases where the open source driver outperforms the proprietary Linux driver. But I haven't seen a case where a game runs faster on Linux than on Windows with AMD hardware.

IIRC L4D does. Valve wrote a blog post about it

The comparison was 270 FPS (Windows D3D) vs 315 FPS (Linux OpenGL). At those high framerates I don't think that any performance boost is relevant anymore. Furthermore Left 4 Dead is using D3D 9 IIRC, so pretty outdated.

http://blogs.valvesoftware.com/linux/faster-zombies/

That's a massive cherry pick right there. People who, you know, buy GPUs to play games, run windows because the experience is _always_ better.

Laptop's with switchable gpu seem to be challenging to get to work as designed wherein the integrated gpu is the only one used much of the time.

You probably don't use Linux.

That's a weird assumption to make but I use Linux for both my work and personal laptop as well as my work desktop.

I have Linux running on a Lenovo T420s. Intel everything - chipset, graphics, etc. - stuff that should be supported in Linux great. If I run Chromium, it hard locks my screen with visual artifacts unless I disable GPU acceleration. No problems running in Windows at all (including the normal composited desktop environment, Chrome, etc.). I find that even with proven non-bleeding-edge hardware with known drivers, battery life and graphics stability are far worse.

Lenovo T420S has an Nvidia GPU from what I can see, it's probably trying to use that but there's some issue getting nouveau's acceleration working.

And battery life on thinkpads often needs tinkering, either through kernel parameters or tp_smapi if supported.

Same problem with the T420, T420 and T440 I have had. The T440 I couldn't get more than 3 hours out of it compared to 7-8 on windows. They had the PM set up properly. Powertop shows loads of things constantly waking the devices up.

Have you tried TLP? It was originally written specifically with Thinkpads in mind, but can be used on any laptop now: http://linrunner.de/en/tlp/docs/tlp-linux-advanced-power-man...

Yeah I burned hours on powertop and TLP.

I installed Linux on my company Mac and it feels like my battery charge lasts less long

Linux on mac will always have shorter battery live. That's mostly because of Apple proprietary code required for power management.

I will say it's mostly due to the amount of effort creating power-saving features in OSX that are not present in other OS's. It's quite interesting all the little things they do to prolong the battery.

I have Fedora 25 with TLP, it operates fine, there is nothing proprietary blocking it (but you do need TLP, install it asap to get any batt life out of Linux) on my macbook and I can almost reach 5 1/2 hours web browsing (Opera with battery saver). On OSX I can easily obtain 9 hours doing similar work. This is all on a 6 year old battery too.

Can confirm. Also, Windows actually saves batteries compared to a vanilla Ubuntu install. I don't know why but Linux GUI is extremely wasteful for some reason.

It's not the GUI, power management tools in Linux are somewhat behind the curve. Or so I'm told anyway

> not to mention make graphics more stable (at least, given FOSS-friendly graphics hardware).

I hope you're not thinking about Intel integrated graphics :-). I had a terrible experience with stock drivers in Ubuntu LTS (both current and previous) so much that I had to install oibaf PPA to be able to play the few 2D indie games that I care of.

Not my experience, actually rather the opposite especially in terms of battery life and sleep/wake issues.