Windows 8 Outperforming Ubuntu Linux With Intel OpenGL Graphics

[przemoli]

It doesn't make sense to benchmark using games which are designed for discrete GPU on Intel GPU. If you are serious gamer you will need discrete GPU anyway.

Still AAA publishers/game dev do want to target Intel GPU's, as those are popular. So good results on Intel GPU are very good news for everybody (especially if you just need to update your gpu drivers..).

[Rexilion]

It's quite possible Linux/Open Source Intel drivers will never reach the performance level of Windows drivers, because the latter are close source, thus Intel can secretly cheat in Windows but they cannot cheat in Linux, 'cause their secrets/hacks/subterfuges will be instantly unearthed by their competitors and journalists.

I find that really hard to believe. If Intel is already developing drivers for their hardware, aren't they already giving away all the details?

You are right if you are talking about specific parts of a video card like third-party video decoders and such.

And besides, optimizing a driver is not cheating. Intel is just not putting as much energy in optimizing the Linux drivers.

And what is wrong with placing a dirty hack in the driver you develop on your own? Practically speaking, nouveau was full of hacks since the developers were still figuring things out on how to handle it. After a big rewrite in the kernel, libdrm and mesa things are starting to look 'nice'. A very nice example is this commit (link!). It is a hack because it will not work under all states of the world. But it works!

EDIT:

People start thinking for a moment..

1) Valve DO NOT ALLOW for benchmarks of Steam nor games on Steam for Linux... RTFE(ULA). Its beta. So no nice TF2 benchmarks in near future...

2) Valve support UBUNTU, in STOCK configuration. Benchmarking stock Ubu vs stock Win8 make sense.

3) Installing another DE/tweaking Unity/Compiz is easy, so people wanting to get most perf will do it. But general benchmarks should be useful for general audience :P

4) You can rerun those tests on your configurations, and upload to openbenchmarking.org Show the results, instead of molesting poor keyboards

Przemoli for president

[schmidtbag]

People start thinking for a moment..

1) Valve DO NOT ALLOW for benchmarks of Steam nor games on Steam for Linux... RTFE(ULA). Its beta. So no nice TF2 benchmarks in near future...

2) Valve support UBUNTU, in STOCK configuration. Benchmarking stock Ubu vs stock Win8 make sense.

3) Installing another DE/tweaking Unity/Compiz is easy, so people wanting to get most perf will do it. But general benchmarks should be useful for general audience :P

4) You can rerun those tests on your configurations, and upload to openbenchmarking.org Show the results, instead of molesting poor keyboards

First comment you posted in this topic I fully agree with.

[ChrisXY]

2) Valve support UBUNTU, in STOCK configuration. Benchmarking stock Ubu vs stock Win8 make sense.

Yes, but the article sounds like it is supposed to be a driver benchmark. So it would still be nice to know if this is the best the driver can do. So maybe additionally, ubuntu with openbox or so and windows without aero could be benchmarked (can aero be disabled in windows 8?)

It doesn't make sense to benchmark using games which are designed for discrete GPU on Intel GPU. If you are serious gamer you will need discrete GPU anyway.

By the way, Team Fortress 2 and Counter Strike Source and Half Life 2 Deathmatch run pretty good on intel. Only serious sam 3 is not really optimized for it (and not officially supported at the moment).

[Vim_User]

2) Valve support UBUNTU, in STOCK configuration. Benchmarking stock Ubu vs stock Win8 make sense.

3) Installing another DE/tweaking Unity/Compiz is easy, so people wanting to get most perf will do it. But general benchmarks should be useful for general audience

This actually doesn't make any sense. This benchmark was neither to compare Steam nor Steam games with their Windows counterpart, so your argument #2 is irrelevant. So is argument number three, since this was not a benchmark about squeezing the last bit performance out of a DE/WM.
This was a comparison for OpenGL rendering on Windows and Linux on Intel hardware. If you can't tell if a lower performance is caused by a well known, but unpredictable, performance regression in Unity or if lower performance is indeed caused by the Linux Intel driver the whole article is missing the point totally and therefore nothing but a waste of time.
The astonishing thing is that this was told to Michael several times, but it seems that he does not understand that.

[schmidtbag]

This actually doesn't make any sense. This benchmark was neither to compare Steam nor Steam games with their Windows counterpart, so your argument #2 is irrelevant. So is argument number three, since this was not a benchmark about squeezing the last bit performance out of a DE/WM.
This was a comparison for OpenGL rendering on Windows and Linux on Intel hardware. If you can't tell if a lower performance is caused by a well known, but unpredictable, performance regression in Unity or if lower performance is indeed caused by the Linux Intel driver the whole article is missing the point totally and therefore nothing but a waste of time.
The astonishing thing is that this was told to Michael several times, but it seems that he does not understand that.

I see what you mean but he brought up his argument #2 in response to his argument #1, so you should probably look at them as 1A and 1B. As for argument #3, that is relevant. A good benchmark will test all major conditions. With your argument, that would mean testing different screen resolutions doesn't matter either. Previous tests have shown that Unity has had a pretty drastic performance drop compared to other DEs in multiple types of tests. AFAIK, compositing doesn't turn off when you run full-screen programs, which drains performance. Windows 8 most likely disables Aero during fullscreen programs, which gives it an automatic lead. Unlike Windows, Linux isn't a 1-size-fits-all OS, so comparing 1 variation of it isn't enough of a test when it comes to a test like this.

[startzz]

Yep, this thread is troll
1) Like its a big surprise, that windows is faster in everything than linux... lioooool
2) Linux cant be really tested for gaming, because it doesnt have any real top games to be tested, same few shitty low level games are being tested like for 50 years... who fkin cares about ice age gaming in linux...
2.1) steam for linux still didnt do much in linux gaming for now, it still doesnt have any great games like "call of duty : blackcops 2" or "need for speed most wanted 2012", just some primitives game for babies... for this situation to change, it will take many years, i mean, it will take at least 10 years for first big games to appear in linux, and it will be just a few top games, not like hundreds of games that are available for windows right now
3) linux isnt developed for being number one, it is made for learning purposes, so the situation about gaming in linux wont change.

[gururise]

Great benchmark Michael! I agree with others that the games benchmarked (Open Arena, Xonotic, etc) are very relevant to a driver comparison, because you want to keep the same codepath (openGL) and reduce the number of different variables as much as possible when looking at driver performance between two distinct platforms.

Also, to those saying Intel Gfx are irrelevant... Intel has the best OSS graphics drivers of any gfx manufacturer. Would you prefer to see a comparison of Radeon OSS vs Windows Catalyst? We all know who would win that one.

[totex71]

ofcource even windows 8 has better performance than ubuntu. ubuntu is slow as hell compared to other distros and specially source based ones like gentoo.
i get over 3 times better graphics performance in gentoo than on ubuntu, so no. this test dosent show at all that windows 8 has better performance than linux. only that it has better than ubuntu.

[Michael]

As shown in other more recent Phoronix tests, most of the Unity/Compiz overhead has been eliminated with the latest Raring packages... As used in this benchmark. But for those in disbelief that the Linux OpenGL performance can lose to Windows, I'll happily run some more tests from Xfce and KDE on this laptop to show those results...