Intel HD 4000 Ivy Bridge Graphics On Linux

[apholux]

Windows OGL version support on HD4000 - OGL3.3? HD4000 Win vs. Linux report?

I read somewhere that HD4000 supports OGL3.3 on Windows. Is it true? When comparison between HD4000 Linux&Windows will be available, like the one done some time back for HD3000? That report was a great job.
Thanks

[smitty3268]

I'm surprised how close AMD and Intel are in some of the tests. I thought AMD is much faster.

Llano is a lot more powerful when it comes to texturing and shaders, but Ivy Bridge actually has the lead in raw pixel pushing power and memory bandwidth.

So it's no surprise to see it hanging around in tests with engines like Quake 3. Expect the differences to really show up in tests with Unigine and WINE tests that are more shader bound.

I really wish Michael would have set the 3870 to high power profile, so we could have seen a fair test between the hardware on Mesa drivers.

[russofris]

Why? Catalyst has excellent 3D performance compared the to open-source radeon driver. Have you ever seen a changelog for Catalyst? There are often a bunch of per-app optimizations.

The short answer is this. In my experience (5750 and others) catalyst-linux tends to underperform its Window's and Xbox counterpart by a remarkable amount. To refer to something as "optimized", when you're only achieving 30-50% of the hardware's maximum performance does a disservice every other piece of optimized software out there.

Call it "AMD's proprietary Catalyst Linux driver". Call it "AMD's highly-polished Catalyst Linux turd". Calling it "highly optimized" only serves to diminish the concerns of those that have had to live with the performance shortcomings of FGLRX for the past 5 years.

F

[jrch2k8]

In windows, the Llano chips seem to usually beat the Intel ones fairly easily in graphics workloads, but it seems either the Intel Linux drivers are much better than the Intel Windows drivers (very possible), or the Catalyst drivers in Linux are slower than in Windows.

fglrx is a lot slower than DX windows catalyst and Xbox and even on windows opengl is considerably slower than DX but is still faster than opengl in linux

[Kayden]

Thanks for the great article, Michael!

For what it's worth, the VDrift stable release has a few bugs---it tries to use the texture2DLod function in the fragment shader, which isn't available unless you enable the GL_ARB_shader_texture_lod extension or use GLSL 1.30. Which it doesn't, so we say "no", and they get bad rendering. I went to go submit a patch for that (it is open source after all!), and it turned out that they already fixed it; they just haven't released an updated version. Git snapshots of VDrift seem to work just fine on Sandybridge/Ivybridge.

[Alejandro Nova]

This is one part of what we were waiting for.

The other one is: We know that in Windows the Brazos platform destroys Atom. I'd like to see a comparison between Brazos and Atom, with both Catalyst and free drivers. I think Atom could destroy Brazos in Linux, thanks to better drivers.

[Kivada]

I'm surprised how close AMD and Intel are in some of the tests. I thought AMD is much faster.

The Intel HD4000 is a year newer then the tested AMD 6550D and still comes up short, in a few weeks AMD's new parts in the same price bracket will be out, with the even faster HD7660D http://www.legitreviews.com/news/12666/ So again to Intel, better luck next time guys.

[Kano]

Well maybe compare intel vaapi against amd xvba (if you like xbmc fernet menta or via xvba vaapi wrapper) - then you know what chip is better...

[jrch2k8]

Well maybe compare intel vaapi against amd xvba (if you like xbmc fernet menta or via xvba vaapi wrapper) - then you know what chip is better...

intel prolly wins hand down, well ofc unless michael finds an combination of settings and a video that xvba can actually render without quality loss or creepy render error or kernel lockups, but even then since the cpu parts is a lot faster than llano i think is hard to determine which one is more effective since cpu usage in IB should be shockingly low compared to llano even if the gpu part is more efficient in llano. so dunno it probably will be hard to tell

[Kano]

mpeg2 is not accellerated with xvba, h264 l5.1 is also not possible with xvba. if you select other testfiles xvba should work. in theory there is a h264 encoder you can use via (intel) vaapi es well but i don't know if there is an app using is yet (besides the test tools from libva).