NVIDIA Slaughters AMD Catalyst On Linux In OpenGL 4.x Micro-Benchmarks

[Britoid]

Still waiting for good open source NVIDIA drivers so I can use their hardware, until then its a paper weight.

Or just use the blobs.

[mitcoes]

Can someone explain why GTX 750ti is outperforming the GTX 780ti in a bunch of the tests? The performance just doesn't match. Shouldn't it be 780ti > 680 > 750ti? That's almost never the case. I'm not 100% confident about this test. It all smells fishy I tell ya!

Not at every test the 750 Ti is new generation Maxwell GPU, in fact it would have to be 8xx generation and 780 Ti and 680 are Kepler

[mitcoes]

1920x1080 is easier to read 1080p(2k)

I do think that with the 4k name resolution and in order to fast benchmark ...
1920x1080 must me named 1080p (2k) and
1280x720 must be named 720p (1k) and Phoronix should test ...

1k + 2k + 4k resolutions

Said this, OpenGL 4 is a great victory for NVIDIA drivers, and a great defeat for ...

Xbox, PS4 and AMD APUs cheap Steam Machines

So my actual CHEAP Steam Machine config should start with a CHEAP Nvidia 750Ti + any CHEAP AMD CPU configuration.

[brad0]

Or just use the blobs.

You're really that stupid eh?

[dungeon]

I wonder how many of those tests have alternative implementations using AMD extensions.

Apitest use OpenGL 4.2+ and only ARB extensions, that is part of any 4.2 implementation .

[_ONH_]

Yep apitest is just writted by Nvidia for one GDC session to show that there is already "fast path" in their OpenGL dirver .

etc. http://www.slideshare.net/CassEverit...river-overhead

Just to point out... when they say "OpenGL", they mean our OpenGL implementation ;D.

For example one thing that acceleratess OpenGl perf is ARB_indirect_parameters, ARB_shader_draw_parameter, ARB_bindles_textures, ... and all the other GL extentions mentioned in this presentation, will only give OpenGL 4.5/5 "a fast path" until the it is not the case, Nv may hav implementet it in ther prop. driver, but thas about it. To date OpenGl does not have this. Hope this gets implemented son and also get part of core specification.

[Darksurf]

My main issue with FGLRX right now is I can't get my A10-7850K and R7 250 to crossfire
My 6870s crossfire just fine. Wish AMD would fix this, my Steam machine would like some extra GPU power.

Past that, not having any issues. Love my AMD cards. thinking of upgrading my CF HD6870s for an R9 280X.

[dungeon]

But looking in code i see there is actually some NV extensions used in apitest

GL_NV_shader_buffer_load
GL_NV_vertex_buffer_unified_memory
GL_NV_bindless_multi_draw_indirect

And no AMD one i see there .

@_ONH_

Yep that is why i say _their_ implementation .

[Britoid]

You're really that stupid eh?

No, the blobs work much better than the open source drivers, therefore when using nVidia hardware under Linux, I use the blobs.

However the case is the opposite for AMD hardware.

[Sdar]

Yep apitest is just writted by Nvidia for one GDC session to show that there is already "fast path" in their OpenGL dirver .





etc. http://www.slideshare.net/CassEverit...river-overhead

Just to point out... when they say "OpenGL", they mean our OpenGL implementation ;D.

AMD and intel participated on that same conference (Approaching Zero Driver Overhead in OpenGL)