11-Way Low-End & Older Open-Source GPU Linux Comparison

[dungeon]

I did kernel bisecting on my single-core Intel Atom tablet. It gets faster the further you get into bisecting, as fewer files need to be recompiled every time.

That sounds to me you write a patch then .

Just kidding... maybe i will do if i got time for that or maybe developers get fixed it in the meantime especially if newer radeons are also affected .

[GreatEmerald]

I would like to avoid git bicesting because of compilation time which can easely be numbered by days on Athlon XP .

I did kernel bisecting on my single-core Intel Atom tablet. It gets faster the further you get into bisecting, as fewer files need to be recompiled every time.

[dungeon]

Sounds like you need to bisect it.

I would like to avoid git bicesting because of compilation time which can easely be numbered by days on Athlon XP .

[GreatEmerald]

No i use much more older than that, it is r200 . I can even play 720p videos with subtitles with mplayer wihout tearing on that and Athlon XP with 70-80% CPU usage , but after this 3.14 regression i can't anymore ...

Sounds like you need to bisect it.

[dungeon]

@dungeon

Do you use r600g or radeonsi? r600g shouldn't tear in XV unless you explicitly disabled vsync, glamor, well, its other name is tearing until someone creates a new GL extension.

No i use much more older than that, it is r200 . I can even play 720p videos with subtitles with mplayer wihout tearing on that and Athlon XP with 70-80% CPU usage , but after this 3.14 regression i can't anymore ...

So doesn't metter, if someone want to try to play these with any radeon (i am not sure is there also some difference in CPU usage?) and if it is global regression that is more arguments for me to fill a better bug . For example some of these with kernel 3.13 i 3.14 using gl output:

http://samples.ffmpeg.org/game-formats/idroq/

mplayer -vo gl http://samples.ffmpeg.org/game-formats/idroq/jk02.roq

Is there some difference in CPU usage with 3.13 and 3.14 kernels no metter whatever mesa version is ?

[curaga]

@dungeon

Do you use r600g or radeonsi? r600g shouldn't tear in XV unless you explicitly disabled vsync, glamor, well, its other name is tearing until someone creates a new GL extension.

[dungeon]

I wolud like if someone with radeon can try to play some video with mplayer using gl video output:

mplayer -vo gl some.file

With 3.14 and 3.13 kernel and to watch CPU usage.

For me in 3.14 CPU utilazation is doubled (45% more) . xv output not affected, but i not use it because of diagonal tearing and subtitles is not good like with gl .

Also in most gl games when some video is played it use more CPU like with that mplayer example .

[Azrael5]

hd3850 agp

Is it possible to bench ati HD3850 AGP, one of the best if not the best latest AGP card ever? It should outperform the hd4670. Thanks.

[dungeon]

All of the tested graphics cards were easily CPU bottlenecked when running World of Padman on the AMD E1 Kabini APU system. Why these results are being shown though is that while CPU bottlenecked the Nouveau performance was slightly better than the R600/RadeonSI Gallium3D performance, due to potentially lower CPU utilization.

Running Xonotic with low quality image settings showed the GPUs again being bottlenecked by the low-end dual-core AMD APU, but it's interesting to show again the GPUs with the Nouveau graphics driver were running faster than the R600/RadeonSI drivers when being CPU bottlenecked.

I also spotted that high CPU utilazation with 3.14 kernels in some games at cartain places with older r200 radeon too, 3.13 was better .

[GreatEmerald]

I don't do that for every article due to only having one WattsUp USB power meter and am oftenly doing tests concurrently on many systems...

That's fair enough. Though it would be nice to have an article on low-end power consumption comparison (including an idle power consumption graph), since low-end parts usually end up in places where power consumption matters (netbooks, laptops, HTPCs).