AMD A10-7850K Kaveri: Windows 8.1 vs. Ubuntu Linux

[_SXX_]

Are the Catalyst drivers really that bad in terms of system stability? I've been wanting to slap together a mini-PC and was considering using an AMD APU.

If you want to use something like Debian stable they're not that bad, but they still have lot of issues in games I play.
With never kernels, software and X proprietary drivers have lot of issues for me. Though I'm didn't tested latest proprietary drivers yet so I can't say for sure what if there many issues with them.

[guido12]

Are the Catalyst drivers really that bad in terms of system stability? I've been wanting to slap together a mini-PC and was considering using an AMD APU.

[jrch2k8]

I don't play most of those games, doesn't make my point any less valid.

Anyway, after a month of talking with people at various forums, reddit and most important, exchanging emails with AMD support lead engineer which, so he told me , was in contact with Linux team about my issue (I've send him things like xorg logs, atisysteminfo and reports about performance of multiple games under 3 different Ubuntu versions with 3 different drivers, so far) conclusion is simple - AMD drivers are problem here.

point 1: you are bitching to the wrong crowd, you need to bitch those studios to include benchmarks tools at least equal to their windows counterparts (or make a fund to pay the article since it has to be done manually)

point 2: yes, fglrx is always the problem, even if 1 thing go fine and fast other 50 will totally break because of it. OSS drivers are dreamy though for all but heavy 1080p gaming

[djdoo]

Yes, it's ridiculous. Talk about entitlement.

- Oh Michael, all this work you do for free, in the site I use for free, well, I want you to do in in my terms, or I'll bitch like there is no tomorrow. Because I deserve it.

And yes, I'll ignore the troll, no need to read any more bitching. Luckily I can add it to the troll-list.

Sigh, kids!

I totally agree.
For me Unigine Valley and Heaven 4 are two of the most grafically impressive 3D apps I ve seen so far and work fully on linux. Tessellation is a feature of openGL 4.x and both of them use it strongly!

[gutigen]

If the games you care about so much don't play well when tests like Heaven prove capable, that's either your fault or the developer's.

I don't play most of those games, doesn't make my point any less valid.

Anyway, after a month of talking with people at various forums, reddit and most important, exchanging emails with AMD support lead engineer which, so he told me , was in contact with Linux team about my issue (I've send him things like xorg logs, atisysteminfo and reports about performance of multiple games under 3 different Ubuntu versions with 3 different drivers, so far) conclusion is simple - AMD drivers are problem here.

[ua=42]

Sigh.... I wonder if time would be saved if I wrote a script to auto-post this to the forums each time a benchmark is done....

Phoronix does not benchmark <insert game here> because it does not have a benchmark mode for Linux or has a benchmark that is hard to automate.

These common tests use the following opengl levels.

Unigine requires openGL 3.2 (and higher if available by the driver)
Xonotic openGL 2.0
Open Arena openGL ES

As you can see, they cover a selection of the openGL feature set. Performance improvements detected by these benchmarks are likely to indicate improvements in other games using the same opengl level.

If you want things to change. Please kindly ask game/benchmark developers to create a benchmark of their game engine that can be launched from the command line. Preferably one that is self contained so it can easily be transferred to the test machines.

[mendieta]

50% faster than Haswell

So, I have a Haswell i5-4670k with pretty fast RAM. I am getting abou 15fps in Unigine Sanctuary, same resolution as in the article (1920x1080)

http://openbenchmarking.org/result/1401139-MEND-140103012

That means that Kaveri is 50% faster in the test. Having said that, I still prefer the open source support of my chip, but it's nice to put the numbers in perspective. This APU is quite nice.

[mendieta]

Unigine Valley hardly is OpenGL 2.0...

Trolling...

Yes, it's ridiculous. Talk about entitlement.

- Oh Michael, all this work you do for free, in the site I use for free, well, I want you to do in in my terms, or I'll bitch like there is no tomorrow. Because I deserve it.

And yes, I'll ignore the troll, no need to read any more bitching. Luckily I can add it to the troll-list.

Sigh, kids!

[nadro]

TF2
CS:source
DOTA2
L4D2
Portal

Those games use the same engine - Source Engine, so all those games should show the same bugs. In this case you can benchmark just 1-2 games.

[schmidtbag]

Given the age of Linux, the age of catalyst/FOSS and the growing age of decent games that use catalyst,mesa,steam,SteamOS,latest kernal im fairly mind blown someone doesn’t on the entire planet create a decent easy to follow benchmark site for games on Linux.

Well that's sort of the issue with many visitors on phoronix - they treat this website as a place primarily for gaming and hardware recommendations, when that's really just a side objective (and yes, I know the title says otherwise but the title really needs to be changed to suit the content). There really needs to be an entirely separate website dedicated to benchmarking games and hardware in a consumerist perspective; phoronix seems to largely target enthusiasts, admins, and developers. I see no problem in how phoronix runs tests, and I think taking a consumerist approach would taint the website.


I've considered starting a website myself for strictly linux gaming and hardware reviews, but considering I don't have the time or money to pursue such a thing, I dismissed the idea pretty quickly. Besides, you'll never satisfy everyone. Linux users are hard to please - they're generally arrogant, fussy, and nitpicky.