05-05-2010, 08:14 AM
Now I got what susikala is standing for. His first choice of words were quite poor, indicating a personal attack on phoronix, but now that the talks are again in a decent level, we can focus back on the article.
I have to agree that the text doesn't point straight where it should be clearer than water: the fact that those results are great for Linux.
Gaming is one concern of a myriad, surely. But an important one. We cannot ignore that the gaming industry is quite powerful and has recently been the fuel to GPGPU development. CellBE is the best example.
Talking about CellBE/PS3, we should always remember that gaming console systems are usually a tweaked FreeBSD close-sourced (as the PS3 is). Showing the potential of Linux in this area is indeed important and I believe this article is exactly about that.
Is the article imperfect? Lacking information? Unclear conclusion? Ok, but as always, there's no other place to gather such information but on phoronix and I hope they continue to do so.
05-05-2010, 08:18 AM
Nvidia CUDA is the best example of GPGPU, I cited CellBE as a devel model that used the same principles: gamers paying for development.
Originally Posted by Caveira
05-05-2010, 08:31 AM
For the ones that are not boolean options, I provided the options that should be selected. 1000Hz is an example of that.
Originally Posted by deanjo
05-05-2010, 07:45 PM
I'm just saying that Ubuntu's kernel config isn't leaning one way or another. It's more of a neutral config then anything. There are a hell of a lot of tweaks to their base kernel that would optimize it for server use. The configs they have by default are not ideal for server use either.
Originally Posted by Shining Arcanine
05-06-2010, 06:08 AM
From the perspective of a desktop user, the kernel configuration is optimized for servers in that the settings that a desktop user would want to toggle have been toggled in the opposite direction. What settings would you need to toggle in addition to those if you wanted to optimize it for a server?
Originally Posted by deanjo
05-06-2010, 07:40 AM
This is a very interesting article.
What I gather from it is:
- there's still an obvious gap between Linux and Windows performance, however, the gap is certainly smaller than I expected.
It looks like Ubuntu *could be* considered also for some gaming.
- I cannot believe some of the comments stating that Ubuntu and Windows are substantially equivalent from gaming perspective: the data clearly shows otherwise. I love Ubuntu and all, but ignoring a gap is not going to help closing it.
- I would have liked also to see some comparisons with Directx9, 10 and 11 titles.
How does the graphic of a Directx11 title compare with Ubuntu?
What about PhysX engine? Is it accelerated in Nvidia cards?
Considering how 'crude' are the drivers in Linux compared to Windows, I'm actually quite amazed on how good they were able to keep up with gaming...
05-06-2010, 08:10 AM
When Unigine Heaven would be officially relased with OpenGL 4 support then you could bench it against DX11. Currently you could bench it against DX10. As there is no PhysX support for Linux you can not compare that anyway. Wine can run games with PhysX just like those games run on ATI hardware without hardware support but benching that is somehow pointless. PhysX is using CUDA so in theory it could run on Linux too - UT3 engine could be one of those, but not available for Linux. It seems the UT3 Linux port is dead.
05-06-2010, 03:38 PM
Originally Posted by Kano
my point about DirectX9/10/11 was that graphic quality also plays an important role in gaming. Say that 2 games have the same exact performance in the two platform, but the quality of the graphics is very different, would you say that it's a fair comparison?
We're comparing apples to oranges anyway, but it's important to understand what the differences are.
Frankly, if one is a hard-core gamer and/or is into the latest and greatest titles, then the choice is simple, whether we like it or not.
I am a casual gamer and I'm plenty satisfied with Linux's games and performance, and I am certainly not switching back to Windows just so I could play some more titles.
One thing is clear from this test: Ubuntu is a little behind, but considering how much more effort goes in Windows (in terms of drivers, game developers ...) it's quite amazing actually.
05-06-2010, 03:45 PM
The DX9 vs DX10 question only interests XP users (there you can not use DX10), basically OpenGL 3.x provides the same features as DX10. But when you test DX10 games using wine then you will suffer from a huge performance penalty. Best is when the game itself provides different renderes. OpenGL based games for Win usually do not run that much slower than native when executed via wine - but still there are not that many which would be a good comparision. Maybe UT2004 (demo) could be used too - with different backends.
05-06-2010, 04:48 PM
I wasn't suggesting to use Wine.
Originally Posted by Kano
Choosing the same games is an obvious choice, but it's limiting. Most of Windows games use DirectX.
I think none of the games tested support directX, which makes the comparison a bit more fair, but prevents us from drawing too many conclusions:
1) Is Ubuntu faster/slower than Windows in gaming? Well, the test seems to indicate that it is a bit slower, but not by far.
2) Is it Ubuntu faster/slower than Windows in general? We can't say, because using games puts a lot of weight on the video drivers and are not really representative of regular (= non-gaming) PC use.
3) Can we say that Ubuntu is a good gaming platform?
If you're only looking at OpenGL, then yes, but, in general, today, much of the gaming market is -unfortunately- on DirectX, which means if you want to ride the latest&greatest titles, you have to stick to Windows.
4) What does it mean "faster" in a non-gaming environment? Does your HD start cranking for minutes, inexplicably? Do you need to reboot when you install an update? Do you need to run several background services to reduce the risk of being infected? Does your boot time increase, constantly, for no known reason? Do you see a longer and longer lag between mouse clicks and windows reactions? Does your PC become sluggish when you have several programs open?
How do you quantify all of the above?
Not easy. But if it was, I think we all know what the answer is.