Page 1 of 1

SuperTuxKart 0.8.2

Posted: Mon Feb 02, 2015 11:49 am
by xeno74
Hi All,

SuperTuxKart 0.8.2 beta2 was released last week. I tested it with my Mac Pro 2009 yesterday. It requires a very powerful computer. It works without any problems with an AMD Radeon HD 5870 PCI express 2.0 x16 graphics card on ubuntu MATE 14.04 LTS x86-64. STK told me the driver is too old but I have an AMD Catalyst driver installed. I installed the latest AMD Catalyst driver 14.12 and the message about too old graphics driver was gone. I'm very impressed about the new STK. The graphics are fantastic. There are a lot of animated things in the tracks. I love it.

Screenshots:
1416193997.jpg
1416192730.jpg
1416194166.jpg
I also tested STK 0.8.2 beta 2 with the open source Mesa graphics driver on the same computer yesterday. There wasn't a message about an old graphics driver but at the begin the race, my system hanged and there wasn't any screen output. STK didn't respond to the keyboard. My system was froze and I couldn't reboot it.

Code: Select all

OpenGL vendor string: X.Org
OpenGL renderer string: Gallium 0.4 on AMD CYPRESS
OpenGL core profile version string: 3.3 (Core Profile) Mesa 10.1.3
OpenGL core profile shading language version string: 3.30
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile
OpenGL core profile extensions:
OpenGL version string: 3.0 Mesa 10.1.3
OpenGL shading language version string: 1.30
OpenGL context flags: (none)
OpenGL extensions:
OK back to our AMIGA one computers. Our latest Mesa is the unofficial version 10.0.4. It's not so fast than the old Mesa 8.0 official but it's fast enough for STK 0.8.1. The unofficial Mesa works on A-EON AMIGA ones and on Power Macs G5 (Late 2005).

STK 0.8.2 has following Linux system requirements:

- Mesa 10.3.X OpenGL 3.1 or a new AMD Catalyst driver
- Unfortunately a powerful computer

I compiled the latest Mesa 10.4.3 on my AMIGA one X1000 last week. Unfortunately STK 0.8.1 is slower with the new Mesa. At least 15% slower. 3D acceleration works but slower on Debian Sid. Mesa 10.4.3 is definitely slower than Mesa 10.0.4.

I will compile a Mesa version 10.3.X next week. But before, I'd like to know if it makes any sense to release the new STK for our AMIGA ones.

Could you run glxgears with vblank_mode=0 glxgears on your AMIGA one X5000, please? I'd like to know how is the 3D performance of the Cyrus boards. Could you also test STK 0.8.1 on your Cyrus board too, please? Please enable all graphics effects and drive all tracks with 8 karts. Please send me the results via PM because of the NDA.

Thanks in advance.

Cheers,

Christian

Re: SuperTuxKart 0.8.2

Posted: Mon Feb 02, 2015 11:53 am
by xeno74
More screenshots:
paradiso-2.png
STK-0.8.2-screenshot1.jpg
STK-0.8.2-screenshot2.jpg

Re: SuperTuxKart 0.8.2

Posted: Sat Feb 07, 2015 3:31 pm
by xeno74
FYI: Unfortunately Mesa 10.3 is slower than Mesa 10.0.4.

Re: SuperTuxKart 0.8.2

Posted: Mon Feb 09, 2015 7:07 pm
by Srtest
xeno74 wrote:FYI: Unfortunately Mesa 10.3 is slower than Mesa 10.0.4.
Speed is not everything Xeno. I've been using Nexuiz for a while now and the frame rate behaviour is something that should be investigated. An equal version on the pc works worse than on the X1000 (supposedlly less raw power and old HD5870). It is not just about straight fps. For a long time I remember a talk on the pc about frame times and micro-stuttering. A lot of that behaviour was overshadowd by heavilly worked-on drivers and the most recent hardware. That is until you got to certain games and discovered all that... streamlined power didn't translate as it should have. Example from the past few years have been console ridden games that are badly ported but also the game engines themselves. When I play some stuff on the X1000 the Pwrefficiant (or else) has something I can only refer to as recoverability - it optimizes its own frame rate use to a smoother experience on less than optimal drivers. So as far as the influence of the Mesa drivers on the performance of the X1k its seems we haven't even scratched the surface. Why not reach the point where we are in line with the most recent standards and do what we do best and that optimize.

Re: SuperTuxKart 0.8.2

Posted: Mon Feb 09, 2015 8:06 pm
by mechanic
Srtest,
Thank You.

I evidently did not have the courage to speak your words, but they reflect my sentiments exactly.

Re: SuperTuxKart 0.8.2

Posted: Tue Feb 10, 2015 9:39 am
by zzd10h
@Xeno74
Is it possible to test this latest version, please ?

Re: SuperTuxKart 0.8.2

Posted: Wed Feb 11, 2015 7:01 am
by xeno74
zzd10h wrote:@Xeno74
Is it possible to test this latest version, please ?
We need Mesa 10.3. And we have to optimize it.

Re: SuperTuxKart 0.8.2

Posted: Wed Feb 11, 2015 8:08 am
by zzd10h
Understood, thanks for reply.

And courage for Mesa

Re: SuperTuxKart 0.8.2

Posted: Thu Feb 12, 2015 12:59 am
by Srtest
I hope for now that the Rumblepad II will finally work in-game. Mesa will be done when its done I guess. The speed comment wasn't about STK in particular. It was more about pushing a certain standard while others will fall into place. As far as benchmarking goes the isolation of variables doesn't always present the way it will work in the actuall app/game while other stuff are also using resources. It is also a give and take situation like the newest kernel which I said was slower than 3.18 all around and now appears to boot faster.