View previous topic :: View next topic |
Author |
Message |
lamer0 n00b
Joined: 31 Mar 2003 Posts: 29
|
Posted: Wed Aug 18, 2004 6:42 am Post subject: Kernel 2.6 sorrows. |
|
|
Alright I tried giving the 2.6 kernel a try and hell I am pretty disappointed with the lack of performance. I have these benchmarks to back me up.
System: 1.74ghz athlon xp(256k cache) / 512 ram(ddr2700) / geforceti4600
Drivers: media-video/nvidia-kernel-1.0.6111
2.4.26 / preempt / low latency patches.
ut2k4 demo benchmark.
17.989159 / 58.136852 / 173.751846 fps -- Score = 57.414062
2.6.7 / udev
ut2k4 demo benchmark.
7.751876 / 33.050697 / 95.413155 fps -- Score = 33.055866
Has anyone else noticed 2.6 being slapped around by 2.4 in benchmarks such as these? From what I can see its pointless to go to 2.6 unless you have a newer athlon64 (bigger cache) w/ sata drives.
I have to hand it to 2.6 for being very responsive on the desktop but 2.4.26 with preempt+low latency patches just does dandy in that department as well.
Just by using kernel 2.4 - ut2k4 is totally playable without needing upgrades just to play it at a good frame rate. I hope this helps some people who are struggling with kernel 2.6 ut2k4 fps issues.
also check this link http://www.2cpu.com/articles/98_1.html. I thought 2.6 was supposed to be the be all & end all to smp systems? Yet again 2.6 bites it. Is Linus and crew just pleasing IBM mainframes where 2.6 really shines in benchmarks?
So what are your opinions? |
|
Back to top |
|
|
DarrenM l33t
Joined: 25 Apr 2002 Posts: 653 Location: Sydney, Australia
|
Posted: Wed Aug 18, 2004 6:57 am Post subject: |
|
|
More likely you've missed something somewhere. I didn't see any performance difference when switching to 2.6
Try doing both your benchmarks without sound to see if that could be the problem. |
|
Back to top |
|
|
lamer0 n00b
Joined: 31 Mar 2003 Posts: 29
|
Posted: Wed Aug 18, 2004 6:38 pm Post subject: |
|
|
What is there to miss to get degraded performance? I am fairly confident its configured right. It 2.6's cpu scheduler which evens out the resources to thin and does not dedicate enough cpu time to processes that need it such as games.
If you do have time try out that ut2k4 benchmark in both kernels. |
|
Back to top |
|
|
LeTene Guru
Joined: 02 Mar 2004 Posts: 348 Location: Ah'll glass ye!
|
Posted: Wed Aug 18, 2004 7:13 pm Post subject: |
|
|
I don't have UT24K, but I can assure you that playing Enemy Territory and Return to Castle Wolfenstein yields much smoother gameplay in 2.6 than 2.4.
My system:
Code: | # uname -a
Linux tux 2.6.8-rc2-love3 #6 Sat Aug 7 15:09:43 BST 2004 i686 AMD Athlon(tm) XP 2500+ AuthenticAMD GNU/Linux
# emerge nvidia-kernel -p
[ebuild R ] media-video/nvidia-kernel-1.0.6111 |
_________________ Docs, Tips & Tricks at the Gentoo Wiki page. |
|
Back to top |
|
|
lamer0 n00b
Joined: 31 Mar 2003 Posts: 29
|
Posted: Wed Aug 18, 2004 10:45 pm Post subject: |
|
|
Please post benchmarks. Lots of things 'feel' smoother but its all psychological. oh and dont forget q3/et are old engines so they should run fast no matter what kernel, but I am fairly sure you should see some fps decrease in 2.6. |
|
Back to top |
|
|
DarrenM l33t
Joined: 25 Apr 2002 Posts: 653 Location: Sydney, Australia
|
Posted: Thu Aug 19, 2004 2:33 am Post subject: |
|
|
I've been coming here for a long time and this is the first time I've ever seen anyone who thought 2.6 was slower than 2.4 by any significant margin. That is why I'm more inclined to beleive that you have mis-configured or mis-tested it somehow. Same video drivers? Same color depth? Same exact settings for fsaa etc? If 2.6 was really that much slower, I assure you it would have been discovered by now and we'd all be running 2.4.
As you can see in the benchmarking ut2004-demo thread, the sound drivers make a big difference. Did you compile your own openal for the testing? Did you recompile it for each kernel? Are you using the same sound drivers in each kernel? That's why I suggest doing your tests again without sound. From memory, you can just move the openal.so out of the ut2004demo/System directory to do this.
Also, you mention task scheduling. If there aren't any other active processes, why should this make a difference? Or are you running the benchmark with seti@home or somehting running in the background?
Post your kernel config in a code block too. Maybe someone will be able to spot the problem. |
|
Back to top |
|
|
|